Grant Performance and Health Financing: Analysis of 12 Years of Financing Decisions at the Global Fund to Fight AIDS, Tuberculosis and Malaria

Abstract

Abstract—The Global Fund to Fight AIDS, Tuberculosis and Malaria (“the Global Fund”) was established to attract, manage, and disburse additional financial resources toward Millennium Development Goal (MDG) 6. To increase the efficiency and impact of resources provided, performance-based financing (PBF) was included among the founding principles of the Global Fund. Since 2002, the Global Fund has disbursed over 28 billion USD to 145 countries or regional programs worldwide. This article investigates the degree to which PBF principles were followed in practice as well as the equity implications for the Global Fund's grant portfolio. We obtained performance and financial data on all disbursements (N = 6,064) and grant renewals (N = 892) between 2003 and 2014. Multivariate regression analysis was used to estimate the associations between performance ratings and financing amounts. The dependent variable was the fraction of original budgeted amount disbursed or renewed by the Global Fund. The main independent variable was the performance rating received by the grant. We also compared average funding shares by country income group before and after PBF adjustments. Our analysis found that lower performance ratings were associated with a smaller amount of the next period's budget disbursed; in addition, lower ratings were associated with a higher probability of the grant not being renewed after two years of implementation and a smaller proportion of recommitted funds. Low and lower-middle income countries obtained marginally lower grant performance ratings but still received 90% of the total grant renewal amounts from the Global Fund. These results suggest that the Global Fund has succeeded in implementing PBF principles over the past 12 years without shifting resources away from lower-income countries. Larger impact may be possible in the long run if health outcomes rather than only outputs or coverage indicators are also incentivized in PBF contracts between countries and the Global Fund.

Keywords:

• The Global Fund to Fight AIDS
• Tuberculosis and Malaria
• grant performance
• health financing
• MDG 6
• performance-based funding
• results-based financing

INTRODUCTION

The Global Fund to Fight AIDS, Tuberculosis and Malaria (“the Global Fund”) was established following the Genoa G-8 Summit in 2001 to “attract, manage and disburse additional resources [to] make a sustainable and significant contribution to the reduction of infections, illness and death,” thereby mitigating the impact of these pandemics and contributing “to poverty reduction as part of the MDGs [Millennium Development Goals].”¹

In order to maximize health outcomes of available resources and avoid disappointing results from past aid efforts, ^2-4 performance-based financing (PBF) was included among the founding principles of the Global Fund: “In making its funding decisions, the Global Fund will […] focus on performance by linking resources to the achievement of clear, measurable and sustainable results.”¹ As opposed to traditional input-based financing where resources are provided according to budgetary needs, performance-based models use the achievement of prespecified outputs and outcomes as the primary criterion for funding decisions.^4-8

Since 2002, the Global Fund has disbursed over 28 billion USD to 145 countries or regional programs worldwide.⁹ Though the positive effects of financing provided by the Global Fund have been widely acknowledged, and earlier studies recognized the merits of its performance-based model,^10,11 the funding decisions of the Global Fund have come under criticism in recent years. Two recent reviews conclude that “the Global Fund's existing performance-based financing system provides only a weak link between grant performance and disbursement amounts”¹² and that “because there is no clear link between performance and at least a portion of overall funding, the Global Fund does not effectively transmit performance incentives to its implementing partners.”¹³

The present study analyzes all disbursement and grant renewal decisions made by the Global Fund between 2003 and 2014. Specifically, we assess the degree to which the Global Fund's financing decisions were performance based and also the extent to which these decisions altered the allocation of funding across country income groups.

METHODS

Background

The original funding model of the Global Fund (before it was modified in 2014) was divided into two stages: a proposal stage where funding was assigned to countries and negotiated into grant agreements based on the technical merit of applications, notably soundness of approach, feasibility, potential for sustainable outcomes, and value-for-money (Step 1, shown in the light grey box in Figure 1), and an implementation stage in which period-specific disbursements and grant renewals were made conditional on performance criteria (Steps 2 and 3, shown in the dark grey boxes in Figure 1).

Figure 1. Global Fund Grant Life Cycle (2002–2014)

Each Global Fund grant life cycle started with a five-year funding application submitted by a country's multistakeholder partnership composed of representatives from government, civil society, and the private sector. These applications were assessed by an independent technical review panel appointed by the Global Fund.¹⁴ Ten rounds of funding application cycles occurred between 2002 and 2010. The success rate of applications varied between 31% in 2005 (Round 5) and 54% in 2008 (Round 8).

