Footnotes for "How we work, #3: Our analyses involve judgment calls" blog post

[1] As an aside, GiveWell is often associated with randomized controlled trials (RCTs), which underpin many of our recommendations, but we also consider other experimental or observational evidence, especially when RCTs aren't feasible. RCTs provide particularly strong causal evidence because of how they're designed: the idea is to randomly allocate people or sets of people (e.g., villages or schools) to either a "treatment" group (which receives the intervention in question) or a "control" group (which doesn't). Randomization should mean the two groups are similar as long as the study population is large enough, so any substantial differences in outcomes can be attributed to the intervention.

Examples of recent research questions for which we've reviewed observational evidence and for which experimental evidence is hard to come by (for ethical or other reasons) include:

  • The effect of community delivery of oral rehydration solution on mortality from diarrhea
  • The effect of benzathine penicillin G (an antibiotic) in preventing negative health outcomes from congenital syphilis
  • The health effects of alcohol consumption

[2] Studies on the mortality impact of historical water quality improvements suggest that water quality interventions may reduce mortality from non-waterborne diseases. This is called the Mills-Reincke phenomenon and provides a plausible explanation for the unexpectedly large estimate of the impact of chlorination on mortality in children under five.

[3] See this section of our intervention report for more detail.

[4] Briefly, Haushofer et al. 2021 finds a 63% reduction in all-cause mortality in children under five (95% confidence interval of 13% to 105%). It estimates that chlorination was 18 percentage points higher in the treatment than the control group. At face value, that result implies that drinking clean water averts most of the deaths that would otherwise occur among young children and/or there were very large spillover effects from water chlorination programs (even at the low end of the 95% confidence interval). Our view is that this is unlikely to be the case. See footnote 39 on this page for more detail.

[5] See footnote 2 here.

[6] See this section of our cost-effectiveness analysis.

[7] Or, to give another point of comparison, we're using an estimate where averting a year of clubfoot is half as good as doubling a person's income for a year. For a table of our moral weights for averting deaths and increasing consumption, see this document.

[8] See this grant page, including the table in the summary. The figures in the above paragraph don't include the additional benefit we estimate from potential increases in income in adulthood if clubfoot is successfully treated.

[9] See this grant page for data.