Saving expected lives at $10 apiece?

On this fo­rum, I have referred sev­eral times to a pa­per show­ing strik­ing cost-effec­tive­ness of get­ting pre­pared for global agri­cul­tural catas­tro­phes. It is now pub­lished here. I ac­knowl­edge sev­eral EAs who re­viewed the pa­per. The ab­stract is be­low; we also an­a­lyze re­turn on in­vest­ment and find ex­tremely high val­ues. We do not quan­ti­ta­tively com­pare to the effec­tive­ness of work­ing on other global catas­trophic risks, but be­cause this is such a lev­er­aged op­por­tu­nity, it is likely to com­pare fa­vor­ably. The num­ber of ex­pected lives lost per day de­lay of get­ting pre­pared is what con­vinced me to give sig­nifi­cant frac­tion of my own money to the effort.

I am in­ter­ested in your feed­back on the as­sump­tions, and also how to com­mu­ni­cate the cost-effec­tive­ness to EAs and the gen­eral pub­lic. The char­ity we are start­ing would not only do the di­rect work to get pre­pared, but it would also hope­fully mo­ti­vate ad­di­tional fund­ing. This should be even more cost-effec­tive than the di­rect in­ter­ven­tions, but I would prob­a­bly be con­ser­va­tive and ig­nore that. For most au­di­ences, I would also be con­ser­va­tive and ig­nore far fu­ture benefits. Another source of con­ser­vatism is that our bud­get will be small com­pared to the tens of mil­lions of dol­lars re­quired to do sig­nifi­cant prepa­ra­tion, so we can choose the most cost-effec­tive ac­tivi­ties. Much of the prepa­ra­tions for ~10% global agri­cul­tural short­falls would be valuable to pre­pare for ~100% global agri­cul­tural short­falls (large comet/​as­ter­oid, su­per vol­canic erup­tion, and nu­clear win­ter). Ig­nor­ing these benefits is an­other source of con­ser­vatism. There also sources of con­ser­vatism that af­fect over­all cost-effec­tive­ness, but not cost per life saved, in­clud­ing pre­serv­ing bio­di­ver­sity. We also ig­nore the re­duc­tion of the cost of food dur­ing the catas­tro­phe for the peo­ple who would have sur­vived any­way. Pre­limi­nary calcu­la­tions in­di­cate that this would make the cost to de­vel­oped coun­tries (as­sumed to be the donors) of get­ting pre­pared net nega­tive, mean­ing net nega­tive cost to save ex­pected lives. But I have not yet writ­ten that pa­per, so let’s re­turn to the con­clu­sions of the pub­lished pa­per.

The gen­eral pub­lic typ­i­cally does not do very well with un­cer­tainty, so I was think­ing of us­ing the me­dian value of $10 to save an ex­pected life. I think the me­dia would fix­ate on the lower bound of sav­ing ex­pected lives for $.30 apiece. Might this be ok be­