GLOBAL HEALTH FINANCING INITIATIVE WORKING PAPER 3 | SEPTEMBER 2008

HEALTH SYSTEMS STRENGTHENING VIA PERFORMANCE-BASED AID CREATING INCENTIVES TO PERFORM AND TO MEASURE RESULTS

Rena Eichler Amanda Glassman The Brookings Global Economy and Development working paper series also includes the following titles:

• Wolfensohn Center for Development Working Papers

• Middle East Youth Initiative Working Papers

• Global Health Financing Initiative Working Papers

Learn more at www.brookings.edu/global

Rena Eichler is a consultant to The Brookings Institution, a health economist and health policy ana- lyst. She is owner of Broad Branch Associates, a con- sulting fi rm in international health policy.

Amanda Glassman is a Nonresident Fellow and for- mer Deputy Director of the Global Health Initiative within the Global Economy and Development pro- gram at the Brookings Institution. She is now a Senior Social Development Specialist at the Inter-American Development Bank.

Authors’ Note:

Sincere thanks to the team of the Health Financing Task Force that include David de Ferranti, Maria-Luisa Escobar, Charles Griffi n and Gina Lagomarsino for their comments that guided the questions we asked and the hypotheses we explored. We would like to express deep appreciation to Tristan Blanchard, Loren Becker, and Alexandra Beith for their background research. The paper was strengthened by extremely helpful comments from Maria-Luisa Escobar, Natasha Hsi, Ruth Levine, and Philip Musgrove. Finally, deep thanks to Charlie Griffi n for giving the fi nal push to fi nalize this paper and for his valuable contributions. CONTENTS

Abstract ...... 1

Introduction ...... 2

What is Performance-based Aid for Health and How Does it Differ from Other Forms of Aid? . . . 4

What Can We Learn from ExistingPerformance-based Aid Initiatives? ...... 5 GAVI ISS ...... 5 The Global Fund ...... 10 Poverty Reduction Support Credit (PRSC) ...... 11 “Plan Nacer” in Argentina ...... 12 European Commission Performance-Based Budget Support ...... 14 Strengthening EPI through performance-driven lending in Colombia ...... 14

Discussion: Making Performance-based Aid Work Better ...... 16 Challenges at the donor level: ...... 19

Conclusion ...... 21

Endnotes ...... 22 HEALTH SYSTEMS STRENGTHENING VIA PERFORMANCE-BASED AID CREATING INCENTIVES TO PERFORM AND TO MEASURE RESULTS

Rena Eichler Amanda Glassman

ABSTRACT assistance, strengthening the connection between the fi nancial fl ows and results increases in importance. he new global health partnerships, such as the The experience with performance-based assistance TGlobal Alliance for Vaccines and Immunization for health projects has increased in recent years, pre- (GAVI Alliance), the Global Fund for AIDS, TB, and senting the opportunity for donors to consider some Malaria (GFATM), and UNITAID have fundamentally of the lessons and to build them into their own pro- changed the landscape and scale of funding for global grams. This paper reviews some of the experience and health initiatives since 2000. As many more billions derives lessons for shifting more funding into perfor- of dollars fl ow into these organizations and through mance-based instruments. traditional bilateral and multilateral funders of health

HEALTH SYSTEMS STRENGTHENING VIA PERFORMANCE-BASED AID 1 INTRODUCTION ers, individuals, facilities or multi-site institutions, conditional on their achievement of agreed outputs he new global health partnerships, such as the or outcomes;4 and, (4) assistance to parents or indi- TGAVI Alliance, the Global Fund for AIDS, TB, and vidual patients conditional on their achieving specifi c Malaria (GFATM), and UNITAID have fundamentally results. These approaches are familiar in rich and poor changed the landscape and scale of funding for global countries alike, between one level of government and health initiatives since 2000. The GAVI Alliance and another, in programs that fi nance service provision, in the Global Fund for AIDS, TB and Malaria are among programs that create incentives for changes in behav- the pioneers to shift from paying for expenditures on ior, and between donors and recipients. In addition, inputs to performance based aid (PBA) that links por- governments and donors of all types have increas- tions of their funding to progress on pre-agreed per- ingly tried to improve public sector and grantee ac- formance measures. Global partnerships have been countability through performance measurement and a source of inventive efforts to discover and imple- evaluation.5 ment new sources of revenue to fi nance their opera- tions, including the Innovative Financing Facility for This paper is focused on the second type of perfor- Immunizations (IFFIm), Debt2Heatlh, (PRODUCT)RED, mance-based aid, which accounts for most of the as- the Advance Market Commitment, the airline tax, sistance provided in the health sector. Development and others.1 Commitments by bilateral donors for aid reformers have increasingly called for a shift away global health have also grown significantly during from input fi nancing to performance-based aid (PBA) this period, contributing to an increase in estimated that is conditional on development results.6 Efforts to disbursements of health assistance to developing harmonize aid in poor countries and debt relief efforts countries from $4.6 to $8.5 billion between 2001 and to reduce poverty have also led to the increasing use 2005.2 To transform this unprecedented funding suc- of performance-based aid within policy-based lending cessfully into improved health results for the poor and budget support by the European Commission, the means squarely facing the question of how to assure multilaterals and, through Sector-Wide Approaches that more money will bring more results.3 (SWAp), the bilateral donors. Yet in spite of positive re- ports on the effects of PBA, its use remains the excep- Aid givers have for decades tried to improve the con- tion rather than the rule in development aid for health. nection between outside assistance and the intended results. Essentially four types of performance-based This paper contrasts PBA with other forms of aid, aid are currently used: (1) so-called adjustment lend- analyzes why performance-based approaches remain ing, more recently called development policy lending, a marginal share of health aid, and suggests ways which is funding on a large scale from international to modify current and new arrangements to achieve fi nancial institutions (IFIs), such as the World Bank, greater impact on health results. Our focus is on conditional on implementation of agreed changes in funding from a source external to a country, to either government policy; (2) Assistance that is conditioned a national government or an entity established to on performance in achieving the agreed outputs or implement a strategy that conditions a portion of pay- outcomes, such as through projects or through agree- ment on whether health results are achieved. ments with national governments that condition payment on results; (3) payments to service provid-

2 GLOBAL ECONOMY AND DEVELOPMENT PROGRAM We argue that the effectiveness of PBA rests on fi nd- in this paper suggest that the rules linking payment to ing the right balance between predictable funding and results should be explicit, recipient reports of perfor- a portion at risk for results, and on altering incentives mance should be complemented with external valida- to inspire the system-level changes needed to achieve tion, and “demand driven” technical assistance should better health results. Incentives faced by donor agen- be provided to help countries develop and implement cies and their staff and the constraints they face must strategies to achieve the rewarded results. also be considered. The limited cases of PBA examined

