Across low- and middle-income countries (LMICs), millions of children attend school but learn very little. The vast majority do not even learn to read by age ten. This is not a marginal issue; it is one of the most binding constraints to global development. Without being able to read, it is nearly impossible for individuals to acquire employability skills, navigate public health guidance, hold institutions accountable, or adapt to environmental realities. There is clear evidence that resolving this learning crisis would unlock massive, long-term gains in economic productivity, health outcomes, and human flourishing.
Yet, historically, education has received far less attention and funding from the effective altruism community than global health. Drawing on the cause prioritization framework of Importance, Tractability, and Neglectedness (ITN), I argue that we should invest far more in education and specifically in helping children reach the critical threshold of literacy. But I also urge caution in how we choose the programs in which we invest.
The benefits that accrue from learning to read can be divided into two broad categories: direct economic returns and wise decision-making.
First, learning to read is associated with economic benefits for individuals and societies. An analysis by @Vadim Albinsky of Founders Pledge on this forum evaluated five independent lines of evidence, including randomized controlled trials of school access in LMICs and historical cross-sectional data, and found that a one-standard-deviation gain in learning test scores corresponds to an average 19% increase in adult earnings. He concludes that optimized educational interventions can achieve returns upwards of 11 times the impact of direct cash transfers, putting them in the same cost-effectiveness tier as top-rated global health interventions. There is evidence that learning levels (including literacy) are also associated with economic growth at a country level; a 1 standard deviation increase in a country’s score on international standardised assessments is associated with a 2 percentage point higher annual GDP per capita growth rate.
The second category of benefit is more abstract but arguably just as important: wise decision-making by individuals and institutions. As the adage goes, first you learn to read, then you read to learn; Literacy acts as a gateway to a range of benefits because it allows people to inform themselves. For example, in one cross-country study, literacy was associated with a decrease in child mortality of 67% and an increase in women’s empowerment of 0.68 standard deviations. In OECD countries, foundational skills including literacy are associated with higher levels of political efficacy and volunteering. In a review of accountability interventions across low- and middle-income countries, literacy was found to be a key success factor. Literacy also increases the likelihood of remaining in school – for example a study in Kenya found that a one standard deviation increase in literacy was associated with a 40% decrease in school drop-out.
Of course, it is not true that everyone who can read makes wise decisions. However, it is surely true that being able to read is a prerequisite to consuming information that can enhance the ability to make evidence-informed decisions. Effective altruists prioritize interventions that maximize human agency. By leaving massive swathes of the global population without basic literacy, we face a massive talent bottleneck, reducing humanity's collective capacity to solve complex global coordination failures.
While the importance of literacy is clear, its tractability is nuanced. At a surface level, teaching a child to read is a solved problem. But establishing a program that delivers effective literacy instruction to children at scale in locations with weak education systems is notoriously difficult. A recent review revealed that the median evaluated literacy intervention yields very low impact; even programs considered successful in terms of effect sizes generally only increased reading fluency by a few words per minute, far below what is needed for transformational change.
This tractability barrier exists because literacy education is not a technical fix. To draw on the Harvard Building State Capability framework for change processes, it is an “implementation-intensive service delivery” meaning that it is a complex, decentralized behaviour-change problem. However, it is critical to emphasize that high impact is possible. The Life You Can Save has just selected four impactful charities -the Luminos Fund, Building Tomorrow, Educate Girls and Youth Impact - that are demonstrating high impact on learning outcomes in low- and middle- income countries.
The focus on impactful organisations is important. We cannot assume that a new organisation and/or work in a new context will manage to achieve high impact just because an intervention has been proven to work elsewhere. Even approaches such as "Teaching at the Right Level" (TaRL) or "Structured Pedagogy" which can be highly impactful in some cases can also fail to shift instructional practice and therefore fail to result in meaningful impact when transferred to a new context and/or implementer.
To navigate this, our funding strategy should follow two distinct tracks. First, we should fund organizations that are already on the ground and successfully achieving impact within that specific political economy. Second, where there are no existing impactful programs, we should fund promising organizations to test, adapt and scale, with funding growing as the evidence of impact within that context develops.
There is a huge deficit in the spending needed to address the learning crisis in low- and middle-income countries. Globally, UNESCO estimates that an additional $97 billion USD is needed annually to achieve Sustainable Development Goal 4 (SDG4) with the biggest financing gap being in sub-Saharan Africa. Annual donor funding for education covered only 13% of this deficit in 2023, and this proportion has declined substantially due to decreased spending by key bilateral donors. Official Development Assistance targeted to education is less than half the amount allocated to health.
From an EA perspective, the true neglectedness is even more pronounced when we look at effective funding. A sector-wide review by RTI discovered that the vast majority of major international education donors could not point to a single intervention that successfully improved literacy at scale. Consequently, while billions are spent on educational inputs (like building schools or procuring textbooks), targeted, evidence-backed funding directed at actual learning outcomes is profoundly neglected.
Applying the ITN framework reveals that the global learning crisis represents a massive, highly neglected pool of human potential. The historical EA hesitation to fund education was driven by concerns regarding impact and tractability. However, recent empirical models confirm that when an educational intervention successfully improves learning, its lifetime returns match the profiles of our most celebrated global health interventions. And there are outstanding organizations which are proven to deliver impact at scale and which could absorb substantial funding to scale further.
The lesson for effective altruists is to adopt a two-pronged approach:
By balancing the expansion of what works today with a pipeline to discover what works tomorrow, the EA community can transform education funding from a gamble into a highly calculated, impactful investment.
Helpful!
On the impact side, some would suggest smaller numbers than the ones you point to here after discounting, and on the cost side some programs are fairly expensive. This all works to make the cost-effectiveness case less attractive, which explains limited attention in EA, I think.
We recently wrote a paper with colleagues from Gates trying to incorporate health and social benefits alongside economic ones to some set of schooling (and schooling related interventions)*, and find that they more than double the benefits, which makes them more competitive under some bar.
(*Though in our paper we just model additional years of schooling, rather than learning. One would imagine the benefits of more learning would be larger, but I haven't followed that literature as closely.)