The role of the Technical Review Panel of the Global Fund to Fight HIV/AIDS, Tuberculosis and Malaria: an analysis of grant recommendations

The role of the Technical Review Panel of the Global Fund to Fight HIV/AIDS, Tuberculosis and... Abstract The independent Technical Review Panel (TRP) of the Global Fund to Fight HIV/AIDS, Tuberculosis and Malaria is a unique mechanism to review funding proposals and to provide recommendations on their funding. Its functioning and performance have received little attention in the scientific literature. We aimed to identify predictors for TRP recommendations, whether these were in line with the Global Fund’s ambition to give priority to countries most in need, and whether they correlated with grant performance. We combined data on proposals and applications under the Rolling Continuation Channel, TRP recommendations and grant implementation during the rounds-based mechanism (2002–2010) with country characteristics. Ordered logistic and OLS regressions were used to identify predictors for per-capita funding requests, TRP recommendations, Global Fund funding and grant performance ratings. We tested for financial suppression of large funding proposals and whether fragile or English-speaking countries performed differently from other countries. We found that funding requests and TRP recommendations were consistent with disease burden, but independent of other country characteristics. Countries with larger populations requested less funding per capita, but there is no evidence of financial suppression by the TRP. Proposals from fragile countries were as likely to be recommended as proposals from other countries, and resulting grants performed equally well except for lower performance of HIV/AIDS grants. English-speaking countries obtained more funding for TB and malaria than other countries. In conclusion, the independent TRP acted in line with the guiding principles of the Global Fund to direct funding to countries most in need without ex ante country allocation. The Global Fund appears to have promoted learning on how to design and implement large-scale programs in fragile and non-fragile countries. Other pooled financing mechanisms may consider TRP operating principles to generate high-quality demand, to promote learning and to direct resources to countries most in need. Global fund to fight AIDS, tuberculosis and malaria, technical review panel, health financing, sustainable development goals, millennium development goals, decision making, developing countries, evidence-based policy, health planning, health systems, international health policy, overseas development assistance, resource allocation Key Messages The Technical Review Panel (TRP) of the Global Fund to Fight AIDS, Tuberculosis and Malaria is a unique body ensuring that funding decisions are based on independent technical review of program proposals. It has received little attention in the peer reviewed literature. During the rounds-based mechanism, the TRP recommended Global Fund proposals without regard to funding volumes, population size, governance, domestic health expenditure or DAH, making Global Fund funding needs based without ex ante funding allocations. The Global Fund, working with its Technical Partners, was effective at overcoming lower capacity to design and implement programs in poorer and/or fragile countries. This success holds important lessons for financing health systems. Pooled international funding mechanisms in other sectors, such as the Global Environment Facility, the Green Climate Fund, the Global Partnership for Education or the International Fund for Agricultural Development, should study the workings and performance of the TRP and may consider establishing similar procedures. Introduction Since its establishment in 2001, the Global Fund to Fight AIDS, Tuberculosis and Malaria (henceforth the Global Fund) has disbursed $34.6 billion (in 2014 US$) between 2013 and 2015 (IHME 2016), making it the dominant external funding channel for the three diseases except for HIV/AIDS, where it is second to the US President’s Emergency Plan for AIDS Relief (PEPFAR). In late 2016, the Global Fund mobilized some $13 billion at a time when international aid budgets were under severe pressure owing to the refugee crisis in Europe and macroeconomic pressures on donor countries. The success of the Global Fund in generating quality demand for health financing and in mobilizing substantial volumes of development assistance for health (DAH) has been attributed to its unique design principles (Sachs and Schmidt-Traub 2017). These include a focus on ‘national ownership and respect [for] country-led formulation and implementation’; the evaluation of ‘proposals through independent review processes based on the most appropriate scientific and technical standards’; giving ‘due priority to the most affected countries’; and ‘linking resources to the achievement of clear, measurable and sustainable results’ (GFATM 2001). The Global Alliance for Vaccines and Immunization (Gavi) applies similar design principles and has been equally successful in increasing quality DAH demand for vaccines. The Global Fund’s independent Technical Review Panel (TRP), which reviews proposals and makes funding recommendations to the Board, represents a critical innovation (Sachs and Schmidt-Traub 2017). Among all institutions reviewed in the 2016 UK Multilateral Development Review (DFID 2016), Gavi is the only other mechanism that operates an independent review panel. The Global Fund Board has always followed TRP funding recommendations, and every Board-approved country program has been fully funded. So, although the Board took final funding decisions, they have been determined by the TRP. During the rounds-based funding mechanism (2002–2010), the TRP was asked to issue recommendations without ex ante country allocations based on the technical merit of each proposal. Recommendations were made by consensus on a scale of 1–4 (1—recommended; 2—recommended subject to clarifications; 3—not recommended but strongly encouraged to re-submit; 4—not recommended). During the first 10 rounds 42% of proposals were rated 1 or 2. To maintain the technical independence of the evaluation process and to avoid interference with political considerations relating to individual countries, the Global Fund Board voted en bloc on the entire set of TRP recommendations. Countries could appeal against funding decisions, and for most rounds some appeals were granted (GFATM 2016). The Rolling Continuation Channel (RCC) was introduced in November 2006 to allow grantees with strongly performing grants to apply for continuing funding using a simplified procedure under which the TRP reviewed proposals in ‘waves’ held approximately twice per year. As with the rounds-based mechanism, the TRP could invite unsuccessful proposals to reapply in a later wave. Yet, the TRP soon expressed concerns that the RCC overlapped with the rounds-based mechanism and risked fragmenting country efforts (GFATM 2008). The RCC was discontinued after eight waves in August 2010. After 2011, the Global Fund introduced the New Funding Model under which countries can apply for pre-agreed volumes of funding with modest additional ‘incentive funding’ (GFATM 2013a). TRP recommendations now prioritize interventions within a pre-agreed resource envelope. This diminished the discretion of countries and the TRP to set funding volumes, so we limit our analysis to the rounds-based mechanism. Several studies have investigated the performance of Global Fund grants (Lu et al. 2006; Radelet and Siddiqi 2007; Macro International 2009; Katz et al. 2010; Fan et al. 2013), including in fragile countries (Bornemisza et al. 2010; Patel et al. 2015), and their impact on health systems (Samb et al. 2009; Car et al. 2012; de Jongh et al. 2014; iERG 2014). There has been no systematic assessment of the role of the TRP. This gap in the literature is notable since the TRP constitutes a novel mechanism for allocating aid, a key challenge considered in the health financing (Ottersen et al. 2017) and the development economics literature (World Bank 1998; Temple 2010). Processes for aid allocation ought to balance between criteria for countries’ need and the effectiveness with which they can use incremental funding (Collier and Dollar 2002; Daniels and Sabin 2008; Wood 2008; Ottersen et al. 2017). The aid allocation challenge for the three diseases was further compounded by developing countries’ widespread lack of quality demand, expressed as high-quality diseases management strategies, when the Global Fund was established in 2002 (Lu et al. 2006; Atun and Kazatchkine 2009). To elucidate the role of the TRP, this study addresses three sets of questions that are central to understanding the demand-based funding model of the Global Fund. First, we consider which factors explain the volume of funding requested by countries for each disease. In particular, we investigate whether countries’ funding requests were in line with funding needs as defined by income and disease burden or whether there was evidence of financial suppression of large grants (i.e. lower ratings of proposals requesting higher total volumes of funding after controlling for other explanatory variables). Second, we identify factors that explain TRP ratings of grant proposals and whether these ratings were consistent with the Global Fund’s ambition to give priority to countries most in need. Third, we ask whether TRP recommendations and other proposal characteristics were a good predictor of successful implementation of Global Fund grants, and whether there was evidence for changes in the quality of proposals over time. The success of scaling-up health investments despite initial concerns about absorptive capacity (Lu et al. 2006; Sachs and Schmidt-Traub 2017) stands in contrast to other investment priorities. Health systems, non-communicable diseases, environmental health and non-health-related areas have not experienced a similar scaling up of aid (de Jongh et al. 2014; IHME 2016). They have also seen a lower acceleration of progress under the Millennium Development Goals (McArthur and Rasmussen 2017). Achieving the recently adopted Sustainable Development Goals (SDGs) will require large increases in domestic and external funding for health systems, non-communicable diseases, environmental health (Jamison et al. 2013), education and other areas (Schmidt-Traub 2015). With Gavi and the Global Fund routinely rated among the best performing financing mechanisms (DFID 2016) we also consider whether the performance of the TRP may hold lessons for scaling up funding in other areas. Materials and methods Consistent with earlier investigations into the performance of Global Fund grants (Lu et al. 2006; Radelet and Siddiqi 2007; Katz et al. 2010; Fan et al. 2013; Bowser et al. 2014), we constructed statistical models testing dependent variables against two groups of regressors (i) proposal or grant characteristics relating to both the rounds-based mechanism and the RCC, and (ii) country characteristics covering criteria for need and the effectiveness with which aid might be used by a country. Regressors relating to proposals and grants include the volume of funding requested; the year of the round to track changes over time; TRP recommendations and Board decisions; and indicators of grant performance. Two dummy variables identify learning effects (van Kerkhoff and Szlezák 2006, 2016) and other changes made to proposals in repeat submissions. The first tracks whether proposals and the resulting grants were a resubmission of an earlier proposal rated 3 by the TRP. A second dummy identifies countries that have already had one grant approved by the Global Fund. To combine information on funding proposals and approved grants, proposals rated 1 or 2 by the TRP were manually matched with signed grants from the grants dataset using proposal parameters provided in both datasets, including the proposal name, disease component, country, funding volumes and the year of submission. Consistent with earlier studies into Global Fund grants, we grouped proposals and grants into three disease categories: HIV/AIDS including combined proposals for HIV/AIDS and TB, TB, and malaria (Radelet and Siddiqi 2007; Fan et al. 2013). During the rounds-based mechanism, funding requests for health systems strengthening could be either embedded in disease-specific proposals or be submitted separately under the category ‘other’. Since only 54 proposals were submitted under this separate category it was not possible to study proposals and grants relating specifically to health systems using our statistical models. Regressors relating to country characteristics include measures of need employed by the Global Fund (GFATM 2001, 2013a) and the literature on aid (Temple 2010; Ottersen et al. 2017), including GDP per capita and measures of disease burden. We also consider domestic health expenditure and official development assistance received from other donors in line with Lu et al. (2006) and Fan et al. (2013). Coverage of disease interventions, such as long-lasting insecticide-treated malaria bednets (LLINs), artemisinin-based combination therapy (ACT) or anti-retroviral therapy (ART), can be positively or negatively associated with funding needs depending on the intervention. However, data on intervention coverage were too incomplete for inclusion, except for the TB treatment success rate. Finally, the aid allocation literature shows that countries with large populations receive less aid ceteris paribus (Younas 2008; Bourguignon et al. 2009), so we test for financial suppression by including population size among the regressors. Factors associated with the effectiveness with which incremental health financing can be used include health systems, governance, and other country characteristics. The Universal Health Coverage (UHC) Tracer (GBD 2016) tracks the robustness of countries’ health systems, which are an important factor in scaling up complex disease control and treatment programs (Samb et al. 2009; Bowser et al. 2014; iERG 2014). Building on the work of World Bank (1998) and Collier and Dollar (2002), the health and development economics literature has considered the role of governance on the effectiveness with which aid can be used (Temple 2010; Ottersen et al. 2017). Many investigations into the performance of Global Fund grants therefore consider measures of governance (Lu et al. 2006; Radelet and Siddiqi 2007; Katz et al. 2010). The six most commonly used governance indicators published by the World Bank (2016) exhibit high collinearity for the sample of countries considered in this study. We therefore use three variables (control of corruption, government effectiveness, voice and accountability) that are in line with the literature (Katz et al. 2010), the success factors identified by the Global Fund (GFATM 2001, 2013 b), and have been reviewed in external assessments (Macro International 2009; HLIRP 2011). Additional country characteristics include dummy variables for fragile states, which constitute difficult ‘operating environments’ in the language of the Global Fund, and English-speaking countries since it has been reported that non-English speaking countries faced greater difficulties in accessing funding (Kerouedan 2010; French Ministry of Foreign Affairs 2013). We include a PEPFAR dummy variable for the countries in Africa that were eligible for PEPFAR funding. This allows us to test for complementarity with Global Fund funding for HIV/AIDS, as discussed in the literature (Shakow 2006; Oomman et al. 2007). During the rounds-based mechanism, aid allocations by the Global Fund were determined by a combination of the volume of funding requested by each country, TRP ratings and subsequent Board decisions, and the frequency with which countries applied for funding. To understand how the country characteristics related to needs and efficiency contribute to each step in the allocation process, we ensure maximum coherence in the variables included in Regressions 1–3. The Supplementary data describes the Global Fund data and the construction of the dependent variables. It also provides summary statistics and data sources for all variables considered in this study. The accompanying Stata dataset and do-file allow for replication of the analysis. Regression 1(Table 1) tests the first question, by conducting a linear ordinary least squares (OLS) regression for each of the three disease categories. The model was specified as   yi=Xiβi+εi, (1) where the dependent variable yi is the natural log of the sum of total per capita funding requested by disease i per Global Fund round and any RCC waves held since the previous funding round. Xi denotes the matrix of regressors xij including proposal and country characteristics relating to need and effectiveness, as described above and summarized in Table 1, βi is the matrix of regression coefficients βij estimated in this OLS model, and εi is the disturbance term for each disease. Each regressor comprises data from the year preceding the respective Global Fund round since the TRP considered proposals on the basis of this information. To identify possible non-linearities in submissions by round we considered a linear model specification as well as factor variables for each round. Table 1. Predictors of requested funding volumes (OLS) per round, by disease (Regression 1)   (1)  (2)  (3)    HIV  TB  Malaria  ln(GDP pc), 2014 cUS$  −0.0328  −0.0119  0.0929    (0.0825)  (0.0859)  (0.0869)  ln(population)  −0.671***  −0.544***  −0.664***    (0.0293)  (0.0330)  (0.0361)  UHC tracer (0–100)  −1.089**  1.106*  −2.384***    (0.396)  (0.449)  (0.478)  Dom. health exp, %GDP  1.779  −0.196  −1.157    (2.020)  (2.125)  (2.458)  DAH Global Fund, %GDP  41.75***  3.087  7.450    (12.22)  (13.07)  (13.60)  Non-Global Fund DAH, %GDP  1.134  5.077  18.35***    (5.012)  (5.141)  (5.243)  Prevalence, %  0.0960***  1.207***  0.0224***    (0.0160)  (0.274)  (0.00334)  TB treatment success, %    −0.00691        (0.00447)    Year of Round (2002–2010)  0.154***  0.226***  0.269***    (0.0235)  (0.0266)  (0.0311)  Resubmitted proposal (0/1)  −0.0250  0.111  0.0567    (0.0798)  (0.0860)  (0.0911)  Any grant already approved (0/1)  −0.0715  −0.142  0.123    (0.0852)  (0.0892)  (0.0964)  RCC