Once an application was approved, a legally binding grant agreement covering an initial two years of implementation (referred to as Phase 1) was negotiated with recipient organizations. This agreement typically contained three to five outcome and impact indicators, to assess overall progress towards Program goals after two to three years, as well as up to 20 process, output, and/or coverage indicators that allowed the Global Fund to monitor implementation of activities over shorter time periods.

After an initial disbursement to cover the first grant period, recipients had to submit periodic progress reports, every three, six, or 12 months depending on reporting cycles, that were used by the Global Fund to assess grant performance and decide on the next amount to be disbursed. These reports included indicator results vis-à-vis targets, expenditures against budget, and progress on management actions. Each grant received an initial performance rating based on the average achievement rate of indicators as follows.¹⁵ A1: >100%; A2: 90%–100%; B1: 60%–89%; B2: 30%–59%; C: <30%. This initial rating could be downgraded because of grant management issues, such as ineligible expenditures, stock-outs of health products, or inaccurate indicator data. Disbursements were then supposed to follow performance-based financing ranges, expressed as percentage of cumulative budget disbursed (outlined in Figure 2). The rationale for the performance rating and amount disbursed was communicated in writing to grant recipients by the Global Fund.

Figure 2. Performance-Based Financing Disbursement Ranges

In addition to short-term adjustments in expected disbursement amounts, the Global Fund conducted a comprehensive review of grant performance and health outcomes at the end of the second year of implementation (Phase 1) and decided how much, if any, of the original funding application amounts would be confirmed for the remaining three years (referred to as Phase 2). This decision was called “Phase 2 grant renewal” under the original funding model of the Global Fund. Amounts not confirmed for Phase 2 were made available to all countries in subsequent application rounds of financing from the Global Fund.

Data Sources

We obtained data from the Global Fund's databases, also available on the external website, related to (1) all progress updates and disbursement requests submitted by grant recipients between 2003 and 2014, with the corresponding disbursement ratings and amounts from the Global Fund, and (2) all grant renewal decisions, including Phase 2 ratings and amounts, by the Global Fund between 2005 and 2014.

Out of the total number of progress updates and disbursement requests (N = 7,695), we excluded those with no indicators to be reported against (n = 155), those with negative or null budgeted amounts for the next period (n = 745), those with missing or incoherent reporting period dates (n = 446), and those with missing performance ratings (n = 285). The final analytical sample for disbursements consisted of 6,064 decisions with a total disbursement amount of 15.6 billion USD.

Out of the total number of grant renewals (N = 911), we excluded those with a change of grant recipient between Phase 1 and Phase 2 (n = 6) and those for which the renewal decision was consolidated across all grant recipients for the same disease (n = 13). The final analytical sample for grant renewals consisted of 892 decisions amounting to 14.6 billion USD.

Data Analysis

Our empirical analysis starts with descriptive statistics of period-specific disbursements (Step 2 in Figure 1) and Phase 2 grant renewals (Step 3 in Figure 1) from our analytical sample.

We then use ordinary least squares (OLS) models to analyze the associations between performance ratings and subsequent disbursement and renewal amounts granted by the Global Fund. For both disbursement and renewal decisions, the main dependent variable is the proportion of originally budgeted amount disbursed or renewed. For Phase 2 decisions, the Global Fund also had the option of discontinuing the grant altogether; we show separate results for this decision, taking the probability of nonrenewal as a dependent variable. In all models, we use grants with an A rating as our reference group and include binary indicator variables for grades B1, B2, and C as the main independent variables of interest.

In order to ensure that the estimated associations between performance ratings and disbursements or grant renewals are not driven by unobservable country characteristics, we also estimate specifications with country-level fixed effects for all outcome variables. These models only explore differences in disbursement and grant renewal shares within countries and are therefore not affected by differences in actual or perceived need or capacity of specific countries. For period-specific disbursement regressions, all countries have at least two reports with a rating and a subsequent disbursement decision; hence, all countries are included in the fixed effect analysis. For Phase 2 grant renewals, 19 countries had only one renewal and thus do not provide any variation in the fixed effect analysis.

In the last part, we divide countries into four income brackets: those with per capita income of less than 5,000 USD, those with per capita income of 5,000–9,999 USD, those with per capita income of 10,000–19,999 USD, and those with per capita income above this threshold. To investigate concerns that PBF may unduly penalize lower-income countries, we analyze across income levels the distribution of Phase 2 ratings and the proportion of renewal amounts confirmed by the Global Fund.

