Why Effective Altruism Ignores Latin America

GiveWell has directed $2.3 billion in grants since 2014, of which Latin America received 0.02 cents per dollar. Yet 220 million people in the region live in or near poverty.

Effective altruism’s (EA) insistence on evidence and cost-effectiveness produces something rare in philanthropy: a methodology that actually works. GiveWell’s top charities deliver measurable reductions in child mortality at costs that embarrass almost everything else in the development sector. The difference between giving effectively and at random can be a factor of a hundred.

In 2024, GiveWell directed $397 million to organizations working in 22 countries. Nearly all were in Sub-Saharan Africa or South Asia. In 2025 GiveWell expanded to 30 countries, including the Americas, but that expansion was a response to US foreign aid cuts.

GiveWell publishes its full grant database. Of $2.3 billion in grants since 2014, two touch Latin America: a $175,000 iron absorption study in Guatemala and a share of a $250,000 multi-country malaria grant that includes Haiti alongside three African countries. The region’s total—$425,000 at most, and realistically less—amounts to roughly 0.02% of all funding. Latin America, home to 660 million people or just over 8 percent of the global population, has received less from GiveWell than a single mid-sized grant to Nigeria.

Below the line

Around 89 million people in the region live on less than $4.99 a day in today’s dollars, with one in three living in poverty. In Venezuela, Honduras, Guatemala, and Guyana, more than half the population lives in poverty. The region is the most unequal in the world by Gini coefficient, a dubious distinction it shares with Sub-Saharan Africa. GDP per capita, which is responsible for the middle-income label applied to most countries in the region, tells almost nothing about the lived experience of the bottom quartile.

Children in the region are 14 percentage points more likely to be poor than the overall population. In the OECD, that gap is merely two percentage points. 39% of the extremely poor across the region are under 15 years old.

Approximately a third of Latin Americans are classified by the World Bank as “vulnerable”, which means they are above the poverty line but only one setback away from sliding below it: a medical bill, a job loss, a flood, an earthquake, or six months of hyperinflation. These are people living in houses with corrugated iron roofs that leak in the rain, with no reliable heating or plumbing, whose children go to school but whose families have no buffer against anything going wrong.

The evidence trail

The EA approach evaluates causes on importance, neglectedness, and tractability. In principle, these criteria are geographically neutral. In practice, they are not. The randomized controlled trial infrastructure in global health was built predominantly in Sub-Saharan Africa and South Asia, because that is where development economists and international NGOs were operating in the 1990s and 2000s. The interventions GiveWell can evaluate rigorously are the interventions that were already being studied there. Latin America, with different disease burdens, different institutions, and a different economic profile, broadcasts at a frequency the existing instruments were not calibrated to receive. So it looks intractable, which is why nobody builds the evidence base, and so it remains intractable.

The region also fails a softer filter: it does not fit the mental model of where “real” poverty lives. AI safety and biosecurity are preoccupations of wealthy, technologically advanced societies—first-world concerns, imprecise as the term may be. Classic global health interventions are framed as third-world problems, addressed in low-income countries where the evidence base exists. Latin America is the Global South yet too developed for the malaria-net interventions and still too poor and unstable for existential risk to feel urgent. So it slips between the cracks. The middle-income label that covers most of the region does the rest of the work, creating a bureaucratic impression of adequacy that has little to do with reality.

Consider Argentina, a middle-income country. Argentina has experienced annual inflation exceeding 100 percent in recent years. The nation is simultaneously the IMF’s single largest debtor, holding approximately $57 billion in outstanding credit—more than a third of all IMF lending globally, and more than the combined debt of Ukraine, Egypt, Pakistan, and four other countries. The broader pattern of income volatility, institutional fragility, and rapid erosion of apparent gains is common across the region even when it is less extreme than Argentina’s case. It is precisely what a GDP-per-capita average smoothes away.