held since last Round (0/1)  0.704**  0.299  0.709*    (0.255)  (0.249)  (0.290)  Government effectiveness (−2.5/2.5)  −0.431** (0.157)  −0.526** (0.171)  −0.232 (0.203)  Control of corruption (−2.5/2.5)  0.168 (0.143)  0.226 (0.147)  0.208 (0.178)  Voice and accountability (−2.5/2.5)  0.164* (0.0780)  0.126 (0.0797)  0.230* (0.0900)  PEPFAR focus country (0/1)  0.552***        (0.132)      Fragile state (0/1)  −0.156  −0.0283  0.0676    (0.114)  (0.117)  (0.130)  English speaking (0/1)  0.0101  0.102  0.287**    (0.105)  (0.102)  (0.109)  Constant  −297.9***  −445.5***  −529.6***    (46.96)  (53.14)  (62.12)  R2  0.761  0.723  0.801  Adjusted R2  0.752  0.708  0.790  Observations  465  334  299    (1)  (2)  (3)    HIV  TB  Malaria  ln(GDP pc), 2014 cUS$  −0.0328  −0.0119  0.0929    (0.0825)  (0.0859)  (0.0869)  ln(population)  −0.671***  −0.544***  −0.664***    (0.0293)  (0.0330)  (0.0361)  UHC tracer (0–100)  −1.089**  1.106*  −2.384***    (0.396)  (0.449)  (0.478)  Dom. health exp, %GDP  1.779  −0.196  −1.157    (2.020)  (2.125)  (2.458)  DAH Global Fund, %GDP  41.75***  3.087  7.450    (12.22)  (13.07)  (13.60)  Non-Global Fund DAH, %GDP  1.134  5.077  18.35***    (5.012)  (5.141)  (5.243)  Prevalence, %  0.0960***  1.207***  0.0224***    (0.0160)  (0.274)  (0.00334)  TB treatment success, %    −0.00691        (0.00447)    Year of Round (2002–2010)  0.154***  0.226***  0.269***    (0.0235)  (0.0266)  (0.0311)  Resubmitted proposal (0/1)  −0.0250  0.111  0.0567    (0.0798)  (0.0860)  (0.0911)  Any grant already approved (0/1)  −0.0715  −0.142  0.123    (0.0852)  (0.0892)  (0.0964)  RCC held since last Round (0/1)  0.704**  0.299  0.709*    (0.255)  (0.249)  (0.290)  Government effectiveness (−2.5/2.5)  −0.431** (0.157)  −0.526** (0.171)  −0.232 (0.203)  Control of corruption (−2.5/2.5)  0.168 (0.143)  0.226 (0.147)  0.208 (0.178)  Voice and accountability (−2.5/2.5)  0.164* (0.0780)  0.126 (0.0797)  0.230* (0.0900)  PEPFAR focus country (0/1)  0.552***        (0.132)      Fragile state (0/1)  −0.156  −0.0283  0.0676    (0.114)  (0.117)  (0.130)  English speaking (0/1)  0.0101  0.102  0.287**    (0.105)  (0.102)  (0.109)  Constant  −297.9***  −445.5***  −529.6***    (46.96)  (53.14)  (62.12)  R2  0.761  0.723  0.801  Adjusted R2  0.752  0.708  0.790  Observations  465  334  299  Dependent variable (OLS): ln(Funding request pc pa) by disease, 2014 constant US$. Data are OLS regression coefficients (SE). Sample is restricted to grant proposals from individual countries that were considered by the TRP during the rounds-based mechanism (2002–2010) and for which requested funding volumes are recorded. Sources and definitions of the variables are available in the Supplementary data. GDP, gross domestic product; pc, per capita; 2014 cUS$, constant 2014 US$; DAH, Development Assistance for Health; PEPFAR, President’s Emergency Plan for AIDS Relief; RCC, Rolling Continuation Channel. * P<0.05, ** P<0.01, *** P<0.001. Regression 2(Table 2): To address the second question and to investigate predictors of the TRP recommendations, we constructed an ordered logistic regression model regressing TRP ratings (1–4) on the same proposal and country characteristics considered in Regression 1. Note that lower TRP ratings denoted higher-quality proposals. For each disease i, the probability P that TRP rating yi took on the value l (ranging from 1 to 4) is described by   Pyi=l=Fαl-Xiβi-Fαl-1-Xiβi, (2) where F is the logistic cumulative density function, αl the threshold for the observed value l, βi the matrix of regression coefficients βij estimated (including intercept terms) and Xi denotes the matrix of regressors xij described in Table 2. Since proposal review modalities for the RCC differed from the rounds-based TRP review (Rivers 2008), we excluded RCC funding requests from this regression model. As in Regression 1, we considered factor variables for each round as well as a linear model using the year of the round as a regressor. Table 2. Predictors for TRP proposal ratings (ordered logistic model), by disease (Regression 2)   (1)  (2)  (3)    HIV  TB  Malaria  TRP rating (1–4)        ln(GDP pc), 2014 cUS$  −0.709*  −0.00656  0.591    (0.286)  (0.314)  (0.384)  ln(population)  0.0663  −0.0511  0.0766    (0.139)  (0.162)  (0.232)  UHC tracer (0–100)  0.358  −1.985  1.739    (1.372)  (1.701)  (2.358)  Dom. health exp, %GDP  6.118  3.820  −2.430    (7.162)  (8.520)  (11.00)  DAH Global Fund, %GDP  −7.291  97.02*  63.00    (45.65)  (46.12)  (59.66)  Non-Global Fund DAH, %GDP  −15.41  −16.98  19.60    (17.18)  (20.19)  (23.91)  Prevalence, %  0.0496  −1.388  0.0313    (0.0596)  (1.071)  (0.0166)  TB treatment success, %    0.0181        (0.0164)    ln(board approved total pc), 2014 cUS$  −0.239 (0.153)  −0.252 (0.204)  −0.468 (0.256)  Year of Round (2002–2010)  0.276**  0.250*  0.217    (0.0884)  (0.116)  (0.168)  Resubmitted proposal (0/1)  −0.600*  −0.477  −1.388**    (0.281)  (0.335)  (0.471)  Any grant already approved (0/1)  −7.366***  −8.004***  −8.804***    (0.776)  (1.087)  (1.197)  RCC held since last Round (0/1)  −1.008  −0.678  −1.656    (0.894)  (1.019)  (1.216)  Government effectiveness (−2.5/2.5)  −0.382 (0.537)  0.155 (0.694)  −0.294 (0.851)  Control of corruption (−2.5/2.5)  0.0364 (0.487)  0.897 (0.569)  0.0871 (0.750)  Voice and accountability (−2.5/2.5)  0.376 (0.269)  −0.109 (0.326)  0.827* (0.391)  PEPFAR focus country (0/1)  0.638        (0.481)      Fragile state (0/1)  −0.186  0.146  0.876    (0.396)  (0.432)  (0.575)  English speaking (0/1)  0.334  0.0160  −1.260*    (0.371)  (0.383)  (0.504)  Pseudo R2  0.519  0.537  0.639  Observations  464  334  299    (1)  (2)  (3)    HIV  TB  Malaria  TRP rating (1–4)        ln(GDP pc), 2014 cUS$  −0.709*  −0.00656  0.591    (0.286)  (0.314)  (0.384)  ln(population)  0.0663  −0.0511  0.0766    (0.139)  (0.162)  (0.232)  UHC tracer (0–100)  0.358  −1.985  1.739    (1.372)  (1.701)  (2.358)  Dom. health exp, %GDP  6.118  3.820  −2.430    (7.162)  (8.520)  (11.00)  DAH Global Fund, %GDP  −7.291  97.02*  63.00    (45.65)  (46.12)  (59.66)  Non-Global Fund DAH, %GDP  −15.41  −16.98  19.60    (17.18)  (20.19)  (23.91)  Prevalence, %  0.0496  −1.388  0.0313    (0.0596)  (1.071)  (0.0166)  TB treatment success, %    0.0181        (0.0164)    ln(board approved total pc), 2014 cUS$  −0.239 (0.153)  −0.252 (0.204)  −0.468 (0.256)  Year of Round (2002–2010)  0.276**  0.250*  0.217    (0.0884)  (0.116)  (0.168)  Resubmitted proposal (0/1)  −0.600*  −0.477  −1.388**    (0.281)  (0.335)  (0.471)  Any grant already approved (0/1)  −7.366***  −8.004***  −8.804***    (0.776)  (1.087)  (1.197)  RCC held since last Round (0/1)  −1.008  −0.678  −1.656    (0.894)  (1.019)  (1.216)  Government effectiveness (−2.5/2.5)  −0.382 (0.537)  0.155 (0.694)  −0.294 (0.851)  Control of corruption (−2.5/2.5)  0.0364 (0.487)  0.897 (0.569)  0.0871 (0.750)  Voice and accountability (−2.5/2.5)  0.376 (0.269)  −0.109 (0.326)  0.827* (0.391)  PEPFAR focus country (0/1)  0.638        (0.481)      Fragile state (0/1)  −0.186  0.146  0.876    (0.396)  (0.432)  (0.575)  English speaking (0/1)  0.334  0.0160  −1.260*    (0.371)  (0.383)  (0.504)  Pseudo R2  0.519  0.537  0.639  Observations  464  334  299  Dependent variable (ordered logit): TRP rating (1–4), odds ratios. Data are ordered logistic regression odds ratios (SE) without constant. Sample is restricted to grant proposals from individual countries that were considered by the TRP during the rounds-based mechanism (2002–2010). Sources and definitions of the variables are available in the Supplementary data. GDP, gross domestic product; pc, per capita; 2014 cUS$, constant 2014 US$; DAH, Development Assistance for Health; UHC, Universal Health Coverage; PEPFAR, President’s Emergency Plan for AIDS Relief; RCC, Rolling Continuation Channel. * P <0.05, ** P <0.01, *** P <0.001. Regression 3(Table 3): To combine the effects of requested funding volumes per round (dependent variable in Regression 1) with approval rates (dependent variable in Regression 2) and the frequency with which a country applied for funding, Regression 3 (Table 3) used OLS and the same functional form as Regression 1 to assess predictors of total per capita Global Fund funding received by country and by disease during the rounds-based mechanism. The dependent variable yi was defined as the natural logarithm of total per capita funding received by disease i during the rounds-based funding mechanism, including funding allocated under the RCC. The regressors xij are described in Table 3. The regressor DAH from the Global Fund was modified to exclude DAH for the disease each regression focused on. Since we were interested in the effects over the full duration of the rounds-based mechanism, the independent variables were expressed as the mean over the period 2001–2010. Table 3. Predictors for total per capita Global Fund funding during rounds-based mechanism (OLS)—by disease (Regression 3)   (1)  (2)  (3)    HIV  TB  Malaria  ln(av GDP pc), 2014 cUS$  −0.4521*  −0.3205  −0.0518    (0.1949)  (0.1815)  (0.1579)  ln(av population)  −0.5502***  −0.4719***  −0.5633***    (0.0726)  (0.0646)  (0.0652)  av UHC tracer (0–100)  0.5223  2.3904*  −3.2203***    (0.9995)  (1.0042)  (0.9206)  av Dom. health exp, %GDP  4.2055  −0.7168  −2.5929    (5.2168)  (4.3294)  (4.6890)  Global Fund DAH residual, %GDP  82.632 (72.511)  20.052 (32.666)  37.910 (42.666)  av non-Global Fund DAH, %GDP  −2.064 (16.417)  −8.261 (11.756)  19.398 (12.190)  av Prevalence, %  0.1768*  1.8773***  0.0440***    (0.0787)  (0.5438)  (0.0104)  av TB treatment success, %    0.0031        (0.0098)    av Control of corruption (−2.5/2.5)  −0.1758 (0.3847)  −0.2831 (0.3333)  −0.1246 (0.3844)  av Government effectiveness (−2.5/2.5)  −0.4891 (0.4394)  −0.2498 (0.3682)  0.2472 (0.4718)  av Voice and accountability (−2.5/2.5)  0.1614 (0.1871)  0.0563 (0.1682)  −0.1397 (0.1820)  av PEPFAR focus country (0–1)  1.3285**        (0.4978)      Fragile state (0–1)  −0.7395*  −0.3502  −0.2222    (0.3237)  (0.2749)  (0.2711)  English speaking (0/1)  0.1425  0.5373*  0.4968*    (0.2880)  (0.2119)  (0.2141)  Constant  12.6355***  8.3533***  11.5731***    (2.0848)  (1.8610)  (1.7591)  R2  0.653  0.632  0.824  Adjusted R2  0.607  0.575  0.788  Observations  111  97  72    (1)  (2)  (3)    HIV  TB  Malaria  ln(av GDP pc), 2014 cUS$  −0.4521*  −0.3205  −0.0518    (0.1949)  (0.1815)  (0.1579)  ln(av population)  −0.5502***  −0.4719***  −0.5633***    (0.0726)  (0.0646)  (0.0652)  av UHC tracer (0–100)  0.5223  2.3904*  −3.2203***    (0.9995)  (1.0042)  (0.9206)  av Dom. health exp, %GDP  4.2055  −0.7168  −2.5929    (5.2168)  (4.3294)  (4.6890)  Global Fund DAH residual, %GDP  82.632 (72.511)  20.052 (32.666)  37.910 (42.666)  av non-Global Fund DAH, %GDP  −2.064 (16.417)  −8.261 (11.756)  19.398 (12.190)  av Prevalence, %  0.1768*  1.8773***  0.0440***    (0.0787)  (0.5438)  (0.0104)  av TB treatment success, %    0.0031        (0.0098)    av Control of corruption (−2.5/2.5)  −0.1758 (0.3847)  −0.2831 (0.3333)  −0.1246 (0.3844)  av Government effectiveness (−2.5/2.5)  −0.4891 (0.4394)  −0.2498 (0.3682)  0.2472 (0.4718)  av Voice and accountability (−2.5/2.5)  0.1614 (0.1871)  0.0563 (0.1682)  −0.1397 (0.1820)  av PEPFAR focus country (0–1)  1.3285**        (0.4978)      Fragile state (0–1)  −0.7395*  −0.3502  −0.2222    (0.3237)  (0.2749)  (0.2711)  English speaking (0/1)  0.1425  0.5373*  0.4968*    (0.2880)  (0.2119)  (0.2141)  Constant  12.6355***  8.3533***  11.5731***    (2.0848)  (1.8610)  (1.7591)  R2  0.653  0.632  0.824  Adjusted R2  0.607  0.575  0.788  Observations  111  97  72  Dependent variable: ln(total signed funding pc) rounds-based mechanism, 2014 constant US$. Data are OLS regression coefficients (SE). Sample is restricted to grants approved under the rounds-based mechanism (2002–2010). Sources and definitions of the variables are available in the Supplementary data. GDP, gross domestic product; pc, per capita; 2014 cUS$, constant 2014 US$; DAH, Development Assistance for Health; PEPFAR, President’s Emergency Plan for AIDS Relief. * P <0.05, ** P <0.01, *** P <0.001. Regression 4(Table 4): TRP reports (TRP 2009) and earlier investigations (Lu et al. 2006; Radelet and Siddiqi 2007; Katz et al. 2010) suggest that grant performance ratings were poorly specified and subject to significant discretion by the secretariat (Fan et al. 2013), a point also made by the TRP (2009) and external reviewers (HLIRP 2011). In Regression 4 (Table 4), we did not attempt to resolve these issues and focused instead on investigating the relationship between TRP recommendations and the performance of resulting grants. We also considered whether fragile or English-speaking countries performed significantly differently, and whether there were any significant changes across the rounds. An ordered logit model with the same functional form as for Regression 2 was estimated for each disease i the probability P that the average Phase 1 performance rating yi took on the value l (ranging from 1 to 5 with lower values denoting stronger performance). Table 4 describes the regressors. In this way, we assessed whether TRP standards might have changed over time, particularly with repeat submissions of proposals. Table 4. Predictors for Phase 1 performance rating (ordered logistic model), by disease (Regression 4), odds ratios   (1)  (2)  (3)    HIV  TB  Malaria  Av Phase 1 perf rating, discrete (1–5)  TRP rating = 1  −0.4271  −0.0803  0.4777    (0.7115)  (0.5538)  (0.8269)  ln(board approved total pc), 2014 cUS$  0.2483 (0.1707)  0.0342 (0.2090)  −0.1113 (0.2238)  Year of Round (2002–2010)  0.1579  0.0762  0.2390    (0.1068)  (0.1238)  (0.1408)  Resubmitted proposal (0/1)  0.1208  −0.2810  −0.0769    (0.3022)  (0.3447)  (0.3740)  Any grant already approved (0/1)  −0.6664  0.2579  −0.4988    (0.4729)  (0.5856)  (0.7504)  RCC held since last Round (0/1)  0.0466  0.9869  0.1274    (0.9566)  (0.7507)  (0.9847)  PEPFAR focus country (0/1)  −0.4912        (0.5150)      Fragile state (0/1)  0.4514  −0.4183  0.1344    (0.4650)  (0.4907)  (0.5403)  English speaking (0/1)  −0.2702  −0.1301  −0.9377    (0.4197)  (0.4654)  (0.4814)  Additional control variables:  0.4256  −0.0470  1.1809**  ln(GDP pc), 2014 cUS$  (0.2913)  (0.3964)  (0.3719)  ln(population)  0.0139  0.0266  0.0354    (0.1597)  (0.1851)  (0.2171)  UHC tracer (0–100)  −3.5463*  −2.0079  −4.8589*    (1.4135)  (1.9322)  (2.2038)  Dom. health exp, %GDP  −2.462  −20.544*  −2.682    (7.554)  (9.420)  (10.209)  DAH Global Fund, %GDP  −23.410  −37.113  25.357    (41.904)  (73.230)  (46.697)  non-Global Fund DAH, %GDP  16.520  21.483  33.939    (17.919)  (22.402)  (20.005)  Prevalence, %  0.0989  −2.3241  0.0432*    (0.0844)  (1.2855)  (0.0178)  Phase 1 change mortality, abs  −1.87  4.05  12.01    (9.78)  (117.30)  (19.62)  TB treatment success, %    −0.0318        (0.0190)    P1 change TB treatm success, abs    −0.0662        (0.0339)    LFA: KPMG  0.8943  0.3021  1.2306    (0.7741)  (0.7790)  (1.0456)  LFA: PWC  −0.0264  1.0176  0.8987    (0.4885)  (0.5288)  (0.6400)  LFA: STPH  0.0476  0.7357  0.7545    (0.5809)  (0.6906)  (0.6966)  LFA: UNOPS  −0.4459  0.7423  0.1314    (0.6068)  (0.6339)  (0.9494)  PR: Government  0.0348  1.8559*  −0.4335    (0.7598)  (0.8959)  (0.8758)  PR: Local CSO  0.3742  2.8102*  −1.9810    (0.8202)  (1.1244)  (1.0200)  PR: International CSO  −0.0445  0.0050  −2.3343*    (0.8679)  (1.0168)  (1.0274)  PR: Multilateral  −1.0281  1.5553  −1.7861    (0.8212)  (0.9777)  (0.9672)  Government effectiveness (−2.5/2.5)  0.2201  −0.3888  −0.9319    (0.6393)  (0.7257)  (0.9029)  Control of corruption (−2.5/2.5)  −0.5576  −0.2166  0.0136    (0.5266)  (0.6331)  (0.7573)  Voice and accountability (−2.5/2.5)  0.1056  −0.3197  0.1104    (0.2802)  (0.3325)  (0.3956)  Pseudo R2  0.093  0.101  0.124  Observations  192  164  139    (1)  (2)  (3)    HIV  TB  Malaria  Av Phase 1 perf rating, discrete (1–5)  TRP rating = 1  −0.4271  −0.0803  0.4777    (0.7115)  (0.5538)  (0.8269)  ln(board approved total pc), 2014 cUS$  0.2483 (0.1707)  0.0342 (0.2090)  −0.1113 (0.2238)  Year of Round (2002–2010)  0.1579  0.0762  0.2390    (0.1068)  (0.1238)  (0.1408)  Resubmitted proposal (0/1)  0.1208  −0.2810  −0.0769    (0.3022)  (0.3447)  (0.3740)  Any grant already approved (0/1)  −0.6664  0.2579  −0.4988    (0.4729)  (0.5856)  (0.7504)  RCC held since last Round (0/1)  0.0466  0.9869  0.1274    (0.9566)  (0.7507)  (0.9847)  PEPFAR focus country (0/1)  −0.4912        (0.5150)      Fragile state (0/1)  0.4514  −0.4183  0.1344    (0.4650)  (0.4907)  (0.5403)  English speaking (0/1)  −0.2702  −0.1301  −0.9377    (0.4197)  (0.4654)  (0.4814)  Additional control variables:  0.4256  −0.0470  1.1809**  ln(GDP pc), 2014 cUS$  (0.2913)  (0.3964)  (0.3719)  ln(population)  0.0139  0.0266  0.0354    (0.1597)  (0.1851)  (0.2171)  UHC tracer (0–100)  −3.5463*  −2.0079  −4.8589*    (1.4135)  (1.9322)  (2.2038)  Dom. health exp, %GDP  −2.462  −20.544*  −2.682    (7.554)  (9.420)  (10.209)  DAH Global Fund, %GDP  −23.410  −37.113  25.357    (41.904)  (73.230)  (46.697)  non-Global Fund DAH, %GDP  16.520  21.483  33.939    (17.919)  (22.402)  (20.005)  Prevalence, %  0.0989  −2.3241  0.0432*    (0.0844)  (1.2855)  (0.0178)  Phase 1 change mortality, abs  −1.87  4.05  12.01    (9.78)  (117.30)  (19.62)  TB treatment success, %    −0.0318        (0.0190)    P1 change TB treatm success, abs    −0.0662        (0.0339)    LFA: KPMG  0.8943  0.3021  1.2306    (0.7741)  (0.7790)  (1.0456)  LFA: PWC  −0.0264  1.0176  0.8987    (0.4885)  (0.5288)  (0.6400)  LFA: STPH  0.0476  0.7357  0.7545    (0.5809)  (0.6906)  (0.6966)  LFA: UNOPS  −0.4459  0.7423  0.1314    (0.6068)  (0.6339)  (0.9494)  PR: Government  0.0348  1.8559*  −0.4335    (0.7598)  (0.8959)  (0.8758)  PR: Local CSO  0.3742  2.8102*  −1.9810    (0.8202)  (1.1244)  (1.0200)  PR: International CSO  −0.0445  0.0050  −2.3343*    (0.8679)  (1.0168)  (1.0274)  PR: Multilateral  −1.0281  1.5553  −1.7861    (0.8212)  (0.9777)  (0.9672)  Government effectiveness (−2.5/2.5)  0.2201  −0.3888  −0.9319    (0.6393)  (0.7257)  (0.9029)  Control of corruption (−2.5/2.5)  −0.5576  −0.2166  0.0136    (0.5266)  (0.6331)  (0.7573)  Voice and accountability (−2.5/2.5)  0.1056  −0.3197  0.1104    (0.2802)  (0.3325)  (0.3956)  Pseudo R2  0.093  0.101  0.124  Observations  192  164  139  Dependent variable (Regression 