HEALTH SYSTEMS STRENGTHENING VIA PERFORMANCE-BASED AID 3 WHAT IS PERFORMANCE-BASED PBA differs considerably from other forms of aid that AID FOR HEALTH AND HOW DOES pay for inputs rather than the results (outputs or out- IT DIFFER FROM OTHER FORMS OF comes) that the inputs are intended to produce. For AID? example, “investment” loans and grants from the mul- tilateral development banks are structured to fi nance BA ties at least some portion of donor disburse- expenditures on inputs, as are many projects fi nanced Pments to whether recipient countries attain by bilateral donors. The fi nancing is released to the specified results. What gives PBA “teeth” is that recipient based on submission of auditable evidence payment (part or all) is tied to whether the recipi- of expenditure on items agreed to in the loan, grant, ent achieves the targets specifi ed in the terms of the or contractual agreement.7 Although investment loans agreement with the donor. Because recipients are and grants and other forms of project-based aid may accountable for results, and not for enacting policies include an evaluation framework with performance or documenting what they spent, PBA is expected to indicators that are tracked over the life of the proj- catalyze innovation and inspire the efforts needed to ect, barring fi duciary malfeasance or noncompliance improve performance. There are many ways to struc- with other contractual clauses8, disbursements are ture this fi nancial risk. Part of the payment can be made regardless of whether the intended results are transferred as a lump sum, independent of results, achieved. In the health sector, this implies account- and another part tied to attainment of performance ability and management attention to documenting indicators. Another approach is to condition suc- expenditures on purchased items such as buildings, cessive disbursements on adequate achievement of drugs, training, and technical assistance without results in a prior period. Yet another approach is to necessarily demonstrating that these inputs lead to specify the full potential funding envelope and to hold improved health results or systems, such as increased back a proportion of payment until results can be veri- utilization of health services by the poor. It also mini- fi ed. Many approaches can be considered that vary in mizes managers’ fl exibility to alter the mix of inputs to their degree of fi nancial risk, implications for chang- achieve health results. ing behavior, and likelihood that donors actually will enforce the terms.

4 GLOBAL ECONOMY AND DEVELOPMENT PROGRAM WHAT CAN WE LEARN FROM EX- dress concerns about accountability for results that ISTING PERFORMANCE-BASED AID may contribute to increased uptake. INITIATIVES? Table 1 beginning on the next page provides a sum- BA is theoretically appealing, there is some evi- mary of features of the PBA cases that will be detailed Pdence that it can work at the global and country in this section. level, and global appetite for it is growing. There are also considerable challenges to address when structuring performance agreements that provide GAVI ISS incentives that motivate health improving behaviors The GAVI Alliance’s Immunization Services Support where they really matter, at the interface between (ISS) program is an approach to PBA that conditions households and service providers. This section draws funding to recipient governments on increases in the from real-world experiences to identify the elements number of children who have received at least three of performance-based aid most likely to contribute to doses of the vaccine against diphtheria, tetanus and success. pertussis (DTP3). GAVI provides $20, which is based on the average estimated cost of immunizing additional The Paris Declaration on Aid Effectiveness9 calls for children, for each additional child immunized above increased use of budget support and SWAps to reduce the baseline count for the country. Governments have government transaction costs, increase predictable the fl exibility to use the reward funds in any way they funding for basic health services, and to support gov- believe will enhance immunization coverage. A recent ernment-led sector policy and expenditure programs.10 review of six countries found that over 80 percent of Although most SWAps include performance indicators the funds were spent on recurrent costs and the re- relating to health outputs and outcomes, in practice, mainder on capital expenses, primarily vehicles and there are very few examples where program sup- cold chain equipment. Funds were not necessarily port has been interrupted outside of extreme condi- spent in the year they were granted. At the margin, tions relating to corruption or fi duciary issues. In the fl exible cash like this can make a large difference in Bangladesh SWAp, for example, when targets were the capability of program managers to increase the not met, they were extensively discussed in program effectiveness of the immunization program.14 reviews, but disbursements were not interrupted.11 SWAps remain limited in use at only 12 percent of total GAVI ISS payment rules are clear: funding for the aid for health in 2005,12 perhaps surprising given the fi rst two years of a fi ve-to-seven year program is pre- global commitment to reducing transaction costs and dictable and fi xed to enable governments to invest to harmonizing development assistance. Different do- in expanding immunization coverage. The initial two nors provide varying guidance to their staff on decid- payments (“ISS Investment Funds”) are based on the ing amongst aid instruments, but underlying the low estimated number of additional children who will be use of SWAps are concerns related to the inability of immunized over the pre-ISS baseline. Recipient gov- donors to track and earmark spending on donor pri- ernments are required to submit annual progress orities and to attribute results to their particular pro- reports, including audited fi gures of the number of grams.13 Lessons from the following PBA approaches children vaccinated each year. could inform how to better structure SWAps to ad-

HEALTH SYSTEMS STRENGTHENING VIA PERFORMANCE-BASED AID 5 scal bonus between bonus between scal Future phases of of phases Future the loan contingent trig- on meeting established gers in and weighted a performance Structure contract. loan builds on a of performance-based fi national government and municipalities. Colombia-IDB Colombia-IDB Performance-Driven Loan xed and variable and variable xed National Government National Government Economy-wide bud- Economy-wide supportget with a fi Variable component. released component on perfor- based de- on service mance Scores targets. livery assigned. are EC PBA EC ciary. National then Government, National to from Level Provincial Future phases of the of phases Future on loan contingent triggers. meeting receive Provinces potential 60% of each for payment and poor woman in child who enrolls pro- an insurance and 40% is gram of attainment tied to tar- 10 performance per target). (4% gets province Potential is $10 per budget benefi Argentina-World Argentina-World Plan Nacer Bank ed in the National Government Future credits contin- credits Future trig- gent on meeting specifi gers Reduction Poverty Paper Strategy (PRSP). Response be cancellation, can or delay reduction, met. are until terms c indicators indicators c per- by uenced Principal Recipient Recipient Principal gov- (national (PR) ernment or civil soci- group) ety Subsequent funding Subsequent is infl formance on country on country formance specifi and targets. and targets. Yes Yes No Yes No Yes National Government Yes [(# additional chil- over with DPT3 dren )* No baseline pre-GAVI $20]; conditional Payment over score on DQA 80%. No Yes Yes Yes Results Results independently validated? Recipient Explicit Payment Explicit Payment Formula? Incentive Payment Mechanism Funding AgentFunding ISS GAVI GFATM PRSC Table 1: Features of Selected PBA Approaches PBA Selected of 1: Features Table

6 GLOBAL ECONOMY AND DEVELOPMENT PROGRAM nancing provided provided nancing TA programmed programmed TA within the package of fi under the loan. Adapted Data Quality Quality Data Adapted independent by Audit by contracted entity national government reported validate to results immunization sys- and information performance. tem Colombia-IDB Colombia-IDB Performance-Driven Loan Not directly, but TA but TA Not directly, with be procured can funds. EC Performance based based Performance na- by on reports tional government vali- without external on dation that draw information existing and other systems sources. available EC PBA EC cation by by cation TA from World Bank Bank World from TA con- and local staff sultants. prov- to available TA national from inces is “demand TA level. and reported driven” be an important to feature. In APL, performance In APL, performance “rolled on triggers perfor- from up” the pro- of mance at the province gram level. report Provinces and enrollment on tar- progress Verifi gets. and external internal for Penalty auditors. enroll- misreporting ment is 20%. Argentina-World Argentina-World Plan Nacer Bank Not directly, but Not directly, use can country as- procure funds to sistance. Performance based based Performance na- by on reports tional government vali- without external on dation that draw information existing and other systems sources. available su- staff Bank World pervises. cation by Local Local by cation TA from Stop TB Stop from TA may Partnership to contributed have performance better in TB programs or than in HIV/AIDS does malaria. GFATM TA, direct not provide funds can but grant purchase to be used Development it. TA provide partners implement GFATM to programs. Rating system system Rating PR- of comprised progress, reported verifi (LFA), Agent Fund mitigat- for adjusted ing circumstances. TA to implement the to TA was RED strategy concur- provided Not provided rently. but GAVI, by directly by be used funds can procure to country TA. Data Quality Audit Audit Quality Data indepen- by (DQA) con- dent entity to GAVI by tracted reported validate results. DPT3 sment ap- Role of Technical Technical of Role Assistance Performance Performance asses proach Funding AgentFunding ISS GAVI GFATM PRSC Table 1: Features of Selected PBA Approaches (cont.) Approaches PBA Selected of 1: Features Table