Compound interest

According to the ILO, nearly half of all workers in the region are informally employed. Without formal employment, workers have no social protection, no pension, no access to credit, and no clear path out of poverty. This traps generation after generation, because a generation with formal employment produces children with better nutrition, education, and health outcomes. The returns are slow and hard to attribute, but the impact is real.

The climate bill

The IPCC classifies Latin America as extremely vulnerable to climate change. Eight countries in the region with more than 100 million people appear in the highest-risk category.

Andean glaciers that supply freshwater to tens of millions are retreating. The southern cone of the continent sees significant reductions in rainfall, too. And up to 47% of the Amazon may reach ecological tipping points by 2050 if deforestation and climate change continue unabated. Climate change is a multiplier on every other problem the region faces, as it compounds poverty, food security, and displacement. The research that would tell us what works has barely started.

The marginal dollar

There are two obvious objections. First, Latin America’s problems are political and institutional, resistant to the interventions EA knows how to evaluate. Second, every dollar sent to Guatemala is a dollar not sent to a malaria net in Nigeria, where the cost per life saved may be an order of magnitude lower. If the marginal dollar saves more lives in Nigeria, the approach is working as intended.

Institutional volatility is a different obstacle, but various indicators place Latin America ahead of Sub-Saharan Africa in terms of effectiveness and stability. Political problems are therefore not a barrier to evaluation.

The malaria net only became tractable through sustained research investment. The evidence did not pre-exist the effort that produced it. The evidence base for vitamin A supplementation was not obvious in 1990; it was produced through sustained scientific effort. Before Miguel and Kremer’s first randomized trial in Kenya in 1998, deworming was not on EA’s radar either. The trial’s findings were later reanalysed and disputed, which is precisely the point: even a contested trial was enough to build the cause area. Tractability follows investment, and GiveWell’s criteria reward existing evidence. The effect locks in the current distribution of evidence, which reflects where researchers happened to have worked since the 1990s.

What is more, the argument assumes that the current allocation is the relevant counterfactual. A dollar in Guatemala today almost certainly saves fewer lives than a dollar in Nigeria. That comparison, however, assumes the wrong horizon. The relevant question asks whether a dollar spent building a rigorous evidence base for Latin American interventions could, over a ten-year horizon, generate returns that justify the initial investment. First, for the on-the-ground components of research, such as local staff, field offices, and survey logistics, purchasing power parity means a dollar invested in Guatemala goes further than the same dollar spent in London or San Francisco, where most EA-adjacent research infrastructure is actually built. Second, the disease burden in Latin America is under-studied. The interventions that would prove tractable are simply unknown because nobody has looked. Third, the region’s geographic and economic proximity to the United States creates political economy leverage for scaling successful interventions that is unavailable in West Africa. EA’s own longtermist reasoning accepts this logic when applied to AI safety or pandemic preparedness, where the expected value of investment depends on future returns that do not yet exist.

Across GiveWell’s entire published grant history, Latin America has received at most $425,000. Neglectedness relative to funding—the metric EA actually uses—is therefore close to absolute. A nascent effort (Theory of Change Makers) recently launched eight evidence-based nonprofits from the region, which explicitly cites the gap. The research infrastructure required to make Latin American interventions legible to EA could be built if someone decided it was worth building.

Breaking the cycle

Effective altruism did not set out to ignore Latin America. It followed the evidence trail that existed and built on the infrastructure that was already there. The logic is self-reinforcing: the less research exists, the less tractable a cause appears; and the less tractable it appears, the less research gets funded. The loop runs indefinitely unless someone breaks it deliberately.

If the neglect is driven by absent evidence rather than absent need, the right response is to fund the research infrastructure that would make Latin American interventions legible to the same rigorous approach EA applies elsewhere: randomized trials and institutional partnerships built deliberately for the region’s specific disease burden and institutional context.

The cost of producing this evidence is a fraction of the funding it would unlock. The cost of not producing it is the continued invisibility of 660 million people broadcasting on a frequency the existing instruments were never calibrated to receive. Tune the instruments, and the people stop being invisible.