The aid architecture
Rees Calder · 23 April 2026 · 7 min read
Total Official Development Assistance (ODA) from OECD DAC countries was $223.7 billion in 2023 (OECD preliminary data, April 2024). That's the largest figure ever recorded, up from $186 billion in 2021. It sounds like a lot. It is a lot. But it helps to know where it goes and why.
The aid system is not a simple pipeline of rich-country-gives-to-poor-country. It's a layered architecture with bilateral channels, multilateral institutions, NGO pass-throughs, and increasingly, private capital flowing alongside public money. Understanding the architecture matters because it determines which problems get funded, which get ignored, and where your individual donation fits.
Who pays
Five countries account for roughly 60% of all ODA: the United States ($66.0 billion, 2023), Germany ($36.7 billion), Japan ($19.6 billion), the United Kingdom ($19.4 billion), and France ($15.4 billion). The US number is the largest in absolute terms but among the smallest as a share of Gross National Income: 0.24% of GNI, well below the UN target of 0.7%.
Only five DAC countries consistently hit the 0.7% target: Luxembourg, Sweden, Norway, Denmark, and Germany (which crossed the threshold in 2016 and has stayed above since). The UK hit 0.7% from 2013 to 2020, then cut to roughly 0.5% in 2021 as a post-COVID fiscal measure. The cut removed approximately £4 billion per year from the aid budget.
The GNI percentage matters more than the absolute number for understanding commitment. A country spending 0.7% of GNI on aid is making a genuinely different fiscal choice from one spending 0.24%.
Where it goes
Three main channels.
Bilateral aid (roughly 60%). Government-to-government or government-to-NGO transfers targeted at specific countries. The US bilateral programme (administered by USAID) is the world's largest. Bilateral aid is the most politically influenced channel: donor governments choose recipients based on a mix of development need, strategic interest, and historical ties. Former colonies receive disproportionate aid from their former colonial powers, a pattern the OECD tracks explicitly.
Multilateral aid (roughly 30%). Contributions to international organisations like the World Bank's IDA (International Development Association), the Global Fund to Fight AIDS, Tuberculosis and Malaria, Gavi (the vaccine alliance), and various UN agencies. Multilateral aid is pooled, meaning individual donors have less control over allocation. The tradeoff: pooling reduces political distortion but adds institutional overhead.
Humanitarian aid (roughly 10%). Emergency response to crises: conflicts, natural disasters, displacement. Administered through the UN's OCHA coordination system, ICRC, and NGO partners. Humanitarian aid is the most visible channel (it's what you see on the news) but the smallest share of total ODA.
What it buys
The composition has shifted dramatically over the past two decades.
Health has grown fastest. Global health aid was roughly $13 billion in 2002 and reached $42 billion in 2023 (IHME Financing Global Health data). The growth was driven by three institutions: the Global Fund (founded 2002), PEPFAR (the US President's Emergency Plan for AIDS Relief, launched 2003), and Gavi (scaled from 2005). These three collectively represent the most successful targeted aid programmes in history by most outcome measures.
Education aid has stagnated. Total aid to education was $16.4 billion in 2023 (UNESCO Global Education Monitoring Report), roughly the same in real terms as a decade ago. Education consistently ranks as the most underfunded sector relative to need. The Global Partnership for Education has made the case repeatedly, but education outcomes are harder to measure than health outcomes, and measurability drives funding.
Infrastructure aid declined then partially recovered. After a long decline in the 2000s-2010s (as health dominated), infrastructure aid has partially recovered with the rise of climate-linked development finance. The World Bank's climate co-benefits framework means that infrastructure projects with climate components now attract funding that pure development projects don't.
The effectiveness debate
Does aid work? The honest answer: it depends on what you mean.
At the programme level, the evidence is strong. Specific programmes with clear theories of change, robust implementation, and measurable outcomes work. The evidence base for bed nets (malaria), antiretrovirals (HIV), vaccines (multiple diseases), deworming, and cash transfers is among the strongest in any field of social science. GiveWell's entire model is built on this programme-level evidence.
At the country level, the evidence is mixed. The relationship between total aid received and economic growth is weak and contested. William Easterly ("The White Man's Burden," 2006) and Dambisa Moyo ("Dead Aid," 2009) made influential cases that aid can be counterproductive at the macro level, creating dependency, distorting markets, and propping up bad governance. Jeffrey Sachs ("The End of Poverty," 2005) made the opposite case. The econometric literature (Clemens and Radelet, 2003; Roodman, 2007) finds small positive effects with large uncertainty bands.
The resolution: programme-level giving works. Country-level giving is ambiguous. If you're choosing where to donate, the programme-level evidence is what matters. The macro debate is important for governments allocating bilateral aid, but less relevant for individual donors choosing between specific charities.
Where individual giving fits
Total private charitable giving to international development is roughly $50-60 billion per year globally (Hudson Institute Index of Global Philanthropy, 2024). That's about a quarter of ODA. Individual donations are a meaningful share of total development finance, not a rounding error.
But individual giving is distributed very differently from ODA. Individual donors disproportionately fund: disaster relief (high-visibility, high-emotion), health (measurable, compelling), and specific countries (usually ones with media coverage or personal connections). They underfund: governance, education, institutional capacity building, and prevention.
GiveWell's recommended charities represent a tiny fraction of total individual giving (roughly $600 million of the $50-60 billion) but are selected specifically for programme-level cost-effectiveness. The gap between where most individual money goes and where the evidence says it would do the most good is the core argument of the effective giving movement.
What this means for you
Three implications.
Your donation operates in a market. It's not the only money going to the problem. Understanding the aid architecture means understanding where funding gaps actually exist vs. where they're already covered. GiveWell's "room for more funding" analyses are the best available guide to where marginal donations have the highest impact.
The 0.7% target is a useful benchmark for personal giving too. If rich countries commit to 0.7% of GNI, consider what 0.7% of your personal income would look like. For a £50,000 salary, that's £350 per year, or about £29 per month. The number is deliberately modest.
Programme-level evidence, not country-level emotion. When choosing where to give, follow the programme evidence, not the country narrative. The country that's in the news is rarely the one with the highest-impact funding gap. The programme that's proven to work is almost never the one with the most compelling fundraising video.
One sentence
Global aid is a $224 billion architecture with real successes at the programme level and genuine ambiguity at the country level. Your donation matters most when it follows the programme evidence, not the evening news.
Sources used: OECD DAC ODA Preliminary Data (April 2024), IHME Financing Global Health database (2023), UNESCO Global Education Monitoring Report (2024), Hudson Institute Index of Global Philanthropy (2024), GiveWell Room for More Funding analyses (2024), Easterly "The White Man's Burden" (Penguin, 2006), Moyo "Dead Aid" (Farrar Straus, 2009), Sachs "The End of Poverty" (Penguin, 2005), Clemens and Radelet "Absorptive Capacity" CGD Working Paper (2003), Roodman "The Anarchy of Numbers" CGD Working Paper (2007). Full links in the planning doc.