HEALTH SYSTEMS STRENGTHENING VIA PERFORMANCE-BASED AID 7 Financial risk is introduced in the next three to fi ve is that these programs will become even more focused years when fi nancing switches to “reward funding.” than today on knowing what they are doing and why A data quality requirement is introduced. ISS funds they are succeeding or not. can be earned as long as an independent data quality audit (DQA) achieves at least a score of 80 percent. If An evaluation of the ISS published in 2006, based on the DQA score is below 80 percent, the program is dis- country-level data from 1995 to 2004 (GAVI started in qualifi ed but can strengthen data reporting systems 1999), found that when controlling for other factors, and receive ISS reward funds in the future. In practice, GAVI’s spending had a statistically signifi cant positive many countries have been allowed to continue relying impact on DPT3 coverage in countries with baseline on estimates in the third year because of inadequate coverage 65 percent or less at the start of funding. reporting systems, but by the fourth year, grants must Higher per capita GDP and greater political stabil- be based on audited numbers. In the long run, all ISS ity also contributes to better performance. However, grants in year x would be based on performance in at coverage levels above 65 percent, the statistical year x-1 against the increment over year x-2. The re- model explained improvements poorly.15 Adding one ward is for program performance in the previous year. more year of data and including actual disbursements In 2005, 53 countries received ISS support, of which (rather than the grant amount) as an explanatory vari- 12 had “graduated” to full PBA or reward payments. able allowed a subsequent evaluation to show an im- While GAVI’s support for ISS will eventually evaporate pact for the over-65 percent coverage group.16 as recipients succeed in getting targeted children vac- cinated, GAVI will then be able to shift its focus to ex- It is important to note that as GAVI continues, the panding coverage of other vaccines. ability to evaluate its impact should improve because it will be trying to introduce new vaccines where they It is worth noting some of the conditions that make are not used today, and its PBA approach generates it possible for GAVI-ISS funding to be managed this audited data needed to evaluate its impact, almost way. First, ISS would add probably not more than 15 on a real time basis. Non-PBA approaches typically percent to the resources available for an immuniza- do not generate performance data that would allow tion program, and there is good evidence that it has evaluation. not crowded out other sources of revenue for immu- nizations. Thus removing ISS funds due to bad perfor- Some possible lessons are beginning to emerge from mance will not kill an immunization program, and the the GAVI ISS experience that can inform other PBA initial two years’ investment can only help. Yet GAVI approaches, although these are observations by provides enough money to make the effort worth- practitioners that would require evaluation: 1) It is while. Second, funding is based on counting a specifi c important to ensure that recipients from the national set of immunizations, which is feasible. Performance to local level understand how the performance pay- is thus measurable against a simple metric, notwith- ment will be awarded and associated incentives; 2) standing caveats surrounding the data. The reward Linking payment to audited information is a catalyst system raises the value of information substantially to strengthen health information systems and their – including both the incentive to manipulate it and the use; and, 3) Technical assistance appears to be an im- incentive to assure its accuracy – but the overall result portant contributor to success.

8 GLOBAL ECONOMY AND DEVELOPMENT PROGRAM An August 2004 study of the impact of ISS on im- cause the donor cannot withdraw funding without munization rates17,18 found that limited understanding causing a collapse. In the case of GAVI-ISS, expan- at sub-national levels of how reward system worked sion can commence again once problems in the pro- gram are corrected. impeded results. A more recent evaluation19 suggests that more years of ISS experience has contributed One of the great attractions of performance-based to better understanding of ISS and associated incen- programs is that numbers are routinely generated tives at all levels. In addition, the importance of a that help both donors and recipients understand strong DQA score provides incentives to countries to dimensions of the problem they are trying to solve. strengthen and maintain the information and moni- GAVI has learned some lessons from its initial experi- toring system. A number of countries reported using ence that may cause it to revise elements of the pro- ISS funds to implement the “RED” (Reaching Every gram. The following points come from the evaluation 20 District) program of WHO/UNICEF that is being im- of GAVI’s fi rst fi ve years21: plemented alongside GAVI. RED has fi ve components that are aligned with the goals of GAVI ISS and pro- • Because reward payments are based on the num- vides technical support to implement them. Red helps ber of additional immunizations, countries with to ensure that incentives provided at the national fast population growth tend to do better than those with lower rates. GAVI has supplemented its base level trickle down to affect the service-delivery level formula to adjust for countries with declining birth and is a promising complement to the GAVI incentive rates. structure. • Programs have an incentive to over-estimate the The following design elements of the GAVI ISS model number of children to be vaccinated during the fi rst two years, to increase the initial investment pay- offer lessons to inform the design of other PBA ap- ments. Infl ating these numbers will result, however, proaches: in a lower or no reward payment in year 3 because the performance targets cannot be reached. • The performance indicator is clear and the perfor- mance target is chosen by the recipient at the initial • Some of the countries most in need, such as post stage. confl ict and some of the poorest countries, have been least successful in meeting the criteria for re- • The importance of reliable results reporting has ward payments. caused the funder to go to great lengths to validate results and improve data integrity. • Experience in ISS countries suggests that the cost of additional immunized children increases ex- • A combination of reliable funding initially, combined ponentially at 80 percent coverage, where it hits with performance-contingent funding in the future, an average of $40 per child, rising to over $80 enables immunization programs to gradually move per child at 90 percent coverage. Cost-effective- to a performance-based system and learn the im- ness of the program may drop off quickly as costs plications. rise and herd immunity is achieved. Governments may no longer fi nd the $20 ISS subsidy attractive. • Because GAVI-ISS provides funds at the margin for Therefore, adjusting the subsidy, targeting it across expansion, it does not endanger the entire program and within countries, and other narrower factors – only the expansion is put at risk by poor perfor- may make sense as part of the reward formula – all mance. Often performance-based funding fails be- requiring much fi ner detail in performance mea-

HEALTH SYSTEMS STRENGTHENING VIA PERFORMANCE-BASED AID 9 surement and a complication of GAVI’s attractively Comparison of grant performance across the three simple and easy-to-understand metric. diseases supported by the Global Fund provides some insight into needed complements to implementing The Global Fund performance-based funding that include: 1) effective technical assistance; 2) engaging with stakehold- The Global Fund to Fight AIDS, Tuberculosis and ers; and, 3) strong health management information Malaria is another global initiative that incorporates systems. In a 2006 report, the Global Fund suggests a performance-based element into its funding opera- that better performance among TB grants is due in tions. Approved grants include performance indica- part to technical assistance provided by the Stop tors and targets, along with the condition that phase TB Partnership that includes management, procure- two will be approved if performance is adequate. While ment and targeted implementation support to scale- performance is measured using a rating comprised of: up effective programs. The Global Fund’s emphasis scores on self-reported progress and expenditure re- on including many stakeholders helps to ensure that ports submitted by Principal Recipients; verifi cation funding reaches the level in the system that is the in- by the Local Fund Agent of results and approved ex- terface between communities and service providers. penditures; and contextual information and mitigating In contrast, programs that only work with national circumstances that may affect performance, rules for governments are criticized for failing to reach the how phase two funding decisions are made are not service-delivery and use level. To strengthen monitor- explicit or public.22,23 In practice, better scoring grants ing and reporting, the Global Fund recommends that receive a higher proportion of disbursement requests recipients spend fi ve to 10 percent of their grant funds than those with lower scores, though few Global Fund to enhance health management information and mon- grants have failed to receive any support for their itoring systems. second phase.24 In addition, concerns that the per- formance-based funding approach of the Global Fund The design of Global Fund program fi nancing agree- might penalize poor countries have not been realized; ments offers lessons for other PBA approaches. poor countries have had lower budget reductions Customizing performance measures and targets to in phase two than other recipients, largely because fit a specific country context appears to motivate performance is assessed relative to country-specifi c both countries that begin with a low baseline as well rather than absolute targets. as the already good performers, as evidenced by the observation that grants in poor countries and fragile A 2006 report that examined 140 grants that were states perform as well as in higher income and more at least 18 months old shows that tuberculosis grants stable contexts. The lack of explicit rules about how met all performance targets, HIV/AIDS grants met 87.5 decisions about future phases of grant funding will percent and malaria grants met 60 percent of targets, be made may be both positive and negative: on the on average.25 It is not possible, however, to identify the positive side, grant managers have fl exibility to con- distinct contribution of the Global Fund to attainment sider country specifi c contexts while, on the negative of performance targets because multiple programs side, lack of clear rules may send weak signals about and multiple sources of funding often simultaneously how performance will be rewarded resulting in weaker exist. behavior and system changes than might have been