For all empirical analyses, standard errors are clustered at the country level. Statistical analysis was conducted using the Stata SE 12 software package.¹⁶

RESULTS

Table 1 summarizes data on period-specific disbursements and Phase 2 renewals. The 6,064 observed disbursement decisions ranged between 0 USD and 122 million USD, with an average disbursed amount of 4.4 million USD. Average grant renewal amount (N = 892) was 19.9 million USD. About one third of grants received an A rating at the time of disbursements, 45% received a B1 rating, 17% a B2 rating, and 4% a C rating. HIV/AIDS was the most common health issue targeted by grants, followed by tuberculosis and malaria.

Table 1. Period-Specific Disbursements and Phase 2 Renewal Decisions: Descriptive Statistics

	Period-Specific Disbursement Decisions (N = 6,064)		Phase 2 Renewal Decisions (N = 892)
Amount (million USD), Mean (SD)	4.4	(10.6)	19.9	(33.0)
Performance rating A, N (%)	2,022	(33.3)	247	(27.8)
Performance rating B1, N (%)	2,736	(45.1)	436	(49.0)
Performance rating B2, N (%)	1,060	(17.5)	178	(20.0)
Performance rating C, N (%)	244	(4.0)	28	(3.3)
Approved <20%, N (%)	1,702	(28.1)	44	(4.9)
Approved 20%–50%, N (%)	1,085	(17.9)	43	(4.8)
Approved 50%–80%, N (%)	1,125	(18.6)	237	(26.6)
Approved ≥80%, N (%)	2,147	(35.4)	567	(63.7)
HIV, N (%)	2,666	(44.0)	373	(41.9)
Malaria, N (%)	1,557	(25.7)	237	(26.7)
TB, N (%)	1,709	(28.2)	257	(28.8)

Table 2 summarizes our main multivariable regression results. In Panel A we show the pooled results across all grants (exploring variation in ratings both within and across countries); in Panel B we show estimates with country fixed effects (within country variation only). In both panels, results for three separate dependent variables are displayed: the percentage of the budgeted period amount disbursed (column 1), the probability of a grant not getting renewed in Phase 2 (column 2), and the percentage of the original funding application amount confirmed for Phase 2 (column 3).

Table 2. Associations Between Performance Ratings and Financing Decisions: Multivariable Regression Results. Table shows results from OLS regressions. Each observation corresponds to one financing decision. In column 1, the dependent variable is the fraction of the expected period budget disbursed. In column 2, the dependent variable is the probability of a grant not getting renewed in Phase 2. In column 3, the dependent variable is the fraction of the original Phase 2 amount renewed. Standard errors are clustered at the country level. Panel B includes country fixed effects. ***p < 0.01, **p < 0.05, *p < 0.1

	Panel A: Pooled Analysis
	Decision Point
	Period-Specific Disbursement Relative to Original Budget	Phase 2 Nonrenewal	Phase 2 Renewal Amount Relative to Original Budget
	(1)	(2)	(3)
Rating A	Reference Group
Rating B1	−0.046	0.011*	−0.070***
	(−0.124 to 0.032)	(−0.000 to 0.023)	(−0.106 to −0.035)
Rating B2	−0.185***	0.073***	−0.231***
	(−0.291 to −0.079)	(0.026 to 0.120)	(−0.293 to −0.168)
Rating C	−0.443***	0.483***	−0.623***
	(−0.602 to −0.284)	(0.286 to 0.679)	(−0.752 to −0.494)
Reference group (Rating A) average	0.93	0.0	0.91
Observations	6,064	892	892
R^2	0.047	0.214	0.196
	Panel B: Within Country (Country Fixed Effect) Analysis
	Decision Point
	Period-Specific Disbursement Relative to Original Budget	Phase 2 Nonrenewal	Phase 2 Renewal Amount Relative to Original Budget
Rating B1	−0.068*	0.016	−0.071***
	(−0.145 to 0.009)	(−0.008 to 0.040)	(−0.123 to −0.019)
Rating B2	−0.180***	0.078***	−0.231***
	(−0.287 to −0.073)	(0.020 to 0.137)	(−0.315 to −0.146)
Rating C	−0.433***	0.503***	−0.618***
	(−0.617 to −0.250)	(0.279 to 0.727)	(−0.766 to −0.471)
Observations	6,064	892	892
R^2	0.112	0.350	0.310

