Suppose a rich country, say the UK, wants to give foreign aid to improve education in a poor country, say Malawi. How should it go about that?
There is an enormous literature comparing the cost-effectiveness of everything from buying free eyeglasses for students to paying performance bonuses to teachers. Who exactly should Her Majesty’s Treasury, in the UK’s case, or Janet Yellen, in the US case, make the check out to?
Broadly speaking, there are three options, and right now, the UK employs all of them.
The UK gives money to Malawi directly, aka “bilateral aid.”
The UK gives money to the World Bank, which makes concessional loans to Malawi.
The UK can contribute to the Global Partnership for Education (GPE), a pooled fund, which in turn gives most of its money to the World Bank to spend as grants in places like Malawi.
Everyone involved is painfully aware how redundant this looks. And again, the UK is just an example: the same challenge faces Norway, France, the United States, and others.
Meanwhile, the British Prime Minister Boris Johnson has recently called on donors to open their wallets for the GPE, which launched its new replenishment campaign late last year, asking for at least $5 billion for the period 2021-2025—more than double the amount it raised the last time around for 2018-2020.
GPE’s unique selling point is its single-minded focus on basic education, and its fairly streamlined, no-strings-attached approach to funding poor countries’ own education plans. But is GPE the best channel for education aid? Donors have lots of options, and the prima facie case against GPE can seem compelling at times.
The convoluted path from aid donors to aid recipients in the education sector
Between 2016 and 2018, about $9.45 billion per year in foreign aid was earmarked for education. These funds often follow a convoluted path from their point of origin to the ultimate recipient, sometimes changing hands multiple times.
Figure 1. The global aid architecture for education
Tracking the multiple, parallel flows of foreign aid for education
Source: Authors’ analysis of OECD’s Creditor Reporting System and GPE’s financial documents for the years 2016-18. "Bilateral agencies" include AFD, DFID, IDRC, and SIDA. "UN agencies" are UNESCO and UNICEF. "NGOs" are Oxfam, Save the Children, and CARE. "Other countries" include Australia, Belgium, Canada, Denmark, France, Germany, Japan, and others. All numbers are expressed in millions of US dollars.
We can put numbers on the three different paths from donor to recipient using data from the OECD’s Creditor Reporting System and GPE’s financial documents. We’ll focus here on sub-Saharan Africa, but you can click on the figure above to see numbers for other regions.
One hop. In the most direct channel, donors gave about $5.8 billion in direct bilateral aid for education programs to African governments between 2016 and 2018. Granted some of this bilateral aid goes through implementing agencies, in which case even the most direct route has two hops.
Two hops. Donors contribute about $23 billion on an annualized basis to the World Bank’s International Development Agency (IDA), the bank’s fund for the poorest countries, which then makes concessional loans to developing countries. About 8 percent of that money goes to education projects, of which about 28 percent is spent in Africa.
Three hops. In the least direct channel, donors also give money to pooled funds like GPE. Most of that money passes from GPE to the World Bank or UNICEF, and a large chunk of that then ends up funding education programs in sub-Saharan Africa. Some of the GPE money actually flows back to bilateral aid agencies, such as the UK’s own Foreign and Commonwealth Development Office, to administer alongside the UK’s own bilateral aid programs. In total, GPE spent just under $1 billion in Africa from 2016 to 2018.
Of course, the three channels are not the same. Indeed, it might be quite misleading to think of these as three different ways of giving money to a specific country (e.g., Malawi) for a specific purpose (e.g., to build girls’ latrines in primary schools). The one-hop, bilateral aid channel allows donors to do exactly that. But the whole point of multilateral organizations is that they have fixed rules for how money is allocated across countries, and how projects are chosen and approved. And the recipients, i.e. the Malawian government, generally get some say in how the money is used. The money isn’t always going to end up in the same place, and that’s often for the best. So if the one-hop approach is a girls’ latrine in Malawi, the two-hop approach implies spreading the money across different countries and projects, and the three-hop approach does too—with even more participatory decision making along the way.
For instance, let’s zoom in and focus just on the flows that pass through GPE and see how it does and doesn’t differ from the other channels.
Figure 2. Is GPE redundant?