10 GLOBAL ECONOMY AND DEVELOPMENT PROGRAM generated by rules that show a clear link between trix of PRSP-related conditions and include policy and results and future disbursements. Measuring and veri- administrative actions as well as output and outcome fying performance is more challenging in Global Fund targets (i.e. percentage of pregnant women receiv- grants than in immunization programs supported by ing pre-natal care). There is no explicit weighting, al- GAVI because the diseases are complex and the range though critical conditions are sometimes highlighted of recipients more diverse. Given this complexity, the in project documents as decisive factors for disburse- Global Fund approach to validating performance is a ment, and there has been controversy over measure- useful model to build on. ment of outputs and outcomes. Each successive PRSC comes with its own set of benchmarks, which may or may not be related to the previous credit. Approval Poverty Reduction Support Credit of subsequent credits in a series is contingent upon (PRSC) completion of a set of prior actions laid out in a Letter The Poverty Reduction Support Credit (PRSC) is a of Development Policy (LDP) and matrix of perfor- lending and granting instrument26 created by the mance benchmarks.29 Failure to progress as expected World Bank to provide budget support to governments can result in cancellation or suspension of the PRSC. to develop policies and implement institutional re- However, a graduated response program also allows forms aimed at achieving goals laid out in a country’s for the Bank to choose to adapt the program, reduce Poverty Reduction Strategy Paper (PRSP) and meets funding in the subsequent credit, or delay the subse- the defi nition of PBA used in this paper. PRSC design quent credit until the terms are met.30 refl ects the evolution of thinking among donors and academics in the 1990s that aid is more effective if In the majority of PRSC cases, the conditions for “owned” by countries, geared toward comprehensive, moving forward are met and programs progress as long-term development, focused on measurable per- scheduled. Seven countries have had at least two formance, and harmonized across all donors working successive tranches released, four have had three, in the country.27 PRSC outlays are concessional loans and two have received a fourth disbursement.31 As of or grants, typically taking the form of three annual, February 2005, only 5 percent of countries failed to single-tranche credits. In a three-credit series, the meet established benchmarks; in all but one of these first disbursement typically supports basic needs the Bank opted to take a graduated approach rather such as health, education, and water and sanitation than discontinuing the program entirely. In Guyana, projects. The subsequent disbursements, designed to the Bank opted to abandon its existing PRSC program be triggered by attainment of performance targets in due to political and institutional challenges on the the fi rst credit, are expected to fund rural develop- ground and is instead planning to negotiate a new ment, private-sector development, and post-primary program that is more realistic. The only instance in education.28 which the Bank gave less money than it committed initially was in Ethiopia, which failed to complete one Performance indicators, known as prior actions or of the original prior actions suffi ciently and had its triggers, vary from program to program, and are fi rst tranche cut accordingly. The Bank also delayed negotiated between the government and the Bank. follow-on PRSCs in Tanzania, Nepal, and Sri Lanka due Health targets usually are one section of a larger ma- to inadequate performance toward the benchmarks.32

HEALTH SYSTEMS STRENGTHENING VIA PERFORMANCE-BASED AID 11 While PRSCs are a form of PBA, fi nancial incentives sign an Umbrella Agreement and Annual Performance to perform are weak for both the Bank and recipient. Agreement with the national Ministry of Health that Part of the reason is that disbursements and their specifi es enrollment and performance targets for 10 size are defi ned and programmed as part of the an- tracer maternal and child health and program indi- nual macroeconomic projections agreed upon with cators.36 Provinces negotiate quarterly targets for the IMF, and non-disbursement would imply fi scal im- each “tracer,” expressed as a function of the total balances that are unpalatable. A strength of PRSCs is eligible population, with the only restriction being that they are reported to have increased emphasis on that targets cannot be below those set in the previ- measurement and a focus on performance on agreed ous period. World Bank offi cials note that although benchmarks. 33 provinces were encouraged to set targets much lower than the performance they had been reporting to the World Health Organization, many of them still failed “Plan Nacer” in Argentina to reach these lower targets, revealing that previously The Government of Argentina, with support from a reported performance could not be substantiated World Bank Adaptable Program Loan (APL), began with robust data verifi cation.37 In addition, there is a implementing a program in 2004 called “Plan Nacer” national Project Implementation Unit that provides to reduce the infant and maternal mortality rate, in- technical support to build capacity at both the na- crease effi ciency, and enhance the focus on results by tional and provincial levels to implement and manage providing insurance coverage and access to maternal the process. Included are investments in information and child health services for poor women and chil- systems, fi nancial and human resource management dren.34,35 Of some interest to effective PBA is the APL systems, and support to “streamline” the regulatory itself, which conditions subsequent phases of the loan and planning capacity of the national Ministry of on meeting triggers established in previous phases. Health. Funding is also included for communication The most innovative aspect of this program, however, and community outreach to make target populations is the performance-based transfers from the national aware of this new program, understand their rights, to provincial levels of government and the institu- and encourage enrollment. tional changes, increases in coverage, and utilization by poor women and children that result. Performance-based transfers to provinces are linked to enrollment and attainment of the 10 performance The national Ministry of Health has overall respon- targets. Of the monthly payment of $10 per person/ sibility for meeting performance targets in the APL, per month, half is funded with government funds while provinces have the operational responsibility and the other half by the Bank loan. Of the Bank for implementing Plan Nacer and reaching province- loan portion, 60 percent is transferred based on the level performance targets. This structure enables the numbers of poor women and children enrolled in the central government to have infl uence over health in scheme and 40 percent is linked to evidence on 10 a decentralized context. This decentralized arrange- performance targets, with achievement of each target ment relies on the creation of health purchasers at linked to 4 percent (4 * 10 = 40 percent).38 To make the provincial level to negotiate with and pay provid- Provinces accountable for the quality of enrollment ers for a defi ned list of services. Eligible provinces information, penalties are imposed that equal the