We find strong empirical association between grant performance ratings and amounts disbursed or renewed (Panel A). For period-specific disbursements, a B1 rating reduced the proportion of the expected disbursement amount by 4.6 percentage points (95% confidence interval [CI], −0.124, 0.032), a B2 rating by 18.5 percentage points (95% CI, −0.291, −0.079), and a C rating by 44.3 percentage points (95% CI, −0.602, −0.284), compared to an A rating (Column 1). When looking at the risk of nonrenewal, B2 and C ratings increased the probability of a grant not getting renewed at all for Phase 2 respectively by 7.3 percentage points (95% CI, 0.026, 0.120) and by 48.3 percentage points (95% CI, 0.286, 0.679), compared to an A rating (Column 2). Finally, a B1 rating lowered the proportion of original Phase 2 funding application amount confirmed by 7.0 percentage points (95% CI, −0.106, −0.035), a B2 rating by 23.1 percentage points (95% CI, −0.293, −0.168), and a C rating by 62.3 percentage points (95% CI, −0.752, −0.494), compared to an A rating (Column 3).

Very similar results emerge when we include country fixed effects in our empirical model and only compare performance ratings with disbursements and grant renewals within countries (Panel B). Keeping all other country characteristics the same, a C rating reduced the proportion of the expected disbursement amount by 43.3 percentage points (95% CI, −0.617, −0.250), increased the probability of a grant not getting renewed at all for Phase 2 by 50.3 percentage points (95% CI, 0.279, 0.727), and lowered the proportion of original Phase 2 funding application amount confirmed by 61.8 percentage points (95% CI, −0.766, −0.471), compared to an A rating (Columns 1–3).

In Table 3, we show separate results by time period for disbursements. Overall, the associations between performance ratings and disbursement amounts look similar across the time periods, with slightly stronger responses to C ratings early on and slightly stronger responses to B1 and B2 ratings in later years. In Table 4, we also show results stratified by diseases (HIV/AIDS, malaria, and tuberculosis [TB]); no statistically different patterns were found across diseases. In Table 5, we compare the percentages of the grant renewal amounts in Phase 2 to the indicative renewal ranges. Overall, 47% of funding decisions were exactly in the indicative range, 35% of grants received funding confirmation above the indicative range, and 18% of grants received funding confirmation below the indicative range.

Table 3. Period-Specific Associations Between Performance Ratings and Financing Decisions: Multivariable Regression Results. Table shows results from ordinary least squares regressions. Each observation corresponds to one financing decision. The dependent variable in all models is the fraction of the expected period budget disbursed. Standard errors are clustered at the grant level. ***p < 0.01, **p < 0.05, *p < 0.1

	Period-Specific Disbursement
Period	Years 2003–2005	Years 2006–2008	Years 2009–2011	Years 2012–2014
Rating B1	0.0384	−0.0279	−0.0892*	−0.0470
	(0.171)	(0.0697)	(0.0527)	(0.0832)
Rating B2	−0.109	−0.200**	−0.206**	−0.207*
	(0.179)	(0.0958)	(0.0824)	(0.107)
Rating C	−1.093***	−0.545***	−0.315**	−0.495***
	(0.127)	(0.145)	(0.137)	(0.0984)
Observations	489	1,905	2,222	1,448
R^2	0.007	0.005	0.005	0.008

Table 4. Associations Between Performance Ratings and Phase 2 Renewal Decisions by Disease: Multivariable Regression Results.Table shows results from ordinary least squares regressions. Each observation corresponds to one financing decision. The dependent variable is the fraction of the original phase 2 amount renewed. Standard errors are clustered at the country level. All diseases also include 16 HSS components. ***p < 0.01, **p < 0.05, *p < 0.1

	Phase 2 Renewal Percentage
	All Diseases	HIV	Malaria	TB
Rating B1	−0.0705***	−0.0701***	−0.0307	−0.0949***
	(0.0179)	(0.0261)	(0.0497)	(0.0214)
Rating B2	−0.231***	−0.252***	−0.145***	−0.277***
	(0.0315)	(0.0420)	(0.0497)	(0.0514)
Rating C	−0.623***	−0.628***	−0.606***	−0.550***
	(0.0655)	(0.115)	(0.127)	(0.152)
Observations	892	381	238	257
R^2	0.196	0.216	0.154	0.222