4, ordered logit): Phase 1 Performance rating (1–5), odds ratios. Data are ordered logistic regression odds ratios (SE) without constant. Sample is restricted to grant proposals from individual countries that were approved during the rounds-based mechanism (2002–2010). Sources and definitions of the variables are available in the Supplementary data. GDP, gross domestic product; pc, per capita; 2014 cUS$, constant 2014 US$; TRP, Technical Review Panel; DAH, Development Assistance for Health; UHC, Universal Health Coverage; LFA, Local Fund Agent; PWC, PriceWaterHouse Coopers; STPH, Swiss Tropical and Public Health Institute; UNOPS, United Nations Operations and Project Services; PR, Principal Recipient; CSO, Civil Society Organization; PEPFAR, President’s Emergency Plan for AIDS Relief; RCC, Rolling Continuation Channel. * P <0.05, ** P <0.01, *** P <0.001. Significance of regressors was established at P < 0.05, and each model was subjected to stepwise backward elimination of non-significant predictors (Chatterjee and Hadi 2015) to confirm robustness of predictors. Standard post-regression tests were conducted for data outliers, homoscedasticity, normality of residuals and multi-collinearity of predictors (Supplementary data). We underscore that the factors explaining funding requests, TRP recommendations and funding volumes are highly complex with potential interactions among variables and nonlinear effects, as might be the case for changes in proposal volumes over time. We therefore considered partial regression plot for all regressors and tested interaction terms among variables, but none were found to be significant. Results and discussion This section presents and discusses the results from the regressions summarized in Tables 1–4. Regression 1(Table 1): These regressions generated high adjusted R2 values. After controlling for other factors, higher disease prevalence was associated with higher requested funding volumes (P < 0.001), as would be expected from a needs-based allocation of funding. Population was negatively correlated (P < 0.001) with per capita funding requests, suggesting that large countries exercised financial suppression in Global Fund proposals. This finding was robust under several different specifications and is consistent with the development economics literature (Younas 2008; Bourguignon et al. 2009; Temple 2010), through other studies into DAH (Lu et al. 2010) have not controlled for population size. Coefficients for resubmitted proposals were not significant, so countries did not reduce funding requests following an initial rejection by the TRP. If countries believed the TRP exercised financial suppression they would be expected to reduce funding requests upon resubmission to increase the likelihood of a TRP recommendation. Proposals that followed the successful approval of a first grant to the country did not request significantly different funding volumes. The RCC dummy is significantly associated with higher funding requests for HIV/AIDS and malaria, suggesting that the RCC mechanism did deliver additional resources for high-performing countries. The year of the round was correlated with larger funding requests (P < 0.001) consistent with a scaling-up of program size over time. The coefficient was small, but since over this period the real cost of disease interventions, such as ART (Stover et al. 2011) or malaria interventions (Zelman et al. 2014), fell sharply, the evidence suggests that countries designed their proposals around a substantial scaling-up of interventions. Augmented partial residual plots (Supplementary data) show that ceteris paribus the volume of funding requests increased without major nonlinearities, a finding that is confirmed by alternative specifications that replace the year of round variable with dummy variables for each round (Supplementary data). GDP per capita and domestic health spending were not significantly associated with funding requests, and DAH from non-Global Fund donors was a significant predictor only for higher malaria funding requests. Global Fund DAH was significantly associated with higher funding requests for HIV/AIDS proposals. The association with the strength of health systems, as measured by the Universal Health Care (UHC) tracer, is mixed. Countries with a higher UHC score requested lower per capita funding for HIV/AIDS and malaria (P < 0.01), a finding that is consistent with high unmet financing needs for health system strengthening, as reported widely in the literature (Carrin et al. 2010; Bowser et al. 2014; iERG 2014). Yet, in the case of TB, the association has the opposite sign (P < 0.05). These findings are consistent with Regression 3 below, and they are robust to stepwise elimination (Supplementary data), including the elimination of domestic health spending, which is highly correlated with the UHC tracer. Dropping the UHC tracer from the regressions does not alter the significance of the other coefficients, and replacing the UHC tracer with the density of physicians, a widely used measure for health services (Radelet and Siddiqi 2007; Katz et al. 2010), generates similar results. This evidence reduces the likelihood of a spurious association. A speculative interpretation of the positive association for TB might be that stronger health systems allow for the design of larger-scale programs and that in the case of TB this effect dominates the need for incremental investments to strengthen weak health systems. This issue requires closer scrutiny, possibly by using measures for the specific components of health systems required for operating control and treatment programs for each disease. Countries with lower government effectiveness requested more funding for HIV/AIDS and TB, but this did not translate into higher approved funding volumes (Regression 3). Countries with a higher score on the governance variable ‘voice and accountability’ requested more funding, but this effect was only significant for HIV/AIDS and TB and increased after removing less significant predictors from the model (Supplementary data). Control of corruption was not a significant predictor. PEPFAR focus countries requested significantly more funding per capita (P < 0.001). PEPFAR support to countries would have two competing impacts on funding requests to the Global Fund. On the one hand, PEPFAR reduces residual financing needs for HIV/AIDS programs, but on the other it increases the system capacity for HIV/AIDS program design and execution, which enables PEPFAR countries to scale up programs rapidly (Shakow 2006). On balance, the impact of greater technical capacity appears to have outweighed lower residual funding needs compared with non-PEPFAR countries, a conclusion that is supported by country case studies in Mozambique, Tanzania, and Uganda (Oomman et al. 2007). This finding is also consistent with repeated concerns flagged by the TRP about the low quality of technical assistance to HIV/AIDS grants (TRP 2009), which would help explain the weaker performance of non-PEPFAR countries. On balance, a limiting factor on scaling up appears to be greater funding for technical assistance. Even more significantly, the findings suggest that countries operated below their maximum absorptive capacity for HIV/AIDS programs during the rounds-based mechanism, since additional PEPFAR funding generated a greater demand for Global Fund resources, as discussed further under Regression 2. Contrary to findings elsewhere in the development economics literature (Collier and Dollar 2002; Temple 2010), fragile countries were not associated with lower funding requests to the Global Fund. The dummy for English-speaking countries was not significant, except for a positive correlation of the English-speaking country dummy with malaria funding requests. Regression 2(Table 2): Few predictors were significant in explaining TRP ratings. Funding proposals from countries with large populations were not associated significantly with poorer ratings, suggesting that the TRP did not exercise financial suppression contrary to the prevailing practice of most donors (Bourguignon et al. 2009). This finding is robust to different specifications of Regression 2. It contrasts with the highly significant association of population with the dependent variables in Regressions 1 and 3. After a first grant had been approved by the Global Fund, subsequent proposals from the same country became associated with better TRP ratings (p < 0.001). In addition, resubmitted proposals received better TRP ratings, but the association was not significant for TB proposals. This evidence is consistent with either a lowering of TRP standards for subsequent proposals, a learning effect leading to higher-quality proposals, or a spurious association since high-capacity proponents are more likely to prepare better proposals, which are more likely to be approved over time. Such a spurious association appears less likely because upon removing the variable ‘Any grant already approved’, the significance of ‘Resubmitted proposal’ rises for all three diseases (p < 0.001) without any other major changes to the model results. Since average TRP ratings were positively associated with the year of the round (significant for malaria proposals after stepwise elimination), countries appear to have improved the quality of repeat submissions. Evidence from Regression 4 shows that dummies for resubmitted proposals or proposals following the approval of the first grant for the country were not significantly associated with grant performance ratings, which supports the interpretation that the TRP did not lower its standards for such proposals, and instead we see evidence of learning by countries. The interpretation that the TRP promoted substantial learning in disease program design and implementation is supported by the literature on the Global Fund (Stover et al. 2011; Jamison et al. 2013; Bridge et al. 2016; van Kerkhoff and Szlezák 2016). For example, following two subsequent rejections of its HIV/AIDS proposals by the TRP, China reformed its approach to managing the disease by including international best practice on harm reduction. TRP requirements to adhere to medical best practice by involving communities and people living with the diseases in program design and implementation, also had a deep impact on China’s malaria control program (Wang et al. 2014; Minghui et al. 2015). More broadly, evidence-based funding decisions mediated by the TRP have strengthened harm reduction across the world (Atun and Kazatchkine 2010; Bridge et al. 2016). Similarly, the TRP helped identify gaps in available health interventions, increased adoption, strengthened program design, and drove down the cost of solutions, such as rapid-diagnostic tests for malaria (Zhao et al. 2012), LLINs (WHO 2007; Noor et al. 2009; Zelman et al. 2014), the shift to ACTs (Cohen et al. 2008; Roll Back Malaria Partnership 2008), and ART (Stover et al. 2011). Countries that had been invited to participate in the RCC mechanism, did receive better proposal ratings, but the association is not significant at 5%, even after stepwise elimination. Per capita GDP was not a significant predictor of TRP ratings except for HIV/AIDS proposals where income correlated with better TRP ratings (p < 0.05). Likewise, TRP ratings were not significantly associated with governance variables except for malaria grants where the coefficient for ‘voice and accountability’ was positive (p < 0.05). In line with Bornemisza et al. (2010), there is no evidence that the TRP discriminated against fragile countries or that these countries had greater difficulties in submitting high-quality proposals. English-speaking countries were more likely to receive better TRP ratings for malaria grants (p < 0.05). The relationship was not significant for other diseases. After stepwise elimination of insignificant predictors, PEPFAR countries were significantly associated with better proposal ratings, reinforcing the interpretation (Regression 1) that they had greater capacity to propose high-quality proposals and that residual funding needs were large. Domestic health spending and DAH were not significantly associated with TRP ratings, except for Global Fund DAH for TB, which correlated with worse TRP ratings. Regression 3 shows that this did not translate into a significant effect on overall grant volumes for TB. The UHC tracer, TB treatment success, and prevalence rates were not significant predictors of TRP ratings. On balance, TRP recommendations were not correlated with funding needs measured by disease prevalence, and poorer countries were significantly more likely to have their HIV/AIDS proposals accepted. These findings are robust to different specifications of the regressors reported in the literature (Radelet and Siddiqi 2007; Katz et al. 2010; Lu et al. 2010; Fan et al. 2013), suggesting that TRP decisions were made primarily on the basis of factors independent of country characteristics (except disease prevalence) and funding volumes. This finding supports the hypothesis that TRP recommendations were based on the intrinsic technical quality of the proposals, consistent with the TRP’s mandate during the rounds-based mechanism. Regression 3 (Table 3): Cumulative funding volumes were positively associated with disease prevalence for each disease (P < 0.05), suggesting that the TRP mechanism solicited and recommended high-quality funding requests from higher-burden countries. The association with per capita GDP was significantly negative for HIV/AIDS, and the negative coefficient became significant for TB but not for malaria after stepwise elimination. After controlling for other factors, the Global Fund directed higher per capita funding towards poorer countries. Population size was negatively associated with overall funding volumes (P < 0.001). This finding is robust across different specifications of the model. It suggests that the self-suppression of funding requests from larger countries (Regression 1) resulted in lower funding volumes even though we could identify no evidence of financial suppression in TRP recommendations (Regression 2). These findings challenge the conclusion of the 2011 review of Global Fund operations that ‘large countries put forward enormous proposals’ (HLIRP 2011). As in Regression 1, the UHC tracer generated significant associations with opposing signs for TB and malaria. A stronger health system was associated with higher TB funding, consistent with the hypothesis that more effective health systems support a greater scaling-up of TB interventions. Meanwhile, the association was negative for malaria grants (P < 0.001), possibly due to the dominant role played by the Global Fund in financing malaria LLINs, which could be distributed outside health systems (Hafner and Shiffman 2013; de Jongh et al. 2014) and were most needed in poorer countries that tended to have weaker health systems. These findings reiterate the need for additional statistical analyses of the role of health systems in enabling the scaling up of disease programs. Global Fund funding did not substitute for domestic health expenditure. The associations with DAH and governance variables were also non-significant. Funding for HIV/AIDS was complementary to PEPFAR funding since PEPFAR countries were positively associated with higher per capita funding volumes, reflecting both larger (Regression 1) and higher-quality (Regression 2) proposals. The fragile states dummy was not significantly associated with per capita funding volumes, except for a negative association with HIV/AIDS grants, which is robust to removing the PEPFAR dummy. Considering evidence from the first two regressions, fragile countries submitted fewer proposals, but their proposals did not receive worse TRP ratings. This suggests greater scope for technical assistance to fragile countries to accelerate proposal design. Once again the differential performance of fragile countries on HIV/AIDS grants is consistent with the TRP’s concerns about low-quality technical assistance for HIV/AIDS (TRP 2009). English-speaking countries received more funding (P < 0.05) for TB and malaria grants, but the correlation was not significant for HIV/AIDS. This lends support to concerns (Kerouedan 2010; French Ministry of Foreign Affairs 2013) that non-English-speaking countries were less successful at attracting Global Fund funding. For malaria grants Regressions 1 and 2 suggest that English-speaking countries submitted larger proposals that were more likely to be approved by the TRP. In the case of TB, English-speaking countries submitted proposals more frequently. Regression 4(Table 4): In spite of a large number of regressors, this ordered logistic model yielded few significant associations, possibly due to poorly specified grant proposal ratings during the rounds-based mechanism, as suggested by Fan et al. (2013). As a result, only weak inferences can be drawn from the findings, and we limit ourselves to the role of proposal and grant characteristics considered in Regressions 1–3. After controlling for other factors, average Phase 1 performance ratings did not correlate significantly with TRP recommendations. A plausible explanation is that all proposals rated 2 were revised with guidance from the TRP and the Global Fund secretariat, which increased their quality to a point where they matched or exceeded that of proposals rated 1. Grants resulting from resubmitted proposals or following the approval of an earlier grant did not perform significantly differently. This suggests that higher TRP ratings for such proposals (Regression 2) did not reflect lower TRP assessment standards but signified higher-quality proposals. We found no evidence that grants to fragile countries performed significantly differently from grants to non-fragile countries. Using a smaller dataset, an earlier study (Bornemisza et al. 2010) found a small, negative correlation between grant performance and fragility. Regressions 1–3 suggest that, with the exception of HIV/AIDS, fragile countries were not associated with lower funding requests and funding volumes. We therefore conclude that the Global Fund managed to generate quality demand in the difficult operating environments that are commonly associated with low aid allocation and low grant performance (Collier and Dollar 2002; Collier 