12 GLOBAL ECONOMY AND DEVELOPMENT PROGRAM amount transferred to cover an unauthorized person spending, are under public scrutiny by journalists and plus an additional 20 percent. Enrollment is verifi ed civil society organizations that monitor spending and by crosschecking registers of other social insurance attainment of targets closely.42 schemes. Provinces collect output information for This PBA approach enabled the central gov- each tracer, following explicit guidelines from the na- ernment to have infl uence over health in the- tional program (and assistance to build the required context of provinces that are decentralized, information, verifi cation, and reporting systems). A while still transferring control over resources sworn statement is signed by the provincial authority and local level decisions to the provinces. that reported tracers have been achieved. Registers of enrollees are used to audit clinical information. An Plan Nacer is an innovative PBA program that expands external auditor examines a sample of registrations coverage and access to services by poor women and (enrollment and tracers) for verifi cation. The National children by subsidizing premiums in a targeted social Ministry also conducts a concurrent audit using inter- insurance program, specifying the rules for transfers nal staff. of funds from the national to the provincial level, and by creating health purchasers at the provincial level By May 2006, 46 percent of the eligible population who negotiate with and pay providers for a defi ned in nine provinces (380,000 enrollees) had been en- list of services. One lesson is that there are many de- rolled. Structural changes in health fi nancing and tails in the design and implementation and that much health service management had been introduced attention is needed to each step if this approach is to including: implementation of a health services pur- be replicated in another setting. The details to vali- chasing unit; establishment of performance-based date data and hold provinces accountable are critical. contracting with service providers that includes an This PBA approach enabled the central government to output-based payment mechanism; guaranteed ac- have infl uence over health in the context of provinces cess to a defi ned list of health care services for the that are decentralized, while still transferring control enrolled population; strengthened health informa- over resources and local level decisions to the prov- tion systems; improved population identification inces. It also changed the demand for technical as- and enrollment systems; establishment of a system sistance from supply driven to more effective demand of billing for health services to social insurers; and driven when provinces demanded technical assistance implementation of a strategy to extend coverage to to improve information systems, expand enrollment, identifi ed at-risk populations.39, 40,41 and contract and pay providers. The demand for tech- nical help increased when it became clear through Another aspect of the program that is important is concurrent audits that provinces were not fully on that the Ministry of Health maintains information track and were at risk for not receiving the full fi scal on a public Web site (http://www.nacer.gov.ar/index. transfers. This approach is viewed as having catalyzed asp) showing the performance of provinces and the enduring changes in the health system by profoundly amount of funds disbursed to providers to cover poor changing the roles of key actors and shifting the focus women and children. This has proven to be an impe- to results. tus to mobilize the process as provinces seen to have a large sum in the purchasing account, but limited

HEALTH SYSTEMS STRENGTHENING VIA PERFORMANCE-BASED AID 13 European Commission Performance- whether technical assistance and policy advice might Based Budget Support help governments assess bottlenecks and adjust poli- cies and practices to achieve goals.46 Since 1999, the European Commission has imple- mented a form of PBA that provides budget support The EC relies on government-provided reports of at- through a mechanism that combines a fi xed and condi- tainment of performance targets, without verifi cation tional tranche of aid.43 The fi xed tranche is structured through an audit process. Some indication that re- as “all or nothing” fi nancing and depends on compli- ported results are not reliable comes from the obser- ance with the IMF program. The conditional tranche vation that inconsistent offi cial statistics are reported is released based on reports by the country that pre- to different agencies. For example, offi cial vaccination viously agreed annual service-delivery targets have rates reported to the EC in Burkina Faso do not corre- been reached. Portions of the budget support pack- spond to those reported to the WHO47, but this may be age may be allocated to different sectors, each with attributable to different time periods. their own fi xed and variable components. Within each sector-specifi c tranche, weights may be assigned to It is also possible that the magnitude of the fi nancial each target enabling calculation of a score that deter- incentive is not suffi cient to motivate improved per- mines the share of the variable tranche to be released. formance. The difference in disbursements between Partial credit is given when progress is observed but a the poorest performers and the best performers is target has not been achieved. Guidelines recommend only 20 percent of the already small variable tranche a prior assessment of data quality, but there is no spe- on average or 8 percent of the total amount of the cifi c guidance and it is not known whether these were program, implying a fairly small incentive in monetary conducted.44 The average program size was 78 million terms for countries that perform particularly well. Euros in 2004, with conditional tranches averaging 12 This payment is split amongst several sub-sectors million Euros.45 making the fi nancial incentive for a specifi c sector such as health even smaller. The fact that funds are Support is usually three years in duration and is transferred to the Ministry of Finance rather than the framed within the poverty reduction strategy, relying line ministry or executor may also weaken incentives on indicators and targets from the PRSP whenever to improve performance. possible. The most common indicators used in the health sector are vaccination coverage, assisted deliv- eries, and health service utilization. As with most na- Strengthening EPI through perfor- tional PBA schemes, it is diffi cult to attribute changes mance-driven lending in Colombia in results fully to the EC PBA scheme because multiple In 2003, the Inter-American Development Bank other programs operate simultaneously. Given this launched an experimental program of performance- caveat, the EC reports that roughly half of the credits driven loans (PDL) intended to shift recipient and studied showed full or partial attainment of health Bank focus to the production of and accountability for targets in the 1999-2004 period. In Burkina Faso, for development outcomes.48 The health sector was the example, the EC program reported a modestly improv- fi rst to pick up the PDL, perhaps due to the perceived ing trend in levels of vaccination but no clear improve- “measurability” of health outputs and outcomes in ment in health services utilization. The EC is reviewing

14 GLOBAL ECONOMY AND DEVELOPMENT PROGRAM the medium-term, and operations were approved in brace a simple money for results model. Because of Colombia49, Honduras50 and Nicaragua51. Each op- a perception that an “investment” means an “invest- eration handled the performance contract differently; ment loan,” the IDB Board of Directors sets annual this paper will focus on the model used in Colombia. limits on policy-based lending, or money that can be used for budget support or debt servicing. To avoid The Colombian PDL was developed to strengthen the this limit, the PDL was classifi ed as an “investment Expanded Program on Immunization (EPI).52 While still loan” and tranche sizes were determined based on high compared to world averages, Colombia EPI perfor- actual expenditures during the inter-tranche period, mance had suffered setbacks during a major economic rather than the quality of performance achieved, for recession and as a result of programmatic fragmenta- example. This requirement forces government to ac- tion related to decentralization and the consequent count for actual expenditures to the Bank as they distribution of EPI tasks among different levels of gov- would in a regular investment loan, and reduces the ernment, health insurers and public hospitals. incentives to participate in the experiment since the country must undertake both a new set of evaluation The program intended to align incentives to improve activities and account for and submit receipts to the vaccination rates in the poorest municipalities in Bank ex-post. Colombia by disbursing all-or-nothing tranches to the national government if a pre-defi ned threshold of an- Another issue relates to the creation of an external nual immunization performance was met and verifi ed audit, while recipients must still build and maintain by an independent audit modeled on the GAVI data their own monitoring and evaluation activities in or- quality audit. A performance contract was agreed der to understand if the program is on track to meet containing municipality-specifi c performance indica- performance targets. In a resource-constrained envi- tors and annual goals for the four-year program. Each ronment, governments frequently object to “paying indicator was weighted, an overall performance score twice” for performance monitoring, since the costs of was calculated and disbursement triggers associated the audit are built into the loans. One of the recom- with scores identified up front. Further, incentives mendations of PDL teams at the IDB is that future were to cascade down from the national government external audits be fi nanced by grant monies directly to sub-national governments via a fiscal bonus to by the IDB. municipalities nationwide that met the immunization threshold. Technical assistance and information sys- The Colombia PDL follows best practice on perfor- tem strengthening is built into the program and pro- mance contracting by developing an explicit perfor- vided by the central government or contractors to the mance contract with fi nancial incentives that cascade central government. The program has been in imple- through levels of government and executors. Yet PDL mentation for two years and is disbursing according design decisions taken by the Board to meet bureau- to schedule, indicating that threshold performance cratic requirements have complicated implementation scores have been met as planned.53 and made the instrument less responsive than it could be to performance. A major problem has been the determination of the size of the tranches. The PDL program did not em-