Table 5. Distribution of Phase 2 Renewals Relative to Indicative Range

Rating	Target Range (%)	Number of Renewals	Average % Renewed	% of Renewals Within Target Range	% of Renewals Above Target Range	% of Renewals Below Target Range
A	90–100	248	91.2	56.0	3.2	40.7
B1	60–89	437	84.1	50.6	41.4	8.0
B2	30–59	178	68.1	21.9	65.2	12.9
C	<30	29	28.9	58.6	41.4	NA
Total		892	81.1	46.6	35.5	17.8

Table 6 shows the overall distribution across income-level categories of Phase 2 ratings and grant renewal amounts from the Global Fund. Overall, countries with an income level of less than 5,000 USD per capita in 2010 have lower ratings; only 22% of grants in this bracket obtain an A rating for Phase 2 decisions, compared to 33%, 36%, and 44% in the higher income groups. At the same time, these lower average ratings have only a very limited impact on the overall allocation of funding, with close to 60% of Phase 2 funding going to countries in the lowest income quartile and 90% of Phase 2 funding going to countries in the bottom half of the income distribution.

Table 6. Income per Capita, Average Phase 2 Rating and Grant Renewal Amount

			Proportion of Grants Rated				Total Amount Originally Allocated		Total Amount Confirmed
Gross Domestic Product per Capita Range (USD)	Total Number of Countries in Income Group	Countries with Global Fund Grant	A	B1	B2	C	Million USD	Share (%)	Million USD	Share (%)
(0–4,999)	49	45	0.22	0.51	0.23	0.04	10,197	59.5	8,419	59.7
(5,000–9,999)	44	40	0.33	0.48	0.16	0.03	5,199	30.3	4,266	30.3
(10,000–19,999)	41	31	0.36	0.48	0.16	0.00	1,517	8.8	1,183	8.4
(20,000–)	53	4	0.44	0.42	0.13	0.02	230	1.3	226	1.6

DISCUSSION

This article has three main findings related to performance-based financing decisions at the Global Fund.

First, at the time of grant renewal after two years of implementation, the most poorly performing grants had an increase of almost 50 percentage points in the risk of discontinuation, compared to the best-performing grants. This finding that performance ratings affected the probability of nonrenewal in Phase 2 is consistent with the results of a previous study on PBF at the Global Fund.¹² The consensus on this finding confirms one way in which performance ratings have been used in making funding decisions: poor performance in Phase 1 increased the likelihood of nonrenewal. In addition, our analysis shows that, for grants that were renewed, the most poorly performing grants had financial cuts (in the amount originally allocated for Phase 2) in excess of 60 percentage points compared to best-performing grants. This finding is another way in which the Global Fund has used performance ratings in making grant renewal funding decisions.

Second, our analysis of disbursement decisions over the lifetime of the grants shows that the most poorly performing grants have received nearly 45 percentage points less expected disbursement amounts than the best-performing grants. In short, the individual disbursement decisions for each grant (previously two to four per year) have responded to performance measures, with lower proportions of disbursements approved for lower ratings. This finding suggests that the Global Fund has also largely followed PBF principles on its disbursement decisions between 2003 and 2014.

It is important to note that this analysis of disbursement decisions is different from a previous study that analyzed the associations between average grant performance ratings in Phase 1 and total disbursement amounts (rather than the proportion of the original budget disbursed) in Phase 2.¹² Though this assessment might make sense in a PBF model where the absolute amount granted in a given period or phase was exclusively based on performance ratings and independent of initial five-year programmatic plans and budgets, this is not how the Global Fund's model was designed or implemented. Rather than analyzing total disbursement amounts, our study examines the proportion of original budgeted amounts disbursed, in relation to performance ratings, for the next grant period (previously three to six months), which is more consistent with the disbursement ranges defined by the Global Fund.

Third, though grants from poorer countries received on average lower performance ratings at grant renewals after two years of implementation (Phase 1), this did not have a significant impact on the overall cross-country allocation of funding for Phase 2; our analysis shows that nearly 60% of total grant renewal amounts went to the lowest income quartile of countries and that 90% went to countries with incomes of 10,000 USD or less. Most performance-based funding adjustments were within, rather than across, country income categories. This finding does not corroborate concerns that PBF unduly penalizes lower-income countries because of a presumed inability to demonstrate and/or achieve expected results. This can partly be explained by the performance-based model of the Global Fund where indicator targets are proposed by countries themselves, considering context-specific implementation capacities and operating environments.