A majority of GPE funds go to the World Bank, its main grant agent
Source: Data is from GPE’s consolidated financial reports, for the years 2016-18. All numbers are expressed in millions of U.S. dollars.
GPE accounts for about 7 percent of total aid for basic education, but relies on grant agents to deliver its aid. GPE's largest grant agent is the World Bank, accounting for over 75 percent of GPE spending.
Figure 3. A leaky chain
Overhead rates for the various agencies administering GPE grants
Source: Data is from GPE’s consolidated financial reports, for the years 2016-18. Agency fees are used to defray administrative costs related to the administration of grant funds and are a percentage of the grant allocation. Supervision fees are included within the grant to offset costs not covered by the grant agent’s agency fees.
For each new entity along the GPE funding chain, there are overhead costs, which help run programs but also add up. After funds reach GPE, which charges an average of 5 percent overhead each year for its administrative costs (including the operating expenses and trustee budgets), money then passes on to grant agents, who charge their own agency and supervision fees (both comprising overhead costs). For example, UN agencies are one GPE grant agent and tend to charge 7 percent of the grant amount in agency fees, while the World Bank has a lower agency fee (1.75 percent). Similarly, nongovernmental organizations (NGOs) charge 6 percent in agency fees and 18 percent in supervision fees, though there are only three GPE countries (Somalia, Bhutan, Papua New Guinea) where NGOs are the grant agents. Of course, these fees help run programs, but what we don’t know is the value for money (or expertise delivered per dollar spent) for these widely varying overhead costs. However, the real question is this: if most GPE funds end up going to the World Bank, GPE’s main grant agent, why shouldn’t these funds be directed straight from the donors to the World Bank in the first place instead of passing through GPE?
There are a few possible reasons: geographic focus, sector focus, and governance.
Geographically, GPE allocates more of its money to low-income and fragile countries compared to the World Bank
We know that a large majority of GPE funds are disbursed by the World Bank, and we also know that all the GPE recipient countries (in 2016-2018) are IDA-eligible. This isn’t surprising considering poverty is one of the key determinants of eligibility for GPE grants, along with fragility and population of out-of-school children. But does this overlap between GPE and the World Bank make GPE redundant?
Figure 4. Pro-poor bias
GPE gives a larger share of its money to low-income economies
Source: CRS data on disbursements (US dollars) for 2016-18. GPE allocation data is from their financial reports. World Bank income group classifications are from 2019. Hover over the bars to see the disbursements as expressed in millions of US dollars.
Given the big overlap in GPE and IDA countries, are both funds spending the education aid money on similar sub-groups of countries? Not quite. As shown in Figure 4, GPE spends a majority of its funds on low-income economies, including fragile and conflict-affected countries—a focus that has only increased over time since 2004. GPE uses a need-based allocation formula to determine allocation, with additional weighting for fragile and conflict-affected countries. IDA, in comparison, has a relatively smaller focus on low-income countries as a proportion of its total spending on education, and a much larger focus on “lower middle income” countries, though its total spending in dollar amounts is higher for both low-income and lower-middle income countries due to its larger overall budget envelope. Despite this differentiation in country focus, most GPE funds do end up flowing through the World Bank before going to the recipient countries.
Sectorally, GPE spends a larger proportion of its education funds on basic education compared to the World Bank and bilaterals
More than the type of countries that receive GPE’s money, a bigger difference lies in sectoral prioritization. GPE spends the vast majority (82 percent) of its budget directly on basic education. This is in contrast to IDA and UNICEF, who spend roughly one half and one fifth of their budget on basic education, respectively—though it is likely some of UNICEF’s uncategorized spend goes to teacher training and other programs to support basic education indirectly.
Figure 5. Primary focus
GPE is more focused on basic education compared to IDA and bilateral donors
Source: CRS data on disbursements (USD) for 2016-18. “Early childhood education” is formal and non-formal preschool education. “Basic education” entails primary education, basic life skills for youth, primary education equivalent for adults, and school feeding as well as secondary education and vocational training. “Post-secondary education” refers to higher education (universities, colleges, and scholarships) and advanced technical training. “Education (unspecified)” encompasses education sector policy planning, aid to education ministries, research, and teacher training where the education level is not specified. Hover over the bars to see the disbursements as expressed in millions of US dollars.