HEALTH SYSTEMS STRENGTHENING VIA PERFORMANCE-BASED AID 15 DISCUSSION: MAKING PERFOR- the capacity of a health system to deliver results by MANCE-BASED AID WORK BETTER fundamentally changing the roles and relationships between actors and entities and the incentives that BA approaches such as GAVI ISS and the Global drive them. Altered relationships can occur between PFund suggest that linking aid to performance national and sub-national government entities, as is most effective when viewed as a strategy to demonstrated by Plan Nacer in Argentina and EPI in strengthen the capacity of national and local institu- Colombia, as well as between service providers and tions to deliver results, as well as a way to hold gov- the communities they serve. ernments accountable. If thoughtfully designed and implemented, PBA can catalyze changes that enhance

Box 1. Summary of Design Elements of Effective PBA Terms of Payment • Majority (more than 75 percent?) reliable and predictable and not linked to performance. • A smaller portion (less than 25 percent?) at risk and linked to attainment of explicit and predetermined performance targets. • Explicit payment formula.

Indicators and Targets • Small number (less than 10?). • Targets set relative to country’s own baselines. • Outputs (e.g. immunizations, prenatal care visits according to norms, institutional deliveries) that are closely correlated with health outcomes, complemented with equity measures.

Health Information • Health Information Systems will likely need to be strengthened – from lowest level in the system up to national level. • Provider level “self” reports should roll up to higher levels. • A random audit system needed to validate data. • Considerable penalties for data discrepancies. Careful attention to incentives from national to sub-national levels of government to interface between ser- vice providers and households • To achieve results, the benefi ts of potential fi nancial rewards need to trickle down to affect the actions of providers and households on the ground.

Management, payment, monitoring, and evaluation at the country level • Signifi cant management and administrative systems needed to implement PBA within a country to establish performance targets, measure, monitor, reward, evaluate, and revise. • Technical Assistance • Once recipients are paid based on results, they will ask for technical assistance to develop and implement strategies to achieve the results.

16 GLOBAL ECONOMY AND DEVELOPMENT PROGRAM Features of the cases described in this paper suggest for how payment will depend on results should be a number of lessons about how PBA should be struc- explicitly defi ned, as in the case of GAVI ISS and Plan tured to be most effective. What follows are a number Nacer. Making the rules explicit strengthen the value of design, process and implementation suggestions to of the incentives as recipients are clear about what strengthen PBA, also summarized in Box 1. they may lose if performance targets aren’t reached. Even with explicit rules, as in the case of GAVI ISS, it Payment should be part fixed and reliable and a may take several years for an organization to internal- smaller portion conditional on results. The limited ize and understand how the incentive payment works. evidence presented suggests that terms of payment Performance indicators, baselines and targets should in PBA schemes should be structured as one part a be clear; a mechanism to validate results specifi ed; predictable and reliable stream, with a smaller portion and penalties for discrepancies considerable. at risk for achieving performance. The fi xed stream should be provided for enough time to encourage the A small number of indicators is best. Improvement efforts and investments needed to strengthen health targets should be set relative to a country’s own systems. The portion at risk should provide the incen- baseline rather than to achieve some absolute per- tive to focus on the results that both the donor and formance level. Evidence from cases of performance the recipient value. based incentives between payers and developing country health facilities suggests that recipients In contrast, an all-or-nothing approach to approval fi nd it hard to improve results on more than ten in- of future disbursements appears to make the perfor- dicators54; this guidance may be transferable to PBA mance-based element of the program a challenge to agreements. The approach used by both the Global enforce as evidenced by the preference for a gradu- Fund and GAVI ISS, of rewarding improvements ated response to poor performance in PRSCs. All-or- against a country’s own baseline, provides both low nothing PBA fi nancing schemes can limit leverage and performing countries and those starting from a higher create a tension between the desire to assure predict- level the incentives to improve. able fi nancing for health services and the need to en- force performance requirements. Improvements in health outcomes such as reductions in child and maternal mortality rates are the ultimate Experimentation and evaluation is necessary to de- measures of performance. However, since changes in termine the size of the incentive needed to improve health outcomes take longer than increases in utiliza- performance cost-effectively, and this is likely to vary tion of priority services, and it is diffi cult to attribute depending on the context. However, the model of com- changes in health outcomes purely to health interven- bining a fi xed tranche of fi nancing for predictability tions, outputs that are highly correlated with desired with a variable tranche linked to performance, used outcomes can be a close second best. Examples in- by GAVI ISS and the EC, seems likely to produce better clude: immunization rates, proportion of people sleep- results than all-or-nothing fi nancing structures. ing under malaria bed nets, and percentage of TB patients who complete the full course of treatment. Explicit rules that specify how payment will be linked Process measures are a distant third best but may be to results strengthen the impact of PBA. The rules considered in cases where capacity needs to be en-

HEALTH SYSTEMS STRENGTHENING VIA PERFORMANCE-BASED AID 17 hanced before improvements in outputs are even pos- aid from accountability for spending on every input, sible. Examples of short-term process measures may investing instead in the information needed for coun- include: training health workers and posting them in tries to manage for results and fi nanciers to verify rural areas and strengthening the drug and supply that results occurred. management systems so that the needed inputs are in place. National level indicators that specify broad For PBA to work, incentives need to affect actions access and distributional targets such as: “proportion at all levels – from the national level to the local in- of districts with at least 80 percent of fully immunized terface between service providers and households. children,” may also be considered. The experience of GAVI ISS highlights the importance of ensuring that PBA incentives are understood at all Health Information Systems will need to be strength- levels of the health system and that they trickle down ened – from the lowest level in the system up to the to affect the behavior of implementers. The 2004 national level. Once payment is linked to results, a evaluation found that the ISS performance incentives robust system to report on and validate results in were not well understood by most national and dis- needed. All the examples discussed in this paper use trict offi cials. As a result, offi cials did not introduce performance data reported by the recipient, but only incentives to motivate and reward service providers GAVI ISS, the Global Fund, and Plan Nacer incorporate for increasing DTP3 coverage. Such fi ndings argue an external audit to validate results. A critical com- for a concerted effort to inform staff of incentives ponent of effective PBA is investment in information and, where possible, to pass those incentives directly systems to enable countries to track performance, to implementers and households. Implementing such use data to guide management interventions, and to schemes would maintain the valued fl exibility in the monitor unintended consequences. use of ISS funds while assuring that ISS monies stimu- late the desired results. Because PBA requires independently verifi able mea- surement of outputs and outcomes, they may not Complex government structures at the national be perceived as “light touch” by donors wishing to and sub-national levels also present an important reduce the burden on countries of managing inter- challenge for designers and implementers of PBA national aid. While this element of PBA does confl ict programs. Rewarding performance at a national gov- with the fears of many in global health about cre- ernment level does not necessarily cascade down to ating new rules and requirements associated with infl uence the actions of front line providers or the fi nancing, verifying performance may replace other health-related behaviors of the population unless systems of accountability for how money is spent on careful attention is paid to incentives at all levels and inputs, resulting in a stronger health system in the to the obstacles that prevent the potential benefi ts longer run. The GAVI DQA, for example, has gener- of performance-based incentives from influencing ated demand for information at the country level, as actions where they really matter. These incentive ef- well as the opportunity for technical assistance to fects occur not only within the institutional contexts correct management information system shortcom- of government entities and service providers, but ings and improve results. Instead of under-designing also at the household level, where individuals and PBA, light touch should be achieved by de-linking families decide whether to practice healthy behaviors,