Our study emphasizes that the Global Fund's financing model was by design a mixed model that made an initial allocation of funding for each grant according to country needs and technical merit of applications and then used performance incentives for disbursements and also at the time of grant renewal to encourage recipients to use funds efficiently and effectively. Under its new funding model, the Global Fund is more explicit with respect to the objective of achieving an overall fair allocation of financial resources, directly earmarking funding every three years based on countries' income level and disease burden.

Though all data used in this study are in the public domain and, therefore, third parties could have conducted such an analysis, it would have been difficult and time-consuming because the Global Fund does not provide access to a unified database with information on performance ratings, budget frames, and funding decisions over time and across the portfolio of grants. Considering that transparency and accountability are among its founding principles, easier public access to data would not only benefit the research community, but also enable stronger oversight from external stakeholders. It is unfortunate that the Global Fund has gradually been providing less, rather than more, performance-related information about grant portfolios on its external website.

One of the key challenges with all performance-based mechanisms is defining and measuring “performance.” The Global Fund's selection of indicators to assess performance has been criticized for using too many input and process indicators, which could bias incentives away from health outcomes.¹³ Research suggests that there has been progress over time in the Global Fund's indicator selection, with a decline in the proportion of input/process indicators and an increase in the use of outcome/impact indicators.¹⁷ Furthermore, under its new funding model, the Global Fund is removing input/process indicators altogether from standard grant agreements. However, though routine disbursements continue to be based on shorter-term output and coverage measures, as they should be, because these are more frequently reported and easily verifiable, concerns remain regarding potentially insufficient incentives for programs to achieve longer-term impact in terms of reduced disease morbidity and mortality.

Overall, the results in this article show that the Global Fund has largely followed PBF principles over the past 12 years while ensuring a majority of financial resources to the most affected countries. These efforts at the Global Fund have contributed to a shift in the main paradigm of global health financing in the pursuit of MDG 6 toward performance-based principles. However, as highlighted by other studies, some challenges still remain and should be addressed by the Global Fund. The discontinuation of the grant renewal processes should be evaluated because it may have weakened the performance-based model of the Global Fund. But most important, the Global Fund should consider PBF mechanisms that can also reward improved health outcomes and not only incentivize increase coverage of key interventions.

Moving into its fifth replenishment, the Global Fund could consider adopting a three-tier financing approach for the next 2017–2019 funding cycle that would assure continuity while using performance-based incentives for greater impacts. The three tiers could include (1) “fixed tranches” of funding to cover operating expenses and the costs of maintaining current cohorts of patients on treatment; (2) “additional variable” amounts for increased coverage of target populations with key interventions; and (3) “reward payments” for improved health outcomes in terms of decreased morbidity and mortality. The proportional size of each tier of financing could vary according to the history and profile of the grants, as well as country income levels. Such a tailored approach, with appropriate national counterpart financing arrangements, could help address concerns of funding predictability and sustainability while at the same time maximizing health impacts. Arguably, this could better position the Global Fund as the international community defines the global health architecture to achieve the newly defined Sustainable Development Goals.

DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST

No funding was provided for this analysis. The Global Fund had no role in the study design, data collection, data analysis, data interpretation, or writing of the article. Two of the authors (R.T.B.H. and M.O.O.) are employed by the Global Fund, one of whom (R.T.B.H.) was on unpaid sabbatical leave during the time of this independent analysis. All coauthors had final responsibility for the decision to prepare and submit this article for publication.

Notes

[a] Where outcome generally means behavior change and impact means decreased morbidity and mortality.

[b] Reporting cycles were originally quarterly but then progressively shifted to semi-annual in 2008 and 2009. Disbursement decisions have been transitioned to annually since 2013.

[c] Until 2008 the thresholds for performance ratings were as follows: A: >80%; B1: 50%-79%; B2: 30%-49%; C: <30% (there was no distinction between A1 and A2 ratings).

[d] Data used in our analysis are available on the Global Fund website in “Grant Performance Reports” (disbursement ratings, budget frames, disbursement amounts) and “Grant Score Cards” (grant renewal ratings, original and confirmed Phase 2 amounts).

[e] Because there was no distinction between A1 and A2 ratings until 2008, we decided to collapse these two ratings into one for the purpose of the regression analysis both for disbursements and grant renewals.

APPENDIX A.

Features of the New Funding Model

In preparation of the fourth replenishment cycle for 2014–2016, and in support of its newly developed strategy, the Global Fund decided to roll out in 2013 a new funding model to replace its original one, which had remained the same since its establishment in 2002.