We don’t know how these funds would be spent if GPE didn’t exist. Would the money go instead to IDA or to bilateral education programs? Unclear. What we do know is that GPE does attract commitments from donors that typically don’t have a big focus on basic education in their bilateral spending, such as France and Germany.
GPE’s constituency-based governance gives rich countries relatively less control of the money compared to other funding bodies
Perhaps the biggest difference between GPE and other funding bodies lies in its governance and decision-making structure. Figure 6 compares board member voting shares across GPE, IDA, and bilaterals such as the USAID and the UK’s FCDO.
Figure 6. Shared governance
The board representation of rich and poor countries across different aid channels
Source: GPE board data as noted on their website on September 1, 2020, IDA board data as reported on August 17, 2020, and reflects voting powers of elected directors. Bilaterals refer to USAID and FCDO. UNICEF board data is for 2020. For IDA and UNICEF, “developing countries” include low-income and lower-middle-income economies only, as classified by the World Bank, while all others fall under “rich countries.”
GPE is governed by a constituency-based board of directors, with 20 members, representing developing countries, civil society organizations (CSOs), multilateral agencies, donor countries, and private sector organizations. In IDA, on the other hand, the voting share of each of the 173 members is largely determined by their financial contribution to the World Bank. The five largest donors (US, UK, China, France, and Germany) appoint five members on the 25-member board, while the remaining elected members represent all the remaining countries. The US, for example, holds 10.2 percent of the IDA voting share, which gives it considerable sway over IDA decisions.
GPE has a relatively larger share of developing country representation compared to IDA, and prioritizes consensus-building in its board decisions. If all efforts to build consensus fail, GPE uses majority voting, but the majority vote must include at least one vote in favor from each key constituency, including developing countries, donor countries, multilateral agencies, and others. Similarly at the country level, GPE requires collective oversight by national government and donors, and—while it's true that much of GPE's money is routed through the World Bank as its “grant agent”—GPE recipients generally get to choose between the World Bank or another organization to administer the program in their country, which seems like healthy competitive pressure.
Multi-stakeholder governance has its downsides. The need for consensus on a board representing widely divergent interests can lead to gridlock. GPE’s share of total basic education aid to low-income and lower-middle-income countries fell to 6.7 percent from 11.4 percent between 2014 and 2018—a fall that may potentially be explained by the abrupt decline in disbursements in 2018-2019 after a slowdown in grant approvals. This slowdown was compounded by an average three-year lag between approval and disbursements.
Fights within the big tent of the GPE board have also led to paralysis on key thematic issues. Notably, GPE avoided a policy position on private schooling for several years, and held up country programs with private-sector involvement at the objection of civil society representatives, before finally reaching agreement on a new private-sector strategy in 2019.
Redundancy versus legitimacy
Yes, GPE is a slightly expensive way to channel aid money. And yes, most GPE funds end up going to the World Bank, which raises questions about why this money shouldn’t be directed straight to the World Bank in the first place.
But GPE is different from the World Bank both in terms of who gets the money and what the money is spent on. It has a somewhat stronger focus on basic education, and a much stronger focus on low-income countries.
Ultimately though, the positive case for replenishing GPE instead of spending directly on bilateral aid programs rests primarily on GPE’s shared governance model, with a stronger voice for developing countries and civil society, and less conditionality of the sort the World Bank imposes. “Multilateralism is back” announced Joe Biden’s nominee for ambassador to the United Nations. We’ll see if that holds true with education financing as GPE tries to replenish its coffers this year.
We would like to thank Matt Clancy and Lawrie Harper-Simmonds for their comments on an earlier version of the blog.
 Data quality is an issue with OECD’s Creditor Reporting System. Disbursements from multilateral partnerships and funds, like GPE, are coded as bilateral disbursements to regional or unspecified recipients. These multilateral institutions are listed under "Reported Channel Name", which allows us to identify them (though there are often misspellings). In order to analyze disbursements by country income type, as seen in Figure 4, we had to identify, code, and remove GPE from our CRS dataset and replace these entries with yearly disbursement data from GPE's financial documents. The amounts removed from CRS roughly match the amounts added from GPE's financial documents.
Rights & Permissions
You may use and disseminate CGD’s publications under these conditions.