18 GLOBAL ECONOMY AND DEVELOPMENT PROGRAM seek treatment, and follow medical advice. Helping petite for technical assistance can increase when PBA governments understand existing incentives, assess recipients request help to strengthen the systems constraints, identify bottlenecks, and develop action needed to achieve rewarded results. plans to achieve results may contribute to ensuring that aid tied to results is effective. Challenges at the donor level: Attention needed to management and administra- Donors often provide aid for political and strategic tive systems. Existing government processes may reasons in addition to humanitarian ones, sometimes be incompatible with performance-based aid. For making it diffi cult to condition aid on results and act example, budgeting from national to local levels of on threats to withhold aid for poor results. Although governments and to facilities is most often based on disbursements are conditioned on policy reforms or the costs of inputs determined at the start of a fi scal development results, policy-based loans are used to year. PBA introduces an element of uncertainty and fi ll a government’s fi scal gap, usually as programmed a budget category that can only be realized after the in an IMF agreement, and withholding funding could performance period is over and results are verifi ed. create undesirable macroeconomic and fiscal side Careful attention to what is feasible within current effects. Even looking solely at the health objectives, fi scal structures and to introducing needed changes there is an uneasy balance between assuring predict- will be necessarily part of any PBA approach. In addi- ability of funding for humanitarian and ethical reasons tion, systems to transfer funds, manage performance and enforcing the performance mandate. Further, for agreements, and monitor information will all need to countries that borrow from the multilateral devel- be established. Governments should be encouraged to opment banks, it may be problematic to reserve a incorporate systems to monitor desired impact, iden- portion of lending, on which recipients are paying ad- tify unintended effects of any systems to alter incen- ministration fees or interest, pending the achievement tives, and to make revisions to what will continue to be of performance. a dynamic and evolving strategy. Beyond the questions of whether and how to with- Technical assistance is necessary and PBA increases hold funding, donors also face signifi cant administra- the appetite for it. Our examination of prominent PBA tive hurdles to implementing PBA. For bilaterals, the programs supports growing external evidence that multi-year commitment required for effective perfor- for these capacity enhancing benefi ts to be realized, mance-based aid may be in confl ict with the one-year technical assistance needs to be provided as a com- planning and budgeting cycles determined by the plement to the changed incentives to help countries overall national budget process. Further, the adminis- identify constraints to strong performance and strate- trative processes (i.e., procurement and contracting) gies to improve it. Radelet suggests that for the Global that have been institutionalized to award aid are often Fund to be effective, technical assistance is needed in in confl ict with linking aid to results because a desire addition to money to combat HIV/AIDS, tuberculosis, for transparency and accountability to fi nanciers has and malaria.55 This observation echoes what is found led to systems that follow inputs rather than outputs in all of the cases described in this paper. Argentina’s or outcomes. Shifting from funding inputs to aid that experience with Plan Nacer demonstrates how the ap- rewards performance may imply a change in the way

HEALTH SYSTEMS STRENGTHENING VIA PERFORMANCE-BASED AID 19 donor agencies are organized and staffed, as the new lateral development banks or global initiatives. Some administrative procedures involved may require a implementation constraints may be alleviated if new change in staff skill mix, and a change in the incen- lending and granting instruments are explicitly de- tives staff face. signed to deliver PBA and assure accountability for results that bilateral and other donors can adapt and Where donors face administrative or procurement participate in. constraints to implementing PBA, they may be able to contribute to pooled programs managed by multi-

20 GLOBAL ECONOMY AND DEVELOPMENT PROGRAM CONCLUSION desired results through PBA. Further, impediments at the donor level may prevent many PBA schemes from hrough its ability to demonstrate value added ever getting off the ground at all. T(the additional performance generated condi- tioning part of funding on attainment of results), PBA More work is needed to determine the preconditions has the potential to overcome common limitations that predict a country’s ability to succeed under PBA associated with input-based fi nancing for essential arrangements, as well as when governments or other health services in developing countries. Incorporating representative entities should be recipients. Some effective PBA features into SWAps and budget support preconditions might include basic elements of an programs, considered best practice in aid effective- information system, promise to alter the national to ness by those concerned with the transaction costs local budget process to allow innovation and the fl ex- imposed on countries by multiple donor programs ibility to respond to incentives, and a commitment to and the system fragmentation they imply56, may over- facilitate a strategic planning process to identify dys- come some of the obstacles donors face by linking aid functional incentives and alter them to inspire health- to verifi able results that can be documented and re- improving actions. ported to taxpayers and other stakeholders.

In spite of its promise, the evidence based on PBA The growing donor interest in using PBA mechanisms requires urgent strengthening. Given current donor provides an exciting opportunity to shift the focus of staff skill mix, it is likely that specialized capacity in international health aid from inputs to outputs and contract design, synthesis of norms and best prac- outcomes. Many factors affect the likelihood that PBA tices, and testing low-cost solutions for performance schemes will be effective at improving health status verification is needed. Further, PBA would benefit in developing countries. In particular, inattention to from a dedicated program of policy research on con- incentive structures and technical capacity appear to tract and incentive design and evaluation. be particular stumbling blocks in the achievement of

HEALTH SYSTEMS STRENGTHENING VIA PERFORMANCE-BASED AID 21 ENDNOTES 12. Lane C, Glassman A. Bigger and better? Scaling up and innovation in development aid for health. 1. Please see http://www.brookings.edu/projects/ Health Affairs July/August 2007. global-health/health-snapshot.aspx for brief ex- planations of each of these innovative fi nancing 13. Schieber GJ, Gottret P, Fleisher LK, Leive AA. Fi- mechanisms. nancing global health: mission unaccomplished. Health Affairs, Volume 26, Number 4, July/August 2. Lane C, Glassman A. Bigger and better? Scaling 2007. up and innovation in development aid for health. Health Affairs July/August 2007. 14. Chee, Grace, Natasha His, Kenneth Carlson, Sla- vea Chankova, Patricia Taylor. Evaluation of the 3. Center for Global Development Evaluation Gap fi rst fi ve years of GAVI immunization services sup- Working Group. When will we ever learn? Improv- port funding. Washington, DC: Abt Associates Inc., ing lives through impact evaluation. Washington, September 2007. DC: Center for Global Development, May 2006. 15. C. Lu, C. Michaud, E. Gakidou, K. Khan, C. Murray 4. See http://www.cgdev.org/section/initiatives/_ac- (2006) “Effect of the Global Alliance for Vaccines tive/ghprn/workinggroups/performance for docu- and Immunisation on diphtheria, tetanus, and ments produced by a working group on perfor- pertussis vaccine coverage: an independent as- mance-based incentives. sessment.” The Lancet, Volume 368, Issue 9541, 5. Osborne D, Gaebler T. Reinventing Government. Pages 1088-1095. 1992; US Government Performance and Results 16. Chee et al. (2007) op. cit. act of 1993 (P.L. 103-62). 17. Hsi, Natasha and Rebecca Fields, Evaluation of 6. Meltzer; Radelet S. Challenging foreign aid: a GAVI Immunization Services Supporting Funding policy-maker’s guide to the Millennium Challenge Case Study: Kenya. prepared for The Global Alli- Account. Washington, DC: Center for Global De- ance for Vaccines and immunizations by Abt As- velopment, 2003; Birdsall N, Barder O. their PBA sociates Inc. and Academy for Educational Devel- paper. opment, June 2004. 7. For example, see IDB. New Lending Framework. 18. Chee, Grace, Rebecca Fields, Natasha Hsi, and 2005. Whitney Schott. Evaluation of GAVI Immunization 8. Such as required counterpart fi nancing or other Services Support Funding, prepared for The Glob- administrative requirements. al Alliance for Vaccines and Immunizations by Abt 9. http://www.oecd.org/dataoecd/11/41/34428351. Associates Inc. August 2004. pdf. 19. Chee et al. (2007) op. cit.