2008; Temple 2010). PEPFAR focus countries and English-speaking countries were not associated with significant differences in performance ratings. The latter suggests that the Global Fund overcame language barriers during grant implementation even though English-speaking countries obtained higher funding volumes for some diseases. Limitations The regressions reported in this study investigate associations that are highly complex and depend on a large number of factors with possible interaction terms. Even though the R2 are relatively high, the reported associations could also be due to factors not considered in the models. In particular, we did not have access to data on the intrinsic quality of grant proposals, which could be gathered through structured reviews of random samples of grant proposals. Also, Global Fund practices (e.g. TRP procedures, evaluation standards, the role of the Technical Partners, modalities for Principal Recipients and Local Fund Agents) are widely reported to have improved over time (HLIRP 2011; French Ministry of Foreign Affairs 2013; DFID 2016), and these effects may not have been picked up fully in the regression analyses. Though we did not find any significant relationships for the most common interaction terms reported in the development economics literature (Temple 2010), it is possible that missing interaction terms affect the results. As discussed in the findings, the data quality of grant performance ratings has been questioned, so Regression 4 could only establish the plausibility of consistent TRP assessment standards during the rounds-based mechanism. Historic data on intervention coverage during the rounds-based mechanism was too sparse to construct comprehensive models of Global Fund grant performance. Too few proposals on health systems strengthening were considered during the rounds-based mechanism, so the Global Fund’s impact on health systems could not be considered in this study. Conclusion During the rounds-based mechanism (2002–2010) per capita funding requests to the Global Fund were correlated with disease prevalence rates and they increased over time in the presence of sharp falls in the cost of interventions, consistent with a significant scaling-up of programs in fragile and non-fragile countries alike. The data suggest that the Global Fund promoted substantial learning and improvements in the quality of country proposals, which is supported by findings in other parts of the public health literature. It is likely that the TRP’s transparent rating of proposals, the release of findings from each funding round, and the systematic review of lessons learnt with Technical Partners (e.g. WHO, UNAIDS, Stop TB and Roll-Back Malaria) contributed to propagating knowledge on scaling up interventions across countries. The country-led funding model appears to have encouraged scaling-up since countries were not constrained in the volume of per capita funding they could request, and success in one country inspired others to follow. The data do not support the hypothesis that large countries put forward very large proposals to the Global Fund, which served as a key justification for the shift from rounds-based to the New Funding Model. Instead, per capita funding to and funding requests from countries with larger populations were financially suppressed, even though TRP recommendations were unaffected by population size. This suggests significant unmet funding needs, particularly in countries with larger populations. Large funding requests for HIV/AIDS from PEPFAR-eligible countries lend further support to the presence of large unmet financing gaps during the rounds-based mechanism. Throughout the rounds-based mechanism, the TRP fulfilled its role. It appears to have recommended proposals without regard to funding volumes, population size, or other country characteristics typically associated with different aid volumes, such as governance, domestic health expenditure, and DAH. Global Fund funding went towards higher-burden countries with lower incomes, consistent with the guiding principles of the Global Fund. The findings suggest that the Global Fund, working with its Technical Partners, has been effective at overcoming lower capacity to design and implement programs in poorer and/or fragile countries. Fragile countries requested funding volumes that were not significantly different from those requested by other countries; they were as likely to have their funding requests approved by the TRP; and their grants performed equally well except for HIV/AIDS grants. The TRP flagged concerns about the low quality of technical assistance for HIV/AIDS grants, which might explain the differential performance of fragile countries here. Conversely, higher funding requests and better ratings of proposals from PEPFAR countries suggest that countries that receive greater support in strengthening and scaling-up their response to HIV/AIDS can attract more funding from the Global Fund. This in turn indicates the presence of large unmet funding needs in non-PEPFAR countries, which technical assistance might be able to convert into high-quality proposals. The Global Fund’s success in operating across the full spectrum of country environments, including fragile states, may set an example for financing health systems and other investment priorities under the SDGs. English-speaking countries obtained higher volumes of funding for malaria and TB. TRP reports frequently referred to quality problems with the translation of proposals and supporting documents into English. The Global Fund and other mechanisms must ensure that language does not become a barrier to accessing financing. Taken together, the evidence shows that a demand-based funding mechanism relying on independent technical review of proposals without ex-ante country allocations can generate needs-based funding allocations. It can stimulate quality demand even in poor and fragile countries contrary to widespread expectations when the Global Fund was established in 2002. Pooled international funding mechanisms in other areas, such as the Global Environment Facility, the Green Climate Fund, the Global Partnership for Education, or the International Fund for Agricultural Development, should study the workings and performance of the TRP and may consider establishing similar procedures. Further work is needed to investigate how lessons from the TRP can be applied to other sectors and to understand how the role of the TRP has evolved under the New Funding Model. Supplementary data Supplementary data are available at Health Policy and Planning online. Acknowledgements The Swedish International Development Cooperation Agency (Sida) provided funding for this study. Jeffrey Sachs, Ekko van Ierland, and Jeroen Klomp advised on the design of the study. Christoph Benn, Lucie Blok, Sofia Cordero, David Durand-Delacre, George Gotsadze, Ilze Kalnina, and Katerina Teksoz have provided valuable comments. Three anonymous referees provided valuable comments. David Durand-Delacre and Murad Hirji assisted with data collection and preparation. The views expressed are those of the author and do not represent those of Sida or of the Sustainable Development Solutions Network (SDSN). Funding The research was funded by the Sustainable Development Solutions Network using a grant from the Swedish International Development Cooperation (Sida). Conflict of interest statement. None declared. References Atun R, Kazatchkine M. 2009. Promoting country ownership and stewardship of health programs: The Global Fund experience. Journal of Acquired Immune Deficiency Syndromes  52: S67– 8. Google Scholar CrossRef Search ADS PubMed  Atun R, Kazatchkine M. 2010. The Global Fund’s leadership on harm reduction: 2002–2009. International Journal of Drug Policy  21: 103– 6. Google Scholar CrossRef Search ADS PubMed  Bornemisza O, Bridge J, Olszak-Olszewski M, Sakvarelidze G, Lazarus JV. 2010. Health aid governance in fragile states: the Global Fund experience. Global Health Governance  4: 1– 18. Bourguignon F, Levin V, Rosenblatt D. 2009. International redistribution of income. World Development  37: 1– 10. Google Scholar CrossRef Search ADS   Bowser D, Sparkes SP, Mitchell A. 2014. Global Fund investments in human resources for health: innovation and missed opportunities for health systems strengthening. Health Policy and Planning  29: 986– 97. Google Scholar CrossRef Search ADS PubMed  Bridge J, Hunter BM, Albers E et al.   2016. The Global Fund to Fight AIDS, Tuberculosis and Malaria’s investments in harm reduction through the rounds-based funding model (2002–2014). International Journal of Drug Policy  27: 132– 7. Google Scholar CrossRef Search ADS PubMed  Car J, Paljärvi T, Car M et al.   2012. Negative health system effects of Global Fund’s investments in AIDS, tuberculosis and malaria from 2002 to 2009: systematic review. JRSM Short Reports  3: 70. Google Scholar CrossRef Search ADS PubMed  Carrin G, Buse K, Heggenhougen K, Quah SR. 2010. Health Systems Policy, Finance, and Organization . Amsterdam: Elsevier Academic Press. Chatterjee S, Hadi AS. 2015. Regression Analysis by Example . Hoboken, NJ: John Wiley & Sons. Cohen JM, Singh I, O'Brien ME. 2008. Predicting Global Fund grant disbursements for procurement of artemisinin-based combination therapies. Malaria Journal  7: 200. Google Scholar CrossRef Search ADS PubMed  Collier P. 2008. The Bottom Billion: Why the Poorest Countries Are Failing and What Can Be Done about It . New York: Oxford University Press. Collier P, Dollar D. 2002. Aid allocation and poverty reduction. European Economic Review  46: 1475– 500. Google Scholar CrossRef Search ADS   Daniels N, Sabin JE. 2008. Setting Limits Fairly: Learning to Share Resources for Health . Oxford: Oxford University Press. de Jongh TE, Harnmeijer JH, Atun R et al.   2014. Health impact of external funding for HIV, tuberculosis and malaria: systematic review. Health Policy and Planning  29: 650– 62. Google Scholar CrossRef Search ADS PubMed  DFID. 2016. Raising the Standard: The Multilateral Development Review 2016 . Department for International Development, London. Fan VY, Duran D, Silverman R, Glassman A. 2013. Performance-based financing at the Global Fund to Fight AIDS, Tuberculosis and Malaria: an analysis of grant ratings and funding, 2003–12. The Lancet Global Health  1: e161– 8. Google Scholar CrossRef Search ADS PubMed  French Ministry of Foreign Affairs. 2013. French contributions to the Global Fund to Fight AIDS, Tuberculosis and Malaria: evaluation report synthesis. No. 126. French Ministry of Foreign Affairs, Paris. GBD. 2016. Measuring the health-related Sustainable Development Goals in 188 countries: a baseline analysis from the Global Burden of Disease Study 2015. The Lancet  388: 1813– 50. CrossRef Search ADS   GFATM. 2001. The framework document. Global Fund to Fight AIDS, Tuberculosis and Malaria, Geneva. GFATM. 2008. Report of the Technical Review Panel and the Secretariat on funding recommendations for Wave 3 Rolling Continuation Channel proposals. Global Fund to Fight AIDS, Tuberculosis and Malaria, Geneva. GFATM. 2013a. The Global Fund’s New Funding Model. Global Fund to Fight AIDS, Tuberculosis and Malaria, Geneva. GFATM. 2013b. Guidelines and requirements for Country Coordinating Mechanisms. Global Fund to Fight AIDS, Tuberculosis and Malaria, Geneva. GFATM. 2016. List of successful appeals in previous rounds. Global Fund to Fight AIDS, Tuberculosis and Malaria, Geneva. Hafner T, Shiffman J. 2013. The emergence of global attention to health systems strengthening. Health Policy and Planning  28: 41– 50. Google Scholar CrossRef Search ADS PubMed  HLIRP. 2011. Turning the page from emergency to sustainability: the final report of the High-Level Independent Review Panel on Fiduciary Controls and Oversight Mechanisms of the Global Fund to Fight AIDS, Tuberculosis and Malaria. High-Level Independent Review Panel on Fiduciary Controls and Oversight Mechanisms of the Global Fund to Fight AIDS, Tuberculosis and Malaria, Geneva. iERG. 2014. Every woman, every child: a post-2015 vision. Independent Expert Review Group on Information and Accountability for Women’s and Children’s Health, Geneva. IHME. 2016. Development Assistance for Health Database 1990–2015. Jamison DT, Summers LH, Alleyne G. 2013. Global health 2035: a world converging within a generation. The Lancet  382: 1898– 955. Google Scholar CrossRef Search ADS   Katz I, Aziz MA, Olszak-Olszewski M et al.   2010. Factors influencing performance of Global Fund-supported tuberculosis grants. The International Journal of Tuberculosis and Lung Disease  14: 1097– 103. Google Scholar PubMed  Kerouedan D. 2010. Enjeux politiques de l’évaluation à cinq ans du Fonds mondial de lutte contre le sida, la tuberculose et le paludisme. Bulletin de la Société de Pathologie Exotique  103: 119– 22. Google Scholar CrossRef Search ADS PubMed  Lu C, Michaud CM, Khan K, Murray CJ. 2006. Absorptive capacity and disbursements by the Global Fund to Fight AIDS, Tuberculosis and Malaria: analysis of grant implementation. The Lancet  368: 483– 8. Google Scholar CrossRef Search ADS   Lu C, Schneider MT, Gubbins P et al.   2010. Public financing of health in developing countries: a cross-national systematic analysis. The Lancet  375: 1375– 87. Google Scholar CrossRef Search ADS   Macro International. 2009. The Five-Year Evaluation of the Global Fund to Fight AIDS, Tuberculosis, and Malaria: synthesis of study areas 1, 2 and 3. Macro International, Geneva. McArthur JW, Rasmussen K. 2017. Change of Pace: Accelerations and Advances During the Millennium Development Goal Era . Washington, DC: Brookings Institution. Minghui R, Scano F, Sozi C, Schwartländer B. 2015. The Global Fund in China: success beyond the numbers. The Lancet Global Health  3: e75– 7. Google Scholar CrossRef Search ADS PubMed  Noor AM, Mutheu JJ, Tatem AJ, Hay SI, Snow RW. 2009. Insecticide-treated net coverage in Africa: mapping progress in 2000–07. The Lancet  373: 58– 67. Google Scholar CrossRef Search ADS   Oomman N, Bernstein M, Rosenzweig S. 2007. Following the Funding for HIV/AIDS: A Comparative Analysis of the Funding Practices of PEPFAR, the Global Fund and World Bank MAP in Mozambique, Uganda and Zambia . Washington, DC: Center for Global Development. Ottersen T, Kamath A, Moon S, Martinsen L, Røttingen J-A. 2017. Development assistance for health: what criteria do multi-and bilateral funders use? Health Economics, Policy and Law  12: 223– 44. Google Scholar CrossRef Search ADS   Patel P, Cummings R, Roberts B. 2015. Exploring the influence of the Global Fund and the GAVI Alliance on health systems in conflict-affected countries. Conflict and Health  9: 7. Google Scholar CrossRef Search ADS PubMed  Radelet S, Siddiqi B. 2007. Global Fund grant programmes: an analysis of evaluation scores. The Lancet  369: 1807– 13. Google Scholar CrossRef Search ADS   Rivers B. 2008. Rolling Continuation Channel (RCC), 2nd update. Aidspan. Roll Back Malaria Partnership. 2008. The Global Malaria Action Plan. For a Malaria-Free World . Geneva: Roll Back Malaria Partnership. Sachs JD, Schmidt-Traub G. 2017. Global Fund lessons for Sustainable Development Goals. Science  356: 32– 3. Google Scholar CrossRef Search ADS PubMed  Samb B, Evans T, Dybul M et al.   2009. An assessment of interactions between global health initiatives and country health systems. The Lancet  373: 2137– 69. Google Scholar CrossRef Search ADS   Schmidt-Traub G. 2015. Investment needs to achieve the Sustainable Development Goals. SDSN Working Paper. Sustainable Development Solutions Network, Paris. Shakow A. 2006. Global Fund - World Bank HIV/AIDS Programs comparative advantage study. World Bank and Global Fund to Fight AIDS, Tuberculosis and Malaria, Washington D.C. and Geneva. Stover J, Korenromp EL, Blakley M et al.   2011. Long-term costs and health impact of continued Global Fund support for antiretroviral therapy. PLoS ONE  6: e21048. Google Scholar CrossRef Search ADS PubMed  Temple JRW. 2010. Aid and conditionality. In: Handbook of Development Economics . Amsterdam: Elsevier, 4415– 523. Google Scholar CrossRef Search ADS   TRP. 2009. Report of the Technical Review Panel and the Secretariat on Round 9 proposals. Global Fund to Fight AIDS, Tuberculosis and Malaria, Geneva. van Kerkhoff L, Szlezák NA. 2006. Linking local knowledge with global action: examining the Global Fund to Fight AIDS, Tuberculosis and Malaria through a knowledge system lens. Bulletin of the World Health Organization  84: 629– 35. Google Scholar CrossRef Search ADS PubMed  van Kerkhoff L, Szlezák NA. 2016. The role of innovative global institutions in linking knowledge and action. Proceedings of the National Academy of Sciences  113: 4603– 8. Google Scholar CrossRef Search ADS   Wang R-B, Zhang Q-F, Zheng B et al.   2014. Transition from control to elimination: impact of the 10-year global fund project on malaria control and elimination in China. Advances in Parasitology  86: 289– 318. Google Scholar CrossRef Search ADS PubMed  WHO. 2007. WHO Releases New Guidance on Insecticide-Treated Mosquito Nets . Geneva: World Health Organization. Wood A. 2008. Looking ahead optimally in allocating aid. World Development  36: 1135– 51. Google Scholar CrossRef Search ADS   World Bank. 1998. Assessing Aid: What Works, What Doesn’t, and Why . Oxford: Oxford University Press. World Bank. 2016. Worldwide Governance Indicators database. Younas J. 2008. Motivation for bilateral aid allocation: Altruism or trade benefits. European Journal of Political Economy  24: 661– 74. Google Scholar CrossRef Search ADS   Zelman B, Kiszewski A, Cotter C, Liu J. 2014. Costs of eliminating malaria and the impact of the Global Fund in 34 countries. PLoS ONE  9: e115714. Google Scholar CrossRef Search ADS PubMed  Zhao J, Lama M, Korenromp E et al.   2012. Adoption of rapid diagnostic tests for the diagnosis of malaria, a preliminary analysis of the Global Fund program data, 2005 to 2010. PLoS ONE  7: e43549. Google Scholar CrossRef Search ADS PubMed  © The Author(s) 2018. Published by Oxford University Press in association with The London School of Hygiene and Tropical Medicine. All rights reserved. For Permissions, please e-mail: journals.permissions@oup.com http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Health Policy and Planning Oxford University Press