This new funding model was intended to enable the Global Fund to invest more strategically, maximize its resources, and achieve greater global impact in the fight against HIV/AIDS, tuberculosis, and malaria. It changed the way applicants applied for and received funding and then managed their grants. The key features of this new funding model were as follows:

1. Focus on high–disease burden/low-resource settings: Each eligible country has an allocated amount of predictable funding for the three-year allocation period based on their disease burden and income level.

2. Enhanced engagement: The Global Fund engages in existing country coordination mechanisms in health and development, with a focus on multistakeholder participation, prior to board approval of grants.

3. Flexible timeline: Eligible countries can apply at any point during the three-year allocation period so that funding aligns with country-specific demands and national budgeting cycles.

4. Duration of grant: Grants are signed for three-year periods of identical length to the replenishment cycles of the Global Fund.

5. PBF indicators: Focus on output and coverage indicators for disbursements (every year) and consideration of outcomes and impact for new allocation amounts (every three years).

In summary, the main differences between the original and new funding model can be described as follows:

Table A1. Key Features of Old and New Funding Model of the Global Fund

Features	Original Funding Model	New Funding Model
Funding amount	Unpredictable (dependent on outcome of application process)	Predictable (funding allocated to eligible countries every three years)
Timing of application	Fixed (at the time of rounds)	Anytime (during the three-year allocation period)
Global Fund engagement	Ex post (after submission of application and board approval)	Ex ante (before submission of application and board approval)
Grant duration	Five years (2 + 3)	Three years
Grant renewals	After two years (based on performance; i.e., Phase 2)	No renewal process (funding allocated every three years)
PBF indicators	Input, process, and output for disbursements; outcomes and impact for renewals	Output and coverage for disbursements; Outcomes and impact for new allocations

APPENDIX B.

Disbursement Decision Data Exclusions

Table B1. Data Exclusions. ^aDR, disbursement request; PU, progress update.

Reason for Exclusion^a	Number
No active indicator	155
Negative DR budget	437
Null DR budget	308
Missing DR date	139
Overlap PU and DR periods	307
Missing final rating	285
Grand total	1,631

We needed to exclude a number of observations because of missing or incorrect data that prevented us from having reliable performance or financial information or making appropriate linkages between progress report and disbursement request periods.

These are the following (in sequence of exclusion):

1. No active indicator: 155 disbursement decisions were based on periods for which there were no indicators to be measured, and therefore no performance ratings, even though there was a periodic progress report submitted. Typically, these relate to disbursements made early in the life cycle of grants, generally for start-up costs. The grant recipient would still report to the Global Fund but generally on its financial position rather than on any indicator result.

2. Negative disbursement request (DR) budget: 437 disbursement decisions were made against budgets that are currently negative for the period in question in the systems of the Global Fund. These negative amounts correspond to cases where the budget value was retroactively adjusted to reflect the rollover of unimplemented activities from one phase to another. To avoid double-counting these activities in cumulative budget calculations, the convention was to amend the original phase-specific budget by introducing negative adjustments. Therefore, the current budget value for these periods in the systems cannot be used, because it is unrelated to the original budget amount at the time of the disbursement.

3. Null DR budget: 308 disbursement decisions have a corresponding budget amount of zero for the disbursement period. These can be explained either by negative budget adjustments at the end of phases (as described above) or, more often, by late catch-up in program implementation at the end of a grant where the Global Fund exceptionally disburses some remaining funds before closure for activities that should have taken place previously.

4. Missing DR date: 139 cases had missing dates for the disbursement request period (either start and/or end date) because of data quality issues. This prevents retrieval of the corresponding budget amount for the disbursement decision.

5. Overlap progress update (PU) and DR period: 307 cases had overlapping dates for progress report and disbursement request periods. This is also a data quality issue because these periods should be sequential, with the start of a disbursement request period immediately following the end of a progress report period. These cases were therefore excluded because either the performance information (based on PU period) or the budget (based on DR period) would be wrong in the analysis.

6. Missing performance ratings: 285 disbursement decisions had no related grant performance rating. These are also caused by data quality issues, either because the rating was not entered in the Global Fund systems or, more commonly, because these cases are actually intermediary progress reports with no expected disbursements that were entered incorrectly as a decision to disburse zero.

Though these missing or incorrect data do not allow us to analyze the association between performance ratings and amounts disbursed for the corresponding observations, there is no indication that these exclusions bias in any direction our regression analysis of disbursements at the Global Fund.

APPENDIX C.

Performance Assessment and Funding Decision Methodology

The Global Fund's approach to performance-based financing involves (1) assessing grant performance and (2) relating it to a funding decision (i.e., disbursement and/or grant renewal).

Grant Performance

In assessing performance, the Global Fund focuses both on programmatic achievements (i.e., what has been achieved) and grant management (i.e., how it was achieved). This can be summarized as follows:

• • Have predefined indicator targets been reached within agreed-upon timelines?

• • Have implementers and recipients of funds adhered to good management practices?

Indicator Rating

Programmatic achievements are measured through an indicator rating. This rating is calculated by averaging the achievement rates of process and output/coverage indicators defined in the grant agreement (i.e., results versus targets). The thresholds used to determine the ratings are shown in Table C1:

Table C1. Performance Rating Thresholds. ^aUnless downgraded due to critical management issues.

A1	A2	B1	B2	C
Exceeding	Expected	Adequate	Inadequate	Unacceptable
>100%^a	100%–90%^a	60%–89%^a	30%–59%^a	<30%^a

Since the introduction of its new funding model in 2014, only output/coverage indicators are used in calculating the indicator rating. Until then, process indicators could also be used, but higher weight was attributed to top 10 indicators measuring delivery of key products and services to target populations (see Appendix D). At Phase 2 grant renewals, the indicator rating could be upgraded if there were documented evidence of impact in terms of disease morbidity and mortality.

Management Assessment

Grant Management is assessed in the following functional areas

• (1) Monitoring and evaluation

• (2) Program management

• (3) Financial management and systems

• (4) Pharmaceutical and health product management.

Based on this assessment, a rating is given to each of the four functional areas: (1) no issues; (2) minor issues; and (3) major issues.

Grant Rating

Unless there are critical or repeated issues (e.g., ineligible expenditures, stock-outs of health products, inaccurate indicator data), the grant rating equals the indicator rating; if not, it can be downgraded by one or more points (for example, from A2 to B1, or even B2).

Funding decision

For each grant rating, there are indicative disbursement ranges defined in Table C2:

Table C2. Indicative Disbursement Range Thresholds. ^aSince the new funding model in 2014, these disbursement ranges are now linked to the indicator rating (and not to the grant rating). ^bThe Global Fund has marginally modified these ranges over the years.

Grant Ratings	Indicative Disbursement Ranges^a
Exceeding	More than 100%^b of cumulative budget through the next reporting period
Expected	Between 85% and 105%^b of cumulative budget through the next reporting period
Adequate	Between 55% and 95%^b of cumulative budget through the next reporting period
Inadequate	Between 25% and 65%^b of cumulative budget through the next reporting period
Unacceptable	Less than 35%^b of cumulative budget through the next reporting period

As labeled, these ranges are only indicative and the final disbursement decision can also take into account other factors such as past expenditure rates, perceived ability to catch up, and external factors (e.g., force majeure). At Phase 2 grant renewals, similar ranges existed linking performance ratings to indicative grant renewal amounts.

APPENDIX D. Top 10 Indicators for Routine Global Fund Reporting (Until the New Funding Model)

Table D1. Top 10 Indicators. MDR-TB = Multi Drug Resistant Tuberculosis, HSS = Health System Strengthening

	Disease	Indicators for Routine Global Fund Reporting
1	HIV	Number of adults and children with advanced HIV infection currently receiving antiretroviral therapy
2	TB	Number of (a) new smear-positive TB patients detected, (b) new smear-positive TB patients who were successfully treated, and (c) laboratory-confirmed MDR-TB patients enrolled in second-line anti-TB treatment
3	Malaria	Number of (a) insecticide-treated nets or retreatment kits distributed to people and (b) households (or structures or walls) in designated target areas sprayed by indoor residual spraying in the past 12 months
4	Malaria	Number of people with fever receiving antimalarial treatment according to national policy (specify artemisinin-based combination therapy versus other therapy)
5	HIV	Number of women and men aged 15–49 years who received an HIV test in the last 12 months and who know their results
6	HIV	Number of HIV-positive pregnant women who received antiretrovirals to reduce the risk of mother-to-child transmission
7	HIV	Number of condoms distributed
8	HIV, TB, and malaria	Number of people benefiting from community-based programs: specify (a) care and support including orphan support, home-based management of malaria, and directly observed therapy; (b) behavior change communication outreach activities including specific target groups; and (c) disease prevention for people most at risk (except behavior change communication)
9	HIV/TB	Number of TB patients who had an HIV test result recorded in the TB register
10	HSS, HIV, TB, and malaria	Number of people trained