10. Defi nition fi rst used by Foster M, Brown A, Con- 20. http://www.immunizationbasics.jsi.com/Docs/BA- way T. Sector-wide approaches for health devel- SICS_REDpresentation.ppt. The fi ve elements of opment: a review of experience. Overseas Devel- RED are: (1) Re-establishment of outreach ser- opment Institute and World Health Organization, vices with regular outreach for communities with 2003. poor access; (2) Supportive supervision with on 11. Ibid.; Sundewall J, Forsberg BC, Tomson G. Theory site training by supervisors; (3) Community links and practice – a case study of coordination and with service delivery through regular meetings ownership in the Bangladesh health SWAp. Health between community and health staff; (4) Monitor- Research Policy and Systems 2006, 4:5. ing and use of data for action, such as chart dos- es, map population in each health facility; and (5)

22 GLOBAL ECONOMY AND DEVELOPMENT PROGRAM Planning and management of resources for better 8.60 and Bank Procedures 8.60, August 2004. management of human and fi nancial resources. These are available at http://lnweb90.world- bank.org/institutional/manuals/opmanual.nsf/ 21. Chee et al. (2007) op. cit. 284229c803270fad8525705a00112597/ce3b58a 22. Transparency International Global Corruption Re- b704d4228525705c001a9e68?OpenDocument. port 2006, Ch 6, p104, accessible online at http:// 30. Ibid. www.transparency.org/global_priorities/health/ hiv_aids. 31. Ibid.

23. http://www.gao.gov/new.items/d05639.pdf 32. Ibid.

24. Of the 213 grants approved on or before January 1, 33. World Bank. “Poverty Reduction Support 2005, 91 percent had had phase 2 funding agree- Credits: A Stocktaking.” (Washington: World ments approved by the Global Fund board within Bank, June 24, 2005). Available online at: 2 1/2 years. Of the remaining 9 percent, 5 percent http://siteresources.worldbank.org/PROJ- had grant scorecards recommending approval or ECTS/578280-1120680791169/20573838/ conditional phase 2 approval, 1 percent had been PRSC%20Stocktaking%20_06-28-05_.pdf. transferred to new implementing agencies (Prin- 34. World Bank, Project Appraisal Document (PAD), cipal Recipients), 1 percent were still operating but Report Number 37292-AR, October 23, 2006. had disbursement delays of up to a year, and only 1 percent appeared to be inoperative with no dis- 35. World Bank, PROJECT INFORMATION DOCUMENT bursements for at least a year. Personal commu- (PID) CONCEPT STAGE, Report Number AB2388, nication, Christopher Lane, July 14, 2007. May 16, 2006.

25. See http://www.theglobalfund.org/en/fi les/about/ 36. World Bank, Project Appraisal Document (PAD), replenishment/progress_report_midyear_2006. Report Number 37292-AR, October 23, 2006, p. pdf. 10.

26. In almost 70% of cases, these credits, or subsi- 37. Communication with World Bank Task manager, dized loans, are effectively grants. Cristian Baeza.

27. World Bank. “Poverty Reduction Support 38. Communication with World Bank Task manager, Credits: A Stocktaking.” (Washington: World Cristian Baeza. Bank, June 24, 2005). Available online at: 39. World Bank, Project Appraisal Document (PAD), http://siteresources.worldbank.org/PROJ- Report Number 37292-AR, October 23, 2006. ECTS/578280-1120680791169/20573838/ 40. Ministry of Health Argentina, “Informe de la PRSC%20Stocktaking%20_06-28-05_.pdf. Gestión del Plan Nacer a Diciembre de 2006 28. Nyanzi, Deo. “Introduction to Poverty Reduction Cuadros de detalle por provincia,” PowerPoint Support Credit (PRSC) and Poverty Reduction presentation, February 13, 2007. Growth Facility (PRGF).” Economic Policy Empow- 41. Porto, Alberto, Walter Valle and Alfredo Perazzo. erment Programme, European Network on Debt “ Las Transferencias Intergubernamentales del and Development. www.eurodad.org/epep. Plan Nacer Como Un Cambio en el Sistema de 29. World Bank. Interim Guidelines for Poverty Gestion,” August 15, 2006 draft document. Reduction Support Credits. (PRSCs), May 29, 42. Communication with World Bank task manager, 2001. This document may be diffi cult to fi nd. Cristian Baeza. It has been replaced by Operational Policy

HEALTH SYSTEMS STRENGTHENING VIA PERFORMANCE-BASED AID 23 43. This section is based primarily on: http://www.spa- 55. Lancet article http://www.thelancet.com/journals/ psa.org/resources/2005/EC_GBS_VT_Review.pdf. lancet/article/PIIS0140673607611441/fulltext.

44. http://www.spa-psa.org/resources/2005/EC_ 56. http://www.oecd.org/dataoecd/11/41/34428351. GBS_VT_Review.pdf. pdf.

45. Ibid.

46. Ibid.

47. http://www.who.int/vaccines/globalsummary/im- munization/countryprofi leresult.cfm.

48. Inter-American Development Bank. Proposal for an experimental program of performance- driven lending (GN-2278-1/2). Washington, DC: Inter-American Development Bank, December 16, 2003.

49. Inter-American Development Bank. Colombia. Pro- posal for a loan to strengthen the expanded pro- gram of immunization. Performance-driven loan. PR-2928. Washington, DC: IDB, April 19, 2005. This loan was $107 million USD.

50. Inter-American Development Bank. Honduras. Improvement of Health Conditions in Hondu- ras. Performance-Driven Loan. HO-L1002. Wash- ington, DC: IDB, March 6, 2005. The loan was $16,600,000. Document available at http://www. iadb.org/projects/Project.cfm?project=HO- L1002&Language=English.

51. Inter-American Development Bank. Nicaragua. Proposal for a loan for a project to improve ma- ternal and child health. Performance-driven loan. PR-2908. Washington, DC: IDB, November 23, 2004. This loan was 30,000,000 USD.

52. Inter-American Development Bank. Colombia. Pro- posal for a loan to strengthen the expanded pro- gram of immunization. Performance-driven loan. PR-2928. Washington, DC: IDB, April 19, 2005. The description of the program that follows is based on this reference.

53. Personal communication, Leslie Stone, Inter- American Development Bank, November 17, 2007.

54. PBI working group, CGD.

24 GLOBAL ECONOMY AND DEVELOPMENT PROGRAM HEALTH SYSTEMS STRENGTHENING VIA PERFORMANCE-BASED AID 25 26 GLOBAL ECONOMY AND DEVELOPMENT PROGRAM The views expressed in this working paper do not necessarily refl ect the offi cial position of Brookings, its board or the advisory council members.

© 2007 The Brookings Institution

ISSN: 1939-9383

Printed on recycled paper with soy-based inks.

Selected photos courtesy of the World Bank: cover left to right: (#1) Bill Lyons, (#5) Curt Carnemar 1775 Massachusetts Avenue, NW Washington, DC 20036 202-797-6000 www.brookings.edu/global