The role of the Technical Review Panel of the Global Fund to Fight HIV/AIDS, Tuberculosis and Malaria: an analysis of grant recommendations

Loading next page...
 
/lp/ou_press/the-role-of-the-technical-review-panel-of-the-global-fund-to-fight-hiv-jnHnWGrFeK
Publisher
Oxford University Press
ISSN
0268-1080
eISSN
1460-2237
D.O.I.
10.1093/heapol/czx186
Publisher site
See Article on Publisher Site

Abstract

Abstract The independent Technical Review Panel (TRP) of the Global Fund to Fight HIV/AIDS, Tuberculosis and Malaria is a unique mechanism to review funding proposals and to provide recommendations on their funding. Its functioning and performance have received little attention in the scientific literature. We aimed to identify predictors for TRP recommendations, whether these were in line with the Global Fund’s ambition to give priority to countries most in need, and whether they correlated with grant performance. We combined data on proposals and applications under the Rolling Continuation Channel, TRP recommendations and grant implementation during the rounds-based mechanism (2002–2010) with country characteristics. Ordered logistic and OLS regressions were used to identify predictors for per-capita funding requests, TRP recommendations, Global Fund funding and grant performance ratings. We tested for financial suppression of large funding proposals and whether fragile or English-speaking countries performed differently from other countries. We found that funding requests and TRP recommendations were consistent with disease burden, but independent of other country characteristics. Countries with larger populations requested less funding per capita, but there is no evidence of financial suppression by the TRP. Proposals from fragile countries were as likely to be recommended as proposals from other countries, and resulting grants performed equally well except for lower performance of HIV/AIDS grants. English-speaking countries obtained more funding for TB and malaria than other countries. In conclusion, the independent TRP acted in line with the guiding principles of the Global Fund to direct funding to countries most in need without ex ante country allocation. The Global Fund appears to have promoted learning on how to design and implement large-scale programs in fragile and non-fragile countries. Other pooled financing mechanisms may consider TRP operating principles to generate high-quality demand, to promote learning and to direct resources to countries most in need. Global fund to fight AIDS, tuberculosis and malaria, technical review panel, health financing, sustainable development goals, millennium development goals, decision making, developing countries, evidence-based policy, health planning, health systems, international health policy, overseas development assistance, resource allocation Key Messages The Technical Review Panel (TRP) of the Global Fund to Fight AIDS, Tuberculosis and Malaria is a unique body ensuring that funding decisions are based on independent technical review of program proposals. It has received little attention in the peer reviewed literature. During the rounds-based mechanism, the TRP recommended Global Fund proposals without regard to funding volumes, population size, governance, domestic health expenditure or DAH, making Global Fund funding needs based without ex ante funding allocations. The Global Fund, working with its Technical Partners, was effective at overcoming lower capacity to design and implement programs in poorer and/or fragile countries. This success holds important lessons for financing health systems. Pooled international funding mechanisms in other sectors, such as the Global Environment Facility, the Green Climate Fund, the Global Partnership for Education or the International Fund for Agricultural Development, should study the workings and performance of the TRP and may consider establishing similar procedures. Introduction Since its establishment in 2001, the Global Fund to Fight AIDS, Tuberculosis and Malaria (henceforth the Global Fund) has disbursed $34.6 billion (in 2014 US$) between 2013 and 2015 (IHME 2016), making it the dominant external funding channel for the three diseases except for HIV/AIDS, where it is second to the US President’s Emergency Plan for AIDS Relief (PEPFAR). In late 2016, the Global Fund mobilized some $13 billion at a time when international aid budgets were under severe pressure owing to the refugee crisis in Europe and macroeconomic pressures on donor countries. The success of the Global Fund in generating quality demand for health financing and in mobilizing substantial volumes of development assistance for health (DAH) has been attributed to its unique design principles (Sachs and Schmidt-Traub 2017). These include a focus on ‘national ownership and respect [for] country-led formulation and implementation’; the evaluation of ‘proposals through independent review processes based on the most appropriate scientific and technical standards’; giving ‘due priority to the most affected countries’; and ‘linking resources to the achievement of clear, measurable and sustainable results’ (GFATM 2001). The Global Alliance for Vaccines and Immunization (Gavi) applies similar design principles and has been equally successful in increasing quality DAH demand for vaccines. The Global Fund’s independent Technical Review Panel (TRP), which reviews proposals and makes funding recommendations to the Board, represents a critical innovation (Sachs and Schmidt-Traub 2017). Among all institutions reviewed in the 2016 UK Multilateral Development Review (DFID 2016), Gavi is the only other mechanism that operates an independent review panel. The Global Fund Board has always followed TRP funding recommendations, and every Board-approved country program has been fully funded. So, although the Board took final funding decisions, they have been determined by the TRP. During the rounds-based funding mechanism (2002–2010), the TRP was asked to issue recommendations without ex ante country allocations based on the technical merit of each proposal. Recommendations were made by consensus on a scale of 1–4 (1—recommended; 2—recommended subject to clarifications; 3—not recommended but strongly encouraged to re-submit; 4—not recommended). During the first 10 rounds 42% of proposals were rated 1 or 2. To maintain the technical independence of the evaluation process and to avoid interference with political considerations relating to individual countries, the Global Fund Board voted en bloc on the entire set of TRP recommendations. Countries could appeal against funding decisions, and for most rounds some appeals were granted (GFATM 2016). The Rolling Continuation Channel (RCC) was introduced in November 2006 to allow grantees with strongly performing grants to apply for continuing funding using a simplified procedure under which the TRP reviewed proposals in ‘waves’ held approximately twice per year. As with the rounds-based mechanism, the TRP could invite unsuccessful proposals to reapply in a later wave. Yet, the TRP soon expressed concerns that the RCC overlapped with the rounds-based mechanism and risked fragmenting country efforts (GFATM 2008). The RCC was discontinued after eight waves in August 2010. After 2011, the Global Fund introduced the New Funding Model under which countries can apply for pre-agreed volumes of funding with modest additional ‘incentive funding’ (GFATM 2013a). TRP recommendations now prioritize interventions within a pre-agreed resource envelope. This diminished the discretion of countries and the TRP to set funding volumes, so we limit our analysis to the rounds-based mechanism. Several studies have investigated the performance of Global Fund grants (Lu et al. 2006; Radelet and Siddiqi 2007; Macro International 2009; Katz et al. 2010; Fan et al. 2013), including in fragile countries (Bornemisza et al. 2010; Patel et al. 2015), and their impact on health systems (Samb et al. 2009; Car et al. 2012; de Jongh et al. 2014; iERG 2014). There has been no systematic assessment of the role of the TRP. This gap in the literature is notable since the TRP constitutes a novel mechanism for allocating aid, a key challenge considered in the health financing (Ottersen et al. 2017) and the development economics literature (World Bank 1998; Temple 2010). Processes for aid allocation ought to balance between criteria for countries’ need and the effectiveness with which they can use incremental funding (Collier and Dollar 2002; Daniels and Sabin 2008; Wood 2008; Ottersen et al. 2017). The aid allocation challenge for the three diseases was further compounded by developing countries’ widespread lack of quality demand, expressed as high-quality diseases management strategies, when the Global Fund was established in 2002 (Lu et al. 2006; Atun and Kazatchkine 2009). To elucidate the role of the TRP, this study addresses three sets of questions that are central to understanding the demand-based funding model of the Global Fund. First, we consider which factors explain the volume of funding requested by countries for each disease. In particular, we investigate whether countries’ funding requests were in line with funding needs as defined by income and disease burden or whether there was evidence of financial suppression of large grants (i.e. lower ratings of proposals requesting higher total volumes of funding after controlling for other explanatory variables). Second, we identify factors that explain TRP ratings of grant proposals and whether these ratings were consistent with the Global Fund’s ambition to give priority to countries most in need. Third, we ask whether TRP recommendations and other proposal characteristics were a good predictor of successful implementation of Global Fund grants, and whether there was evidence for changes in the quality of proposals over time. The success of scaling-up health investments despite initial concerns about absorptive capacity (Lu et al. 2006; Sachs and Schmidt-Traub 2017) stands in contrast to other investment priorities. Health systems, non-communicable diseases, environmental health and non-health-related areas have not experienced a similar scaling up of aid (de Jongh et al. 2014; IHME 2016). They have also seen a lower acceleration of progress under the Millennium Development Goals (McArthur and Rasmussen 2017). Achieving the recently adopted Sustainable Development Goals (SDGs) will require large increases in domestic and external funding for health systems, non-communicable diseases, environmental health (Jamison et al. 2013), education and other areas (Schmidt-Traub 2015). With Gavi and the Global Fund routinely rated among the best performing financing mechanisms (DFID 2016) we also consider whether the performance of the TRP may hold lessons for scaling up funding in other areas. Materials and methods Consistent with earlier investigations into the performance of Global Fund grants (Lu et al. 2006; Radelet and Siddiqi 2007; Katz et al. 2010; Fan et al. 2013; Bowser et al. 2014), we constructed statistical models testing dependent variables against two groups of regressors (i) proposal or grant characteristics relating to both the rounds-based mechanism and the RCC, and (ii) country characteristics covering criteria for need and the effectiveness with which aid might be used by a country. Regressors relating to proposals and grants include the volume of funding requested; the year of the round to track changes over time; TRP recommendations and Board decisions; and indicators of grant performance. Two dummy variables identify learning effects (van Kerkhoff and Szlezák 2006, 2016) and other changes made to proposals in repeat submissions. The first tracks whether proposals and the resulting grants were a resubmission of an earlier proposal rated 3 by the TRP. A second dummy identifies countries that have already had one grant approved by the Global Fund. To combine information on funding proposals and approved grants, proposals rated 1 or 2 by the TRP were manually matched with signed grants from the grants dataset using proposal parameters provided in both datasets, including the proposal name, disease component, country, funding volumes and the year of submission. Consistent with earlier studies into Global Fund grants, we grouped proposals and grants into three disease categories: HIV/AIDS including combined proposals for HIV/AIDS and TB, TB, and malaria (Radelet and Siddiqi 2007; Fan et al. 2013). During the rounds-based mechanism, funding requests for health systems strengthening could be either embedded in disease-specific proposals or be submitted separately under the category ‘other’. Since only 54 proposals were submitted under this separate category it was not possible to study proposals and grants relating specifically to health systems using our statistical models. Regressors relating to country characteristics include measures of need employed by the Global Fund (GFATM 2001, 2013a) and the literature on aid (Temple 2010; Ottersen et al. 2017), including GDP per capita and measures of disease burden. We also consider domestic health expenditure and official development assistance received from other donors in line with Lu et al. (2006) and Fan et al. (2013). Coverage of disease interventions, such as long-lasting insecticide-treated malaria bednets (LLINs), artemisinin-based combination therapy (ACT) or anti-retroviral therapy (ART), can be positively or negatively associated with funding needs depending on the intervention. However, data on intervention coverage were too incomplete for inclusion, except for the TB treatment success rate. Finally, the aid allocation literature shows that countries with large populations receive less aid ceteris paribus (Younas 2008; Bourguignon et al. 2009), so we test for financial suppression by including population size among the regressors. Factors associated with the effectiveness with which incremental health financing can be used include health systems, governance, and other country characteristics. The Universal Health Coverage (UHC) Tracer (GBD 2016) tracks the robustness of countries’ health systems, which are an important factor in scaling up complex disease control and treatment programs (Samb et al. 2009; Bowser et al. 2014; iERG 2014). Building on the work of World Bank (1998) and Collier and Dollar (2002), the health and development economics literature has considered the role of governance on the effectiveness with which aid can be used (Temple 2010; Ottersen et al. 2017). Many investigations into the performance of Global Fund grants therefore consider measures of governance (Lu et al. 2006; Radelet and Siddiqi 2007; Katz et al. 2010). The six most commonly used governance indicators published by the World Bank (2016) exhibit high collinearity for the sample of countries considered in this study. We therefore use three variables (control of corruption, government effectiveness, voice and accountability) that are in line with the literature (Katz et al. 2010), the success factors identified by the Global Fund (GFATM 2001, 2013 b), and have been reviewed in external assessments (Macro International 2009; HLIRP 2011). Additional country characteristics include dummy variables for fragile states, which constitute difficult ‘operating environments’ in the language of the Global Fund, and English-speaking countries since it has been reported that non-English speaking countries faced greater difficulties in accessing funding (Kerouedan 2010; French Ministry of Foreign Affairs 2013). We include a PEPFAR dummy variable for the countries in Africa that were eligible for PEPFAR funding. This allows us to test for complementarity with Global Fund funding for HIV/AIDS, as discussed in the literature (Shakow 2006; Oomman et al. 2007). During the rounds-based mechanism, aid allocations by the Global Fund were determined by a combination of the volume of funding requested by each country, TRP ratings and subsequent Board decisions, and the frequency with which countries applied for funding. To understand how the country characteristics related to needs and efficiency contribute to each step in the allocation process, we ensure maximum coherence in the variables included in Regressions 1–3. The Supplementary data describes the Global Fund data and the construction of the dependent variables. It also provides summary statistics and data sources for all variables considered in this study. The accompanying Stata dataset and do-file allow for replication of the analysis. Regression 1(Table 1) tests the first question, by conducting a linear ordinary least squares (OLS) regression for each of the three disease categories. The model was specified as   yi=Xiβi+εi, (1) where the dependent variable yi is the natural log of the sum of total per capita funding requested by disease i per Global Fund round and any RCC waves held since the previous funding round. Xi denotes the matrix of regressors xij including proposal and country characteristics relating to need and effectiveness, as described above and summarized in Table 1, βi is the matrix of regression coefficients βij estimated in this OLS model, and εi is the disturbance term for each disease. Each regressor comprises data from the year preceding the respective Global Fund round since the TRP considered proposals on the basis of this information. To identify possible non-linearities in submissions by round we considered a linear model specification as well as factor variables for each round. Table 1. Predictors of requested funding volumes (OLS) per round, by disease (Regression 1)   (1)  (2)  (3)    HIV  TB  Malaria  ln(GDP pc), 2014 cUS$  −0.0328  −0.0119  0.0929    (0.0825)  (0.0859)  (0.0869)  ln(population)  −0.671***  −0.544***  −0.664***    (0.0293)  (0.0330)  (0.0361)  UHC tracer (0–100)  −1.089**  1.106*  −2.384***    (0.396)  (0.449)  (0.478)  Dom. health exp, %GDP  1.779  −0.196  −1.157    (2.020)  (2.125)  (2.458)  DAH Global Fund, %GDP  41.75***  3.087  7.450    (12.22)  (13.07)  (13.60)  Non-Global Fund DAH, %GDP  1.134  5.077  18.35***    (5.012)  (5.141)  (5.243)  Prevalence, %  0.0960***  1.207***  0.0224***    (0.0160)  (0.274)  (0.00334)  TB treatment success, %    −0.00691        (0.00447)    Year of Round (2002–2010)  0.154***  0.226***  0.269***    (0.0235)  (0.0266)  (0.0311)  Resubmitted proposal (0/1)  −0.0250  0.111  0.0567    (0.0798)  (0.0860)  (0.0911)  Any grant already approved (0/1)  −0.0715  −0.142  0.123    (0.0852)  (0.0892)  (0.0964)  RCC held since last Round (0/1)  0.704**  0.299  0.709*    (0.255)  (0.249)  (0.290)  Government effectiveness (−2.5/2.5)  −0.431** (0.157)  −0.526** (0.171)  −0.232 (0.203)  Control of corruption (−2.5/2.5)  0.168 (0.143)  0.226 (0.147)  0.208 (0.178)  Voice and accountability (−2.5/2.5)  0.164* (0.0780)  0.126 (0.0797)  0.230* (0.0900)  PEPFAR focus country (0/1)  0.552***        (0.132)      Fragile state (0/1)  −0.156  −0.0283  0.0676    (0.114)  (0.117)  (0.130)  English speaking (0/1)  0.0101  0.102  0.287**    (0.105)  (0.102)  (0.109)  Constant  −297.9***  −445.5***  −529.6***    (46.96)  (53.14)  (62.12)  R2  0.761  0.723  0.801  Adjusted R2  0.752  0.708  0.790  Observations  465  334  299    (1)  (2)  (3)    HIV  TB  Malaria  ln(GDP pc), 2014 cUS$  −0.0328  −0.0119  0.0929    (0.0825)  (0.0859)  (0.0869)  ln(population)  −0.671***  −0.544***  −0.664***    (0.0293)  (0.0330)  (0.0361)  UHC tracer (0–100)  −1.089**  1.106*  −2.384***    (0.396)  (0.449)  (0.478)  Dom. health exp, %GDP  1.779  −0.196  −1.157    (2.020)  (2.125)  (2.458)  DAH Global Fund, %GDP  41.75***  3.087  7.450    (12.22)  (13.07)  (13.60)  Non-Global Fund DAH, %GDP  1.134  5.077  18.35***    (5.012)  (5.141)  (5.243)  Prevalence, %  0.0960***  1.207***  0.0224***    (0.0160)  (0.274)  (0.00334)  TB treatment success, %    −0.00691        (0.00447)    Year of Round (2002–2010)  0.154***  0.226***  0.269***    (0.0235)  (0.0266)  (0.0311)  Resubmitted proposal (0/1)  −0.0250  0.111  0.0567    (0.0798)  (0.0860)  (0.0911)  Any grant already approved (0/1)  −0.0715  −0.142  0.123    (0.0852)  (0.0892)  (0.0964)  RCC held since last Round (0/1)  0.704**  0.299  0.709*    (0.255)  (0.249)  (0.290)  Government effectiveness (−2.5/2.5)  −0.431** (0.157)  −0.526** (0.171)  −0.232 (0.203)  Control of corruption (−2.5/2.5)  0.168 (0.143)  0.226 (0.147)  0.208 (0.178)  Voice and accountability (−2.5/2.5)  0.164* (0.0780)  0.126 (0.0797)  0.230* (0.0900)  PEPFAR focus country (0/1)  0.552***        (0.132)      Fragile state (0/1)  −0.156  −0.0283  0.0676    (0.114)  (0.117)  (0.130)  English speaking (0/1)  0.0101  0.102  0.287**    (0.105)  (0.102)  (0.109)  Constant  −297.9***  −445.5***  −529.6***    (46.96)  (53.14)  (62.12)  R2  0.761  0.723  0.801  Adjusted R2  0.752  0.708  0.790  Observations  465  334  299  Dependent variable (OLS): ln(Funding request pc pa) by disease, 2014 constant US$. Data are OLS regression coefficients (SE). Sample is restricted to grant proposals from individual countries that were considered by the TRP during the rounds-based mechanism (2002–2010) and for which requested funding volumes are recorded. Sources and definitions of the variables are available in the Supplementary data. GDP, gross domestic product; pc, per capita; 2014 cUS$, constant 2014 US$; DAH, Development Assistance for Health; PEPFAR, President’s Emergency Plan for AIDS Relief; RCC, Rolling Continuation Channel. * P<0.05, ** P<0.01, *** P<0.001. Regression 2(Table 2): To address the second question and to investigate predictors of the TRP recommendations, we constructed an ordered logistic regression model regressing TRP ratings (1–4) on the same proposal and country characteristics considered in Regression 1. Note that lower TRP ratings denoted higher-quality proposals. For each disease i, the probability P that TRP rating yi took on the value l (ranging from 1 to 4) is described by   Pyi=l=Fαl-Xiβi-Fαl-1-Xiβi, (2) where F is the logistic cumulative density function, αl the threshold for the observed value l, βi the matrix of regression coefficients βij estimated (including intercept terms) and Xi denotes the matrix of regressors xij described in Table 2. Since proposal review modalities for the RCC differed from the rounds-based TRP review (Rivers 2008), we excluded RCC funding requests from this regression model. As in Regression 1, we considered factor variables for each round as well as a linear model using the year of the round as a regressor. Table 2. Predictors for TRP proposal ratings (ordered logistic model), by disease (Regression 2)   (1)  (2)  (3)    HIV  TB  Malaria  TRP rating (1–4)        ln(GDP pc), 2014 cUS$  −0.709*  −0.00656  0.591    (0.286)  (0.314)  (0.384)  ln(population)  0.0663  −0.0511  0.0766    (0.139)  (0.162)  (0.232)  UHC tracer (0–100)  0.358  −1.985  1.739    (1.372)  (1.701)  (2.358)  Dom. health exp, %GDP  6.118  3.820  −2.430    (7.162)  (8.520)  (11.00)  DAH Global Fund, %GDP  −7.291  97.02*  63.00    (45.65)  (46.12)  (59.66)  Non-Global Fund DAH, %GDP  −15.41  −16.98  19.60    (17.18)  (20.19)  (23.91)  Prevalence, %  0.0496  −1.388  0.0313    (0.0596)  (1.071)  (0.0166)  TB treatment success, %    0.0181        (0.0164)    ln(board approved total pc), 2014 cUS$  −0.239 (0.153)  −0.252 (0.204)  −0.468 (0.256)  Year of Round (2002–2010)  0.276**  0.250*  0.217    (0.0884)  (0.116)  (0.168)  Resubmitted proposal (0/1)  −0.600*  −0.477  −1.388**    (0.281)  (0.335)  (0.471)  Any grant already approved (0/1)  −7.366***  −8.004***  −8.804***    (0.776)  (1.087)  (1.197)  RCC held since last Round (0/1)  −1.008  −0.678  −1.656    (0.894)  (1.019)  (1.216)  Government effectiveness (−2.5/2.5)  −0.382 (0.537)  0.155 (0.694)  −0.294 (0.851)  Control of corruption (−2.5/2.5)  0.0364 (0.487)  0.897 (0.569)  0.0871 (0.750)  Voice and accountability (−2.5/2.5)  0.376 (0.269)  −0.109 (0.326)  0.827* (0.391)  PEPFAR focus country (0/1)  0.638        (0.481)      Fragile state (0/1)  −0.186  0.146  0.876    (0.396)  (0.432)  (0.575)  English speaking (0/1)  0.334  0.0160  −1.260*    (0.371)  (0.383)  (0.504)  Pseudo R2  0.519  0.537  0.639  Observations  464  334  299    (1)  (2)  (3)    HIV  TB  Malaria  TRP rating (1–4)        ln(GDP pc), 2014 cUS$  −0.709*  −0.00656  0.591    (0.286)  (0.314)  (0.384)  ln(population)  0.0663  −0.0511  0.0766    (0.139)  (0.162)  (0.232)  UHC tracer (0–100)  0.358  −1.985  1.739    (1.372)  (1.701)  (2.358)  Dom. health exp, %GDP  6.118  3.820  −2.430    (7.162)  (8.520)  (11.00)  DAH Global Fund, %GDP  −7.291  97.02*  63.00    (45.65)  (46.12)  (59.66)  Non-Global Fund DAH, %GDP  −15.41  −16.98  19.60    (17.18)  (20.19)  (23.91)  Prevalence, %  0.0496  −1.388  0.0313    (0.0596)  (1.071)  (0.0166)  TB treatment success, %    0.0181        (0.0164)    ln(board approved total pc), 2014 cUS$  −0.239 (0.153)  −0.252 (0.204)  −0.468 (0.256)  Year of Round (2002–2010)  0.276**  0.250*  0.217    (0.0884)  (0.116)  (0.168)  Resubmitted proposal (0/1)  −0.600*  −0.477  −1.388**    (0.281)  (0.335)  (0.471)  Any grant already approved (0/1)  −7.366***  −8.004***  −8.804***    (0.776)  (1.087)  (1.197)  RCC held since last Round (0/1)  −1.008  −0.678  −1.656    (0.894)  (1.019)  (1.216)  Government effectiveness (−2.5/2.5)  −0.382 (0.537)  0.155 (0.694)  −0.294 (0.851)  Control of corruption (−2.5/2.5)  0.0364 (0.487)  0.897 (0.569)  0.0871 (0.750)  Voice and accountability (−2.5/2.5)  0.376 (0.269)  −0.109 (0.326)  0.827* (0.391)  PEPFAR focus country (0/1)  0.638        (0.481)      Fragile state (0/1)  −0.186  0.146  0.876    (0.396)  (0.432)  (0.575)  English speaking (0/1)  0.334  0.0160  −1.260*    (0.371)  (0.383)  (0.504)  Pseudo R2  0.519  0.537  0.639  Observations  464  334  299  Dependent variable (ordered logit): TRP rating (1–4), odds ratios. Data are ordered logistic regression odds ratios (SE) without constant. Sample is restricted to grant proposals from individual countries that were considered by the TRP during the rounds-based mechanism (2002–2010). Sources and definitions of the variables are available in the Supplementary data. GDP, gross domestic product; pc, per capita; 2014 cUS$, constant 2014 US$; DAH, Development Assistance for Health; UHC, Universal Health Coverage; PEPFAR, President’s Emergency Plan for AIDS Relief; RCC, Rolling Continuation Channel. * P <0.05, ** P <0.01, *** P <0.001. Regression 3(Table 3): To combine the effects of requested funding volumes per round (dependent variable in Regression 1) with approval rates (dependent variable in Regression 2) and the frequency with which a country applied for funding, Regression 3 (Table 3) used OLS and the same functional form as Regression 1 to assess predictors of total per capita Global Fund funding received by country and by disease during the rounds-based mechanism. The dependent variable yi was defined as the natural logarithm of total per capita funding received by disease i during the rounds-based funding mechanism, including funding allocated under the RCC. The regressors xij are described in Table 3. The regressor DAH from the Global Fund was modified to exclude DAH for the disease each regression focused on. Since we were interested in the effects over the full duration of the rounds-based mechanism, the independent variables were expressed as the mean over the period 2001–2010. Table 3. Predictors for total per capita Global Fund funding during rounds-based mechanism (OLS)—by disease (Regression 3)   (1)  (2)  (3)    HIV  TB  Malaria  ln(av GDP pc), 2014 cUS$  −0.4521*  −0.3205  −0.0518    (0.1949)  (0.1815)  (0.1579)  ln(av population)  −0.5502***  −0.4719***  −0.5633***    (0.0726)  (0.0646)  (0.0652)  av UHC tracer (0–100)  0.5223  2.3904*  −3.2203***    (0.9995)  (1.0042)  (0.9206)  av Dom. health exp, %GDP  4.2055  −0.7168  −2.5929    (5.2168)  (4.3294)  (4.6890)  Global Fund DAH residual, %GDP  82.632 (72.511)  20.052 (32.666)  37.910 (42.666)  av non-Global Fund DAH, %GDP  −2.064 (16.417)  −8.261 (11.756)  19.398 (12.190)  av Prevalence, %  0.1768*  1.8773***  0.0440***    (0.0787)  (0.5438)  (0.0104)  av TB treatment success, %    0.0031        (0.0098)    av Control of corruption (−2.5/2.5)  −0.1758 (0.3847)  −0.2831 (0.3333)  −0.1246 (0.3844)  av Government effectiveness (−2.5/2.5)  −0.4891 (0.4394)  −0.2498 (0.3682)  0.2472 (0.4718)  av Voice and accountability (−2.5/2.5)  0.1614 (0.1871)  0.0563 (0.1682)  −0.1397 (0.1820)  av PEPFAR focus country (0–1)  1.3285**        (0.4978)      Fragile state (0–1)  −0.7395*  −0.3502  −0.2222    (0.3237)  (0.2749)  (0.2711)  English speaking (0/1)  0.1425  0.5373*  0.4968*    (0.2880)  (0.2119)  (0.2141)  Constant  12.6355***  8.3533***  11.5731***    (2.0848)  (1.8610)  (1.7591)  R2  0.653  0.632  0.824  Adjusted R2  0.607  0.575  0.788  Observations  111  97  72    (1)  (2)  (3)    HIV  TB  Malaria  ln(av GDP pc), 2014 cUS$  −0.4521*  −0.3205  −0.0518    (0.1949)  (0.1815)  (0.1579)  ln(av population)  −0.5502***  −0.4719***  −0.5633***    (0.0726)  (0.0646)  (0.0652)  av UHC tracer (0–100)  0.5223  2.3904*  −3.2203***    (0.9995)  (1.0042)  (0.9206)  av Dom. health exp, %GDP  4.2055  −0.7168  −2.5929    (5.2168)  (4.3294)  (4.6890)  Global Fund DAH residual, %GDP  82.632 (72.511)  20.052 (32.666)  37.910 (42.666)  av non-Global Fund DAH, %GDP  −2.064 (16.417)  −8.261 (11.756)  19.398 (12.190)  av Prevalence, %  0.1768*  1.8773***  0.0440***    (0.0787)  (0.5438)  (0.0104)  av TB treatment success, %    0.0031        (0.0098)    av Control of corruption (−2.5/2.5)  −0.1758 (0.3847)  −0.2831 (0.3333)  −0.1246 (0.3844)  av Government effectiveness (−2.5/2.5)  −0.4891 (0.4394)  −0.2498 (0.3682)  0.2472 (0.4718)  av Voice and accountability (−2.5/2.5)  0.1614 (0.1871)  0.0563 (0.1682)  −0.1397 (0.1820)  av PEPFAR focus country (0–1)  1.3285**        (0.4978)      Fragile state (0–1)  −0.7395*  −0.3502  −0.2222    (0.3237)  (0.2749)  (0.2711)  English speaking (0/1)  0.1425  0.5373*  0.4968*    (0.2880)  (0.2119)  (0.2141)  Constant  12.6355***  8.3533***  11.5731***    (2.0848)  (1.8610)  (1.7591)  R2  0.653  0.632  0.824  Adjusted R2  0.607  0.575  0.788  Observations  111  97  72  Dependent variable: ln(total signed funding pc) rounds-based mechanism, 2014 constant US$. Data are OLS regression coefficients (SE). Sample is restricted to grants approved under the rounds-based mechanism (2002–2010). Sources and definitions of the variables are available in the Supplementary data. GDP, gross domestic product; pc, per capita; 2014 cUS$, constant 2014 US$; DAH, Development Assistance for Health; PEPFAR, President’s Emergency Plan for AIDS Relief. * P <0.05, ** P <0.01, *** P <0.001. Regression 4(Table 4): TRP reports (TRP 2009) and earlier investigations (Lu et al. 2006; Radelet and Siddiqi 2007; Katz et al. 2010) suggest that grant performance ratings were poorly specified and subject to significant discretion by the secretariat (Fan et al. 2013), a point also made by the TRP (2009) and external reviewers (HLIRP 2011). In Regression 4 (Table 4), we did not attempt to resolve these issues and focused instead on investigating the relationship between TRP recommendations and the performance of resulting grants. We also considered whether fragile or English-speaking countries performed significantly differently, and whether there were any significant changes across the rounds. An ordered logit model with the same functional form as for Regression 2 was estimated for each disease i the probability P that the average Phase 1 performance rating yi took on the value l (ranging from 1 to 5 with lower values denoting stronger performance). Table 4 describes the regressors. In this way, we assessed whether TRP standards might have changed over time, particularly with repeat submissions of proposals. Table 4. Predictors for Phase 1 performance rating (ordered logistic model), by disease (Regression 4), odds ratios   (1)  (2)  (3)    HIV  TB  Malaria  Av Phase 1 perf rating, discrete (1–5)  TRP rating = 1  −0.4271  −0.0803  0.4777    (0.7115)  (0.5538)  (0.8269)  ln(board approved total pc), 2014 cUS$  0.2483 (0.1707)  0.0342 (0.2090)  −0.1113 (0.2238)  Year of Round (2002–2010)  0.1579  0.0762  0.2390    (0.1068)  (0.1238)  (0.1408)  Resubmitted proposal (0/1)  0.1208  −0.2810  −0.0769    (0.3022)  (0.3447)  (0.3740)  Any grant already approved (0/1)  −0.6664  0.2579  −0.4988    (0.4729)  (0.5856)  (0.7504)  RCC held since last Round (0/1)  0.0466  0.9869  0.1274    (0.9566)  (0.7507)  (0.9847)  PEPFAR focus country (0/1)  −0.4912        (0.5150)      Fragile state (0/1)  0.4514  −0.4183  0.1344    (0.4650)  (0.4907)  (0.5403)  English speaking (0/1)  −0.2702  −0.1301  −0.9377    (0.4197)  (0.4654)  (0.4814)  Additional control variables:  0.4256  −0.0470  1.1809**  ln(GDP pc), 2014 cUS$  (0.2913)  (0.3964)  (0.3719)  ln(population)  0.0139  0.0266  0.0354    (0.1597)  (0.1851)  (0.2171)  UHC tracer (0–100)  −3.5463*  −2.0079  −4.8589*    (1.4135)  (1.9322)  (2.2038)  Dom. health exp, %GDP  −2.462  −20.544*  −2.682    (7.554)  (9.420)  (10.209)  DAH Global Fund, %GDP  −23.410  −37.113  25.357    (41.904)  (73.230)  (46.697)  non-Global Fund DAH, %GDP  16.520  21.483  33.939    (17.919)  (22.402)  (20.005)  Prevalence, %  0.0989  −2.3241  0.0432*    (0.0844)  (1.2855)  (0.0178)  Phase 1 change mortality, abs  −1.87  4.05  12.01    (9.78)  (117.30)  (19.62)  TB treatment success, %    −0.0318        (0.0190)    P1 change TB treatm success, abs    −0.0662        (0.0339)    LFA: KPMG  0.8943  0.3021  1.2306    (0.7741)  (0.7790)  (1.0456)  LFA: PWC  −0.0264  1.0176  0.8987    (0.4885)  (0.5288)  (0.6400)  LFA: STPH  0.0476  0.7357  0.7545    (0.5809)  (0.6906)  (0.6966)  LFA: UNOPS  −0.4459  0.7423  0.1314    (0.6068)  (0.6339)  (0.9494)  PR: Government  0.0348  1.8559*  −0.4335    (0.7598)  (0.8959)  (0.8758)  PR: Local CSO  0.3742  2.8102*  −1.9810    (0.8202)  (1.1244)  (1.0200)  PR: International CSO  −0.0445  0.0050  −2.3343*    (0.8679)  (1.0168)  (1.0274)  PR: Multilateral  −1.0281  1.5553  −1.7861    (0.8212)  (0.9777)  (0.9672)  Government effectiveness (−2.5/2.5)  0.2201  −0.3888  −0.9319    (0.6393)  (0.7257)  (0.9029)  Control of corruption (−2.5/2.5)  −0.5576  −0.2166  0.0136    (0.5266)  (0.6331)  (0.7573)  Voice and accountability (−2.5/2.5)  0.1056  −0.3197  0.1104    (0.2802)  (0.3325)  (0.3956)  Pseudo R2  0.093  0.101  0.124  Observations  192  164  139    (1)  (2)  (3)    HIV  TB  Malaria  Av Phase 1 perf rating, discrete (1–5)  TRP rating = 1  −0.4271  −0.0803  0.4777    (0.7115)  (0.5538)  (0.8269)  ln(board approved total pc), 2014 cUS$  0.2483 (0.1707)  0.0342 (0.2090)  −0.1113 (0.2238)  Year of Round (2002–2010)  0.1579  0.0762  0.2390    (0.1068)  (0.1238)  (0.1408)  Resubmitted proposal (0/1)  0.1208  −0.2810  −0.0769    (0.3022)  (0.3447)  (0.3740)  Any grant already approved (0/1)  −0.6664  0.2579  −0.4988    (0.4729)  (0.5856)  (0.7504)  RCC held since last Round (0/1)  0.0466  0.9869  0.1274    (0.9566)  (0.7507)  (0.9847)  PEPFAR focus country (0/1)  −0.4912        (0.5150)      Fragile state (0/1)  0.4514  −0.4183  0.1344    (0.4650)  (0.4907)  (0.5403)  English speaking (0/1)  −0.2702  −0.1301  −0.9377    (0.4197)  (0.4654)  (0.4814)  Additional control variables:  0.4256  −0.0470  1.1809**  ln(GDP pc), 2014 cUS$  (0.2913)  (0.3964)  (0.3719)  ln(population)  0.0139  0.0266  0.0354    (0.1597)  (0.1851)  (0.2171)  UHC tracer (0–100)  −3.5463*  −2.0079  −4.8589*    (1.4135)  (1.9322)  (2.2038)  Dom. health exp, %GDP  −2.462  −20.544*  −2.682    (7.554)  (9.420)  (10.209)  DAH Global Fund, %GDP  −23.410  −37.113  25.357    (41.904)  (73.230)  (46.697)  non-Global Fund DAH, %GDP  16.520  21.483  33.939    (17.919)  (22.402)  (20.005)  Prevalence, %  0.0989  −2.3241  0.0432*    (0.0844)  (1.2855)  (0.0178)  Phase 1 change mortality, abs  −1.87  4.05  12.01    (9.78)  (117.30)  (19.62)  TB treatment success, %    −0.0318        (0.0190)    P1 change TB treatm success, abs    −0.0662        (0.0339)    LFA: KPMG  0.8943  0.3021  1.2306    (0.7741)  (0.7790)  (1.0456)  LFA: PWC  −0.0264  1.0176  0.8987    (0.4885)  (0.5288)  (0.6400)  LFA: STPH  0.0476  0.7357  0.7545    (0.5809)  (0.6906)  (0.6966)  LFA: UNOPS  −0.4459  0.7423  0.1314    (0.6068)  (0.6339)  (0.9494)  PR: Government  0.0348  1.8559*  −0.4335    (0.7598)  (0.8959)  (0.8758)  PR: Local CSO  0.3742  2.8102*  −1.9810    (0.8202)  (1.1244)  (1.0200)  PR: International CSO  −0.0445  0.0050  −2.3343*    (0.8679)  (1.0168)  (1.0274)  PR: Multilateral  −1.0281  1.5553  −1.7861    (0.8212)  (0.9777)  (0.9672)  Government effectiveness (−2.5/2.5)  0.2201  −0.3888  −0.9319    (0.6393)  (0.7257)  (0.9029)  Control of corruption (−2.5/2.5)  −0.5576  −0.2166  0.0136    (0.5266)  (0.6331)  (0.7573)  Voice and accountability (−2.5/2.5)  0.1056  −0.3197  0.1104    (0.2802)  (0.3325)  (0.3956)  Pseudo R2  0.093  0.101  0.124  Observations  192  164  139  Dependent variable (Regression 4, ordered logit): Phase 1 Performance rating (1–5), odds ratios. Data are ordered logistic regression odds ratios (SE) without constant. Sample is restricted to grant proposals from individual countries that were approved during the rounds-based mechanism (2002–2010). Sources and definitions of the variables are available in the Supplementary data. GDP, gross domestic product; pc, per capita; 2014 cUS$, constant 2014 US$; TRP, Technical Review Panel; DAH, Development