EA - Red-teaming GiveWell's Recommendation of Mass Deworming by Lucas Lewit-Mendes

Link to original articleWelcome to The Nonlinear Library, where we use Text-to-Speech software to convert the best writing from the Rationalist and EA communities into audio. This is: Red-teaming GiveWell's Recommendation of Mass Deworming, published by Lucas Lewit-Mendes on July 23, 2022 on The Effective Altruism Forum. Primary author: Lucas Lewit-Mendes Secondary authors: Isobel Phillips, Sanjay Joshi tl;dr: We (SoGive) estimate that mass deworming is 6-10x more cost-effective than GiveDirectly's cash transfers. This is less cost-effective than GiveWell's estimate of 12-19x, but remains close to, albeit slightly below, their current 10x funding bar. The discrepancy is primarily driven by our approach to quantifying net gains in adult earnings, which we argue are diminished after accounting for labour market competition. Red-Teaming GiveWell's Recommendation of Mass Deworming As noted in our recent post [insert link], we (SoGive) have been conducting red-teaming of some of GiveWell's recommended top charities, starting with mass deworming programmes. Much has been written on deworming, from a landmark trial in Kenya, to the infamous "worm wars", to an apparently encouraging 20-year follow-up study. GiveWell moved around $40 million to support mass deworming programmes in the year to June 2022. After ~200 hours of research, we hope to shed some light on some relatively unexplored aspects of the mass deworming conversation. Our major update is that we now expect deworming to have less chance of huge economic impact, because competition between workers in the labour market may result in "economic losers". Alongside other small updates, this reduces cost-effectiveness to 6-10x GiveDirectly's cash transfers (varying by charity), down from GiveWell's estimate of 12-19x. We would like to express our gratitude to GiveWell for their correspondence throughout the research process. This post will highlight our key findings from our full intervention report, covering the following topics: #SectionOne-sentence summaryIntroductory content1Statistical significanceThe notably large effects on earnings and consumption from the Busia experiment are not statistically significant, but GiveWell sufficiently accounts for this uncertainty. 2Black box problemThe causal mechanism between deworming and long-run earnings gains is a "black box". SoGive differs from GiveWell on economic losers3Economic losersGains in adult earnings for some dewormed children may lead to "economic losers". SoGive believes it is important to give careful consideration to health effects4Direct health effectsThe direct health effects of mass deworming appear to be small on average. 5ComorbiditiesHealth effects are especially uncertain in the presence of comorbidities. 6Drug resistanceHuman worm parasites have not developed drug resistance, but the threat remains. Despite the significant amount of effort already expended, our work on deworming is not complete. We remain excited to continue exploring some of the other areas that we outline towards the end of this post, including unprogrammed deworming (i.e. how much deworming would happen anyway without mass deworming programmes), the implications of climate change for deworming, and more on the health aspects of deworming. Background on Mass Deworming GiveWell-recommended mass deworming programmes currently operate in sub-Saharan Africa and South Asia, treating children with two types of parasitic worm infections - schistosomiasis and soil-transmitted helminthiasis (STH). The scale of infections is enormous. Around 200 million people have schistosomiasis, while around 1.5 billion have STH. The health effects of worm infections can be severe in rare cases, including organ damage, intestinal inflammation, intestinal obstruction, and impairment of nutrient intake. But since worm infections often coexist with other morbidities, precise data on how often worm infections cause particular severe symptoms is hard to come by. Worm infection is not binary ...