President of the Social Science Research Council; Professor of Politics/Data Science/Law and Director of the Public Safety Lab at New York University; and Co-Director of the Criminal Justice Expert Panel.
Anna Harvey
Karma: 27
[Cause Exploration Prizes] Social and Behavioral Science R&D
CGD reports the cost of an Astra Zeneca-like Covid vaccine (0.75 efficacy) as 3$/dose. But perhaps more to the point, the costs of developing, producing, and distributing Covid vaccines are by now largely sunk. They are widely available free of cost in areas with low vaccination rates. Yet vaccination rates in LMIC remain stubbornly low (and aren’t that great in many HIC). We don’t know which interventions might work to increase vaccination rates. We have some evidence that scaled SMS campaigns increase Covid vax rates in the US; could that work in LMIC? We don’t know! Social signaling increased childhood immunization rates in Sierra Leone. Could social signaling increase Covid vax rates in LMIC? We don’t know!
In order for an intervention that increases Covid vax rates to have greater than 1000x ROI, we need each $100K invested in a vaccine uptake intervention to yield greater than $100M in benefits. Using the Open Philanthropy benchmarks of 32 DALYS per adult death, and $100,000 per DALY, averting an adult death is worth approximately $3.2M. Assuming reducing mortality is the only benefit of Covid vaccine uptake, for each $100K in the costs of an intervention to increase Covid vax uptake, we would need to avert more than 31.25 adult deaths ($100M/$3.2M). Recent estimates of excess mortality rates due to Covid in LIC range as high as 0.007. A population-wide COVID mortality rate of even 0.004 implies that we would see 32 deaths in a population of 8,000 unvaccinated adults. With vaccine efficacy of 0.75, we would need to fully vaccinate about 10,667 adults to avert 32 deaths.For an investment of $100,000, we therefore need each full vaccination (as a result of an intervention) to cost no more than about $9.37. Given that SMS and social signaling campaigns are both relatively low cost (and if anything have diminishing marginal costs), and that there are presumably other interventions that could overcome vaccine hesitancy, that seems within reach.
Interesting to see the skepticism about the existence of currently unknown interventions with high ROI (e.g., large impacts on reducing mortality). There seem to be a very large number of problems for which we have not yet identified effective interventions. For example, given what we know about the effectiveness of Covid vaccines at averting death, and the fact that, despite availability, currently only about 18% of residents of LICs have received even just one dose of a Covid vaccine, an intervention that cost-effectively increased Covid vaccination demand would seem to be well worth the investment in R&D to find such an intervention. The same is presumably true for interventions that cost-effectively increase demand for/take-up of other life-saving vaccinations.
Thanks for this.
I think our point is more that there is not enough social and behavioral science R&D anywhere (in or out of academia) specifically targeted at EA goals, i.e., finding interventions that can generate >1000x ROI at scale. If that R&D were really happening outside of academia, then great! But the recent retrenchment at GiveWell to just 4 interventions with sufficient evidence to be “top charities” suggests that we are not in fact investing sufficiently in R&D to find highly cost-effective causes.
Without targeted funding, social and behavioral science R&D to find interventions that clear the >1000x ROI threshold won’t happen on its own: academic journals and public funders of science don’t prioritize that particular goal. But incentives matter! If we want to find those interventions, all we have to do is fund the work.