Assistance for Health; UHC, Universal Health Coverage; LFA, Local Fund Agent; PWC, PriceWaterHouse Coopers; STPH, Swiss Tropical and Public Health Institute; UNOPS, United Nations Operations and Project Services; PR, Principal Recipient; CSO, Civil Society Organization; PEPFAR, President’s Emergency Plan for AIDS Relief; RCC, Rolling Continuation Channel. * P <0.05, ** P <0.01, *** P <0.001. Significance of regressors was established at P < 0.05, and each model was subjected to stepwise backward elimination of non-significant predictors (Chatterjee and Hadi 2015) to confirm robustness of predictors. Standard post-regression tests were conducted for data outliers, homoscedasticity, normality of residuals and multi-collinearity of predictors (Supplementary data). We underscore that the factors explaining funding requests, TRP recommendations and funding volumes are highly complex with potential interactions among variables and nonlinear effects, as might be the case for changes in proposal volumes over time. We therefore considered partial regression plot for all regressors and tested interaction terms among variables, but none were found to be significant. Results and discussion This section presents and discusses the results from the regressions summarized in Tables 1–4. Regression 1(Table 1): These regressions generated high adjusted R2 values. After controlling for other factors, higher disease prevalence was associated with higher requested funding volumes (P < 0.001), as would be expected from a needs-based allocation of funding. Population was negatively correlated (P < 0.001) with per capita funding requests, suggesting that large countries exercised financial suppression in Global Fund proposals. This finding was robust under several different specifications and is consistent with the development economics literature (Younas 2008; Bourguignon et al. 2009; Temple 2010), through other studies into DAH (Lu et al. 2010) have not controlled for population size. Coefficients for resubmitted proposals were not significant, so countries did not reduce funding requests following an initial rejection by the TRP. If countries believed the TRP exercised financial suppression they would be expected to reduce funding requests upon resubmission to increase the likelihood of a TRP recommendation. Proposals that followed the successful approval of a first grant to the country did not request significantly different funding volumes. The RCC dummy is significantly associated with higher funding requests for HIV/AIDS and malaria, suggesting that the RCC mechanism did deliver additional resources for high-performing countries. The year of the round was correlated with larger funding requests (P < 0.001) consistent with a scaling-up of program size over time. The coefficient was small, but since over this period the real cost of disease interventions, such as ART (Stover et al. 2011) or malaria interventions (Zelman et al. 2014), fell sharply, the evidence suggests that countries designed their proposals around a substantial scaling-up of interventions. Augmented partial residual plots (Supplementary data) show that ceteris paribus the volume of funding requests increased without major nonlinearities, a finding that is confirmed by alternative specifications that replace the year of round variable with dummy variables for each round (Supplementary data). GDP per capita and domestic health spending were not significantly associated with funding requests, and DAH from non-Global Fund donors was a significant predictor only for higher malaria funding requests. Global Fund DAH was significantly associated with higher funding requests for HIV/AIDS proposals. The association with the strength of health systems, as measured by the Universal Health Care (UHC) tracer, is mixed. Countries with a higher UHC score requested lower per capita funding for HIV/AIDS and malaria (P < 0.01), a finding that is consistent with high unmet financing needs for health system strengthening, as reported widely in the literature (Carrin et al. 2010; Bowser et al. 2014; iERG 2014). Yet, in the case of TB, the association has the opposite sign (P < 0.05). These findings are consistent with Regression 3 below, and they are robust to stepwise elimination (Supplementary data), including the elimination of domestic health spending, which is highly correlated with the UHC tracer. Dropping the UHC tracer from the regressions does not alter the significance of the other coefficients, and replacing the UHC tracer with the density of physicians, a widely used measure for health services (Radelet and Siddiqi 2007; Katz et al. 2010), generates similar results. This evidence reduces the likelihood of a spurious association. A speculative interpretation of the positive association for TB might be that stronger health systems allow for the design of larger-scale programs and that in the case of TB this effect dominates the need for incremental investments to strengthen weak health systems. This issue requires closer scrutiny, possibly by using measures for the specific components of health systems required for operating control and treatment programs for each disease. Countries with lower government effectiveness requested more funding for HIV/AIDS and TB, but this did not translate into higher approved funding volumes (Regression 3). Countries with a higher score on the governance variable ‘voice and accountability’ requested more funding, but this effect was only significant for HIV/AIDS and TB and increased after removing less significant predictors from the model (Supplementary data). Control of corruption was not a significant predictor. PEPFAR focus countries requested significantly more funding per capita (P < 0.001). PEPFAR support to countries would have two competing impacts on funding requests to the Global Fund. On the one hand, PEPFAR reduces residual financing needs for HIV/AIDS programs, but on the other it increases the system capacity for HIV/AIDS program design and execution, which enables PEPFAR countries to scale up programs rapidly (Shakow 2006). On balance, the impact of greater technical capacity appears to have outweighed lower residual funding needs compared with non-PEPFAR countries, a conclusion that is supported by country case studies in Mozambique, Tanzania, and Uganda (Oomman et al. 2007). This finding is also consistent with repeated concerns flagged by the TRP about the low quality of technical assistance to HIV/AIDS grants (TRP 2009), which would help explain the weaker performance of non-PEPFAR countries. On balance, a limiting factor on scaling up appears to be greater funding for technical assistance. Even more significantly, the findings suggest that countries operated below their maximum absorptive capacity for HIV/AIDS programs during the rounds-based mechanism, since additional PEPFAR funding generated a greater demand for Global Fund resources, as discussed further under Regression 2. Contrary to findings elsewhere in the development economics literature (Collier and Dollar 2002; Temple 2010), fragile countries were not associated with lower funding requests to the Global Fund. The dummy for English-speaking countries was not significant, except for a positive correlation of the English-speaking country dummy with malaria funding requests. Regression 2(Table 2): Few predictors were significant in explaining TRP ratings. Funding proposals from countries with large populations were not associated significantly with poorer ratings, suggesting that the TRP did not exercise financial suppression contrary to the prevailing practice of most donors (Bourguignon et al. 2009). This finding is robust to different specifications of Regression 2. It contrasts with the highly significant association of population with the dependent variables in Regressions 1 and 3. After a first grant had been approved by the Global Fund, subsequent proposals from the same country became associated with better TRP ratings (p < 0.001). In addition, resubmitted proposals received better TRP ratings, but the association was not significant for TB proposals. This evidence is consistent with either a lowering of TRP standards for subsequent proposals, a learning effect leading to higher-quality proposals, or a spurious association since high-capacity proponents are more likely to prepare better proposals, which are more likely to be approved over time. Such a spurious association appears less likely because upon removing the variable ‘Any grant already approved’, the significance of ‘Resubmitted proposal’ rises for all three diseases (p < 0.001) without any other major changes to the model results. Since average TRP ratings were positively associated with the year of the round (significant for malaria proposals after stepwise elimination), countries appear to have improved the quality of repeat submissions. Evidence from Regression 4 shows that dummies for resubmitted proposals or proposals following the approval of the first grant for the country were not significantly associated with grant performance ratings, which supports the interpretation that the TRP did not lower its standards for such proposals, and instead we see evidence of learning by countries. The interpretation that the TRP promoted substantial learning in disease program design and implementation is supported by the literature on the Global Fund (Stover et al. 2011; Jamison et al. 2013; Bridge et al. 2016; van Kerkhoff and Szlezák 2016). For example, following two subsequent rejections of its HIV/AIDS proposals by the TRP, China reformed its approach to managing the disease by including international best practice on harm reduction. TRP requirements to adhere to medical best practice by involving communities and people living with the diseases in program design and implementation, also had a deep impact on China’s malaria control program (Wang et al. 2014; Minghui et al. 2015). More broadly, evidence-based funding decisions mediated by the TRP have strengthened harm reduction across the world (Atun and Kazatchkine 2010; Bridge et al. 2016). Similarly, the TRP helped identify gaps in available health interventions, increased adoption, strengthened program design, and drove down the cost of solutions, such as rapid-diagnostic tests for malaria (Zhao et al. 2012), LLINs (WHO 2007; Noor et al. 2009; Zelman et al. 2014), the shift to ACTs (Cohen et al. 2008; Roll Back Malaria Partnership 2008), and ART (Stover et al. 2011). Countries that had been invited to participate in the RCC mechanism, did receive better proposal ratings, but the association is not significant at 5%, even after stepwise elimination. Per capita GDP was not a significant predictor of TRP ratings except for HIV/AIDS proposals where income correlated with better TRP ratings (p < 0.05). Likewise, TRP ratings were not significantly associated with governance variables except for malaria grants where the coefficient for ‘voice and accountability’ was positive (p < 0.05). In line with Bornemisza et al. (2010), there is no evidence that the TRP discriminated against fragile countries or that these countries had greater difficulties in submitting high-quality proposals. English-speaking countries were more likely to receive better TRP ratings for malaria grants (p < 0.05). The relationship was not significant for other diseases. After stepwise elimination of insignificant predictors, PEPFAR countries were significantly associated with better proposal ratings, reinforcing the interpretation (Regression 1) that they had greater capacity to propose high-quality proposals and that residual funding needs were large. Domestic health spending and DAH were not significantly associated with TRP ratings, except for Global Fund DAH for TB, which correlated with worse TRP ratings. Regression 3 shows that this did not translate into a significant effect on overall grant volumes for TB. The UHC tracer, TB treatment success, and prevalence rates were not significant predictors of TRP ratings. On balance, TRP recommendations were not correlated with funding needs measured by disease prevalence, and poorer countries were significantly more likely to have their HIV/AIDS proposals accepted. These findings are robust to different specifications of the regressors reported in the literature (Radelet and Siddiqi 2007; Katz et al. 2010; Lu et al. 2010; Fan et al. 2013), suggesting that TRP decisions were made primarily on the basis of factors independent of country characteristics (except disease prevalence) and funding volumes. This finding supports the hypothesis that TRP recommendations were based on the intrinsic technical quality of the proposals, consistent with the TRP’s mandate during the rounds-based mechanism. Regression 3 (Table 3): Cumulative funding volumes were positively associated with disease prevalence for each disease (P < 0.05), suggesting that the TRP mechanism solicited and recommended high-quality funding requests from higher-burden countries. The association with per capita GDP was significantly negative for HIV/AIDS, and the negative coefficient became significant for TB but not for malaria after stepwise elimination. After controlling for other factors, the Global Fund directed higher per capita funding towards poorer countries. Population size was negatively associated with overall funding volumes (P < 0.001). This finding is robust across different specifications of the model. It suggests that the self-suppression of funding requests from larger countries (Regression 1) resulted in lower funding volumes even though we could identify no evidence of financial suppression in TRP recommendations (Regression 2). These findings challenge the conclusion of the 2011 review of Global Fund operations that ‘large countries put forward enormous proposals’ (HLIRP 2011). As in Regression 1, the UHC tracer generated significant associations with opposing signs for TB and malaria. A stronger health system was associated with higher TB funding, consistent with the hypothesis that more effective health systems support a greater scaling-up of TB interventions. Meanwhile, the association was negative for malaria grants (P < 0.001), possibly due to the dominant role played by the Global Fund in financing malaria LLINs, which could be distributed outside health systems (Hafner and Shiffman 2013; de Jongh et al. 2014) and were most needed in poorer countries that tended to have weaker health systems. These findings reiterate the need for additional statistical analyses of the role of health systems in enabling the scaling up of disease programs. Global Fund funding did not substitute for domestic health expenditure. The associations with DAH and governance variables were also non-significant. Funding for HIV/AIDS was complementary to PEPFAR funding since PEPFAR countries were positively associated with higher per capita funding volumes, reflecting both larger (Regression 1) and higher-quality (Regression 2) proposals. The fragile states dummy was not significantly associated with per capita funding volumes, except for a negative association with HIV/AIDS grants, which is robust to removing the PEPFAR dummy. Considering evidence from the first two regressions, fragile countries submitted fewer proposals, but their proposals did not receive worse TRP ratings. This suggests greater scope for technical assistance to fragile countries to accelerate proposal design. Once again the differential performance of fragile countries on HIV/AIDS grants is consistent with the TRP’s concerns about low-quality technical assistance for HIV/AIDS (TRP 2009). English-speaking countries received more funding (P < 0.05) for TB and malaria grants, but the correlation was not significant for HIV/AIDS. This lends support to concerns (Kerouedan 2010; French Ministry of Foreign Affairs 2013) that non-English-speaking countries were less successful at attracting Global Fund funding. For malaria grants Regressions 1 and 2 suggest that English-speaking countries submitted larger proposals that were more likely to be approved by the TRP. In the case of TB, English-speaking countries submitted proposals more frequently. Regression 4(Table 4): In spite of a large number of regressors, this ordered logistic model yielded few significant associations, possibly due to poorly specified grant proposal ratings during the rounds-based mechanism, as suggested by Fan et al. (2013). As a result, only weak inferences can be drawn from the findings, and we limit ourselves to the role of proposal and grant characteristics considered in Regressions 1–3. After controlling for other factors, average Phase 1 performance ratings did not correlate significantly with TRP recommendations. A plausible explanation is that all proposals rated 2 were revised with guidance from the TRP and the Global Fund secretariat, which increased their quality to a point where they matched or exceeded that of proposals rated 1. Grants resulting from resubmitted proposals or following the approval of an earlier grant did not perform significantly differently. This suggests that higher TRP ratings for such proposals (Regression 2) did not reflect lower TRP assessment standards but signified higher-quality proposals. We found no evidence that grants to fragile countries performed significantly differently from grants to non-fragile countries. Using a smaller dataset, an earlier study (Bornemisza et al. 2010) found a small, negative correlation between grant performance and fragility. Regressions 1–3 suggest that, with the exception of HIV/AIDS, fragile countries were not associated with lower funding requests and funding volumes. We therefore conclude that the Global Fund managed to generate quality demand in the difficult operating environments that are commonly associated with low aid allocation and low grant performance (Collier and Dollar 2002; Collier 2008; Temple 2010). PEPFAR focus countries and English-speaking countries were not associated with significant differences in performance ratings. The latter suggests that the Global Fund overcame language barriers during grant implementation even though English-speaking countries obtained higher funding volumes for some diseases. Limitations The regressions reported in this study investigate associations that are highly complex and depend on a large number of factors with possible interaction terms. Even though the R2 are relatively high, the reported associations could also be due to factors not considered in the models. In particular, we did not have access to data on the intrinsic quality of grant proposals, which could be gathered through structured reviews of random samples of grant proposals. Also, Global Fund practices (e.g. TRP procedures, evaluation standards, the role of the Technical Partners, modalities for Principal Recipients and Local Fund Agents) are widely reported to have improved over time (HLIRP 2011; French Ministry of Foreign Affairs 2013; DFID 2016), and these effects may not have been picked up fully in the regression analyses. Though we did not find any significant relationships for the most common interaction terms reported in the development economics literature (Temple 2010), it is possible that missing interaction terms affect the results. As discussed in the findings, the data quality of grant performance ratings has been questioned, so Regression 4 could only establish the plausibility of consistent TRP assessment standards during the rounds-based mechanism. Historic data on intervention coverage during the rounds-based mechanism was too sparse to construct comprehensive models of Global Fund grant performance. Too few proposals on health systems strengthening were considered during the rounds-based mechanism, so the Global Fund’s impact on health systems could not be considered in this study. Conclusion During the rounds-based mechanism (2002–2010) per capita funding requests to the Global Fund were correlated with disease prevalence rates and they increased over time in the presence of sharp falls in the cost of interventions, consistent with a significant scaling-up of programs in fragile and non-fragile countries alike. The data suggest that the Global Fund promoted substantial learning and improvements in the quality of country proposals, which is supported by findings in other parts of the public health literature. It is likely that the TRP’s transparent rating of proposals, the release of findings from each funding round, and the systematic review of lessons learnt with Technical Partners (e.g. WHO, UNAIDS, Stop TB and Roll-Back Malaria) contributed to propagating knowledge on scaling up interventions across countries. The country-led funding model appears to have encouraged scaling-up since countries were not constrained in the volume of per capita funding they could request, and success in one country inspired others to follow. The data do not support the hypothesis that large countries put forward very large proposals to the Global Fund, which served as a key justification for the shift from rounds-based to the New Funding Model. Instead, per capita funding to and funding requests from countries with larger populations were financially suppressed, even though TRP recommendations were unaffected by population size. This suggests significant unmet funding needs, particularly in countries with larger populations. Large funding requests for HIV/AIDS from PEPFAR-eligible countries lend further support to the presence of large unmet financing gaps during the rounds-based mechanism. Throughout the rounds-based mechanism, the TRP fulfilled its role. It appears to have recommended proposals without regard to funding volumes, population size, or other country characteristics typically associated with different aid volumes, such as governance, domestic health expenditure, and DAH. Global Fund funding went towards higher-burden countries with lower incomes, consistent with the guiding principles of the Global Fund. The findings suggest that the Global Fund, working with its Technical Partners, has been effective at overcoming lower capacity to design and implement programs in poorer and/or fragile countries. Fragile countries requested funding volumes that were not significantly different from those requested by other countries; they were as likely to have their funding requests approved by the TRP; and their grants performed equally well except for HIV/AIDS grants. The TRP flagged concerns about the low quality of technical assistance for HIV/AIDS grants, which might explain the differential performance of fragile countries here. Conversely, higher funding requests and better ratings of proposals from PEPFAR countries suggest that countries that receive greater support in strengthening and scaling-up their response to HIV/AIDS can attract more funding from the Global Fund. This in turn indicates the presence of large unmet funding needs in non-PEPFAR countries, which technical assistance might be able to convert into high-quality proposals. The Global Fund’s success in operating across the full spectrum of country environments, including fragile states, may set an example for financing health systems and other investment priorities under the SDGs. English-speaking countries obtained higher volumes of funding for malaria and TB. TRP reports frequently referred to quality problems with the translation of proposals and supporting documents into English. The Global Fund and other mechanisms must ensure that language does not become a barrier to accessing financing. Taken together, the evidence shows that a demand-based funding mechanism relying on independent technical review of proposals without ex-ante country allocations can generate needs-based funding allocations. It can stimulate quality demand even in poor and fragile countries contrary to widespread expectations when the Global Fund was established in 2002. Pooled international funding mechanisms in other areas, such as the Global Environment Facility, the Green Climate Fund, the Global Partnership for Education, or the International Fund for Agricultural Development, should study the workings and performance of the TRP and may consider establishing similar procedures. Further work is needed to investigate how lessons from the TRP can be applied to other sectors and to understand how the role of the TRP has evolved under the New Funding Model. Supplementary data Supplementary data are available at Health Policy and Planning online. Acknowledgements The Swedish International Development Cooperation Agency (Sida) provided funding for this study. Jeffrey Sachs, Ekko van Ierland, and Jeroen Klomp advised on the design of the study. Christoph Benn, Lucie Blok, Sofia Cordero, David Durand-Delacre, George Gotsadze, Ilze Kalnina, and Katerina Teksoz have provided valuable comments. Three anonymous referees provided valuable comments. David Durand-Delacre and Murad Hirji assisted with data collection and preparation. The views expressed are those of the author and do not represent those of Sida or of the Sustainable Development Solutions Network (SDSN). Funding The research was funded by the Sustainable Development Solutions Network using a grant from the Swedish International Development Cooperation (Sida). Conflict of interest statement. None declared. References Atun R, Kazatchkine M. 2009. Promoting country ownership and stewardship of health programs: The Global Fund experience. Journal of Acquired Immune Deficiency Syndromes  52: S67– 8. Google Scholar CrossRef Search ADS PubMed  Atun R, Kazatchkine M. 2010. The Global Fund’s leadership on harm reduction: 2002–2009. International Journal of Drug Policy  21: 103– 6. Google Scholar CrossRef Search ADS PubMed  Bornemisza O, Bridge J, Olszak-Olszewski M, Sakvarelidze G, Lazarus JV. 2010. Health aid governance in fragile states: the Global Fund experience. Global Health Governance  4: 1– 18. Bourguignon F, Levin V, Rosenblatt D. 2009. International redistribution of income. World Development  37: 1– 10. Google Scholar CrossRef Search ADS   Bowser D, Sparkes SP, Mitchell A. 2014. Global Fund investments in human resources for health: innovation and missed opportunities for health systems strengthening. Health Policy and Planning  29: 986– 97. Google Scholar CrossRef Search ADS PubMed  Bridge J, Hunter BM, Albers E et al.   2016. The Global Fund to Fight AIDS, Tuberculosis and Malaria’s investments in harm reduction through the rounds-based funding model (2002–2014). International Journal of Drug Policy  27: 132– 7. Google Scholar CrossRef Search ADS PubMed  Car J, Paljärvi T, Car M et al.   2012. Negative health system effects of Global Fund’s investments in AIDS, tuberculosis and malaria from 2002 to 2009: systematic review. JRSM Short Reports  3: 70. Google Scholar CrossRef Search ADS PubMed  Carrin G, Buse K, Heggenhougen K, Quah SR. 2010. Health Systems Policy, Finance, and Organization . Amsterdam: Elsevier Academic Press. Chatterjee S, Hadi AS. 2015. Regression Analysis by Example . Hoboken, NJ: John Wiley & Sons. Cohen JM, Singh I, O'Brien ME. 2008. Predicting Global Fund grant disbursements for procurement of artemisinin-based combination therapies. Malaria Journal  7: 200. Google Scholar CrossRef Search ADS PubMed  Collier P. 2008. The Bottom Billion: Why the Poorest Countries Are Failing and What Can Be Done about It . New York: Oxford University Press. Collier P, Dollar D. 2002. Aid allocation and poverty reduction. European Economic Review  46: 1475– 500. Google Scholar CrossRef Search ADS   Daniels N, Sabin JE. 2008. Setting Limits Fairly: Learning to Share Resources for Health . Oxford: Oxford University Press. de Jongh TE, Harnmeijer JH, Atun R et al.   2014. Health impact of external funding for HIV, tuberculosis and malaria: systematic review. Health Policy and Planning  29: 650– 62. Google Scholar CrossRef Search ADS PubMed  DFID. 2016. Raising the Standard: The Multilateral Development Review 2016 . Department for International Development, London. Fan VY, Duran D, Silverman R, Glassman A. 2013. Performance-based financing at the Global Fund to Fight AIDS, Tuberculosis and Malaria: an analysis of grant ratings and funding, 2003–12. The Lancet Global Health  1: e161– 8. Google Scholar CrossRef Search ADS PubMed  French Ministry of Foreign Affairs. 2013. French contributions to the Global Fund to Fight AIDS, Tuberculosis and Malaria: evaluation report synthesis. No. 126. French Ministry of Foreign Affairs, Paris. GBD. 2016. Measuring the health-related Sustainable Development Goals in 188 countries: a baseline analysis from the Global Burden of Disease Study 2015. The Lancet  388: 1813– 50. CrossRef Search ADS   GFATM. 2001. The framework document. Global Fund to Fight AIDS, Tuberculosis and Malaria, Geneva. GFATM. 2008. Report of the Technical Review Panel and the Secretariat on funding recommendations for Wave 3 Rolling Continuation Channel proposals. Global Fund to Fight AIDS, Tuberculosis and Malaria, Geneva. GFATM. 2013a. The Global Fund’s New Funding Model. Global Fund to Fight AIDS, Tuberculosis and Malaria, Geneva. GFATM. 2013b. Guidelines and requirements for Country Coordinating Mechanisms. Global Fund to Fight AIDS, Tuberculosis and Malaria, Geneva. GFATM. 2016. List of successful appeals in previous rounds. Global Fund to Fight AIDS, Tuberculosis and Malaria, Geneva. Hafner T, Shiffman J. 2013. The emergence of global attention to health systems strengthening. Health Policy and Planning  28: 41– 50. Google Scholar CrossRef Search ADS PubMed  HLIRP. 2011. Turning the page from emergency to sustainability: the final report of the High-Level Independent Review Panel on Fiduciary Controls and Oversight Mechanisms of the Global Fund to Fight AIDS, Tuberculosis and Malaria. High-Level Independent Review Panel on Fiduciary Controls and Oversight Mechanisms of the Global Fund to Fight AIDS, Tuberculosis and Malaria, Geneva. iERG. 2014. Every woman, every child: a post-2015 vision. Independent Expert Review Group on Information and Accountability for Women’s and Children’s Health, Geneva. IHME. 2016. Development Assistance for Health Database 1990–2015. Jamison DT, Summers LH, Alleyne G. 2013. Global health 2035: a world converging within a generation. The Lancet  382: 1898– 955. Google Scholar CrossRef Search ADS   Katz I, Aziz MA, Olszak-Olszewski M et al.   2010. Factors influencing performance of Global Fund-supported tuberculosis grants. The International Journal of Tuberculosis and Lung Disease  14: 1097– 103. Google Scholar PubMed  Kerouedan D. 2010. Enjeux politiques de l’évaluation à cinq ans du Fonds mondial de lutte contre le sida, la tuberculose et le paludisme. Bulletin de la Société de Pathologie Exotique  103: 119– 22. Google Scholar CrossRef Search ADS PubMed  Lu C, Michaud CM, Khan K, Murray CJ. 2006. Absorptive capacity and disbursements by the Global Fund to Fight AIDS, Tuberculosis and Malaria: analysis of grant implementation. The Lancet  368: 483– 8. Google Scholar CrossRef Search ADS   Lu C, Schneider MT, Gubbins P et al.   2010. Public financing of health in developing countries: a cross-national systematic analysis. The Lancet  375: 1375– 87. Google Scholar CrossRef Search ADS   Macro International. 2009. The Five-Year Evaluation of the Global Fund to Fight AIDS, Tuberculosis, and Malaria: synthesis of study areas 1, 2 and 3. Macro International, Geneva. McArthur JW, Rasmussen K. 2017. Change of Pace: Accelerations and Advances During the Millennium Development Goal Era . Washington, DC: Brookings Institution. Minghui R, Scano F, Sozi C, Schwartländer B. 2015. The Global Fund in China: success beyond the numbers. The Lancet Global Health  3: e75– 7. Google Scholar CrossRef Search ADS PubMed  Noor AM, Mutheu JJ, Tatem AJ, Hay SI, Snow RW. 2009. Insecticide-treated net coverage in Africa: mapping progress in 2000–07. The Lancet  373: 58– 67. Google Scholar CrossRef Search ADS   Oomman N, Bernstein M, Rosenzweig S. 2007. Following the Funding for HIV/AIDS: A Comparative Analysis of the Funding Practices of PEPFAR, the Global Fund and World Bank MAP in Mozambique, Uganda and Zambia . Washington, DC: Center for Global Development. Ottersen T, Kamath A, Moon S, Martinsen L, Røttingen J-A. 2017. Development assistance for health: what criteria do multi-and bilateral funders use? Health Economics, Policy and Law  12: 223– 44. Google Scholar CrossRef Search ADS   Patel P, Cummings R, Roberts B. 2015. Exploring the influence of the Global Fund and the GAVI Alliance on health systems in conflict-affected countries. Conflict and Health  9: 7. Google Scholar CrossRef Search ADS PubMed  Radelet S, Siddiqi B. 2007. Global Fund grant programmes: an analysis of evaluation scores. The Lancet  369: 1807– 13. Google Scholar CrossRef Search ADS   Rivers B. 2008. Rolling Continuation Channel (RCC), 2nd update. Aidspan. Roll Back Malaria Partnership. 2008. The Global Malaria Action Plan. For a Malaria-Free World . Geneva: Roll Back Malaria Partnership. Sachs JD, Schmidt-Traub G. 2017. Global Fund lessons for Sustainable Development Goals. Science  356: 32– 3. Google Scholar CrossRef Search ADS PubMed  Samb B, Evans T, Dybul M et al.   2009. An assessment of interactions between global health initiatives and country health systems. The Lancet  373: 2137– 69. Google Scholar CrossRef Search ADS   Schmidt-Traub G. 2015. Investment needs to achieve the Sustainable Development Goals. SDSN Working Paper. Sustainable Development Solutions Network, Paris. Shakow A. 2006. Global Fund - World Bank HIV/AIDS Programs comparative advantage study. World Bank and Global Fund to Fight AIDS, Tuberculosis and Malaria, Washington D.C. and Geneva. Stover J, Korenromp EL, Blakley M et al.   2011. Long-term costs and health impact of continued Global Fund support for antiretroviral therapy. PLoS ONE  6: e21048. Google Scholar CrossRef Search ADS PubMed  Temple JRW. 2010. Aid and conditionality. In: Handbook of Development Economics . Amsterdam: Elsevier, 4415– 523. Google Scholar CrossRef Search ADS   TRP. 2009. Report of the Technical Review Panel and the Secretariat on Round 9 proposals. Global Fund to Fight AIDS, Tuberculosis and Malaria, Geneva. van Kerkhoff L, Szlezák NA. 2006. Linking local knowledge with global action: examining the Global Fund to Fight AIDS, Tuberculosis and Malaria through a knowledge system lens. Bulletin of the World Health Organization  84: 629– 35. Google Scholar CrossRef Search ADS PubMed  van Kerkhoff L, Szlezák NA. 2016. The role of innovative global institutions in linking knowledge and action. Proceedings of the National Academy of Sciences  113: 4603– 8. Google Scholar CrossRef Search ADS   Wang R-B, Zhang Q-F, Zheng B et al.   2014. Transition from control to elimination: impact of the 10-year global fund project on malaria control and elimination in China. Advances in Parasitology  86: 289– 318. Google Scholar CrossRef Search ADS PubMed  WHO. 2007. WHO Releases New Guidance on Insecticide-Treated Mosquito Nets . Geneva: World Health Organization. Wood A. 2008. Looking ahead optimally in allocating aid. World Development  36: 1135– 51. Google Scholar CrossRef Search ADS   World Bank. 1998. Assessing Aid: What Works, What Doesn’t, and Why . Oxford: Oxford University Press. World Bank. 2016. Worldwide Governance Indicators database. Younas J. 2008. Motivation for bilateral aid allocation: Altruism or trade benefits. European Journal of Political Economy  24: 661– 74. Google Scholar CrossRef Search ADS   Zelman B, Kiszewski A, Cotter C, Liu J. 2014. Costs of eliminating malaria and the impact of the Global Fund in 34 countries. PLoS ONE  9: e115714. Google Scholar CrossRef Search ADS PubMed  Zhao J, Lama M, Korenromp E et al.   2012. Adoption of rapid diagnostic tests for the diagnosis of malaria, a preliminary analysis of the Global Fund program data, 2005 to 2010. PLoS ONE  7: e43549. Google Scholar CrossRef Search ADS PubMed  © The Author(s) 2018. Published by Oxford University Press in association with The London School of Hygiene and Tropical Medicine. All rights reserved. For Permissions, please e-mail: journals.permissions@oup.com

Journal

Health Policy and PlanningOxford University Press

Published: Apr 1, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 12 million articles from more than
10,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Unlimited reading

Read as many articles as you need. Full articles with original layout, charts and figures. Read online, from anywhere.

Stay up to date

Keep up with your field with Personalized Recommendations and Follow Journals to get automatic updates.

Organize your research

It’s easy to organize your research with our built-in tools.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

Monthly Plan

  • Read unlimited articles
  • Personalized recommendations
  • No expiration
  • Print 20 pages per month
  • 20% off on PDF purchases
  • Organize your research
  • Get updates on your journals and topic searches

$49/month

Start Free Trial

14-day Free Trial

Best Deal — 39% off

Annual Plan

  • All the features of the Professional Plan, but for 39% off!
  • Billed annually
  • No expiration
  • For the normal price of 10 articles elsewhere, you get one full year of unlimited access to articles.

$588

$360/year

billed annually
Start Free Trial

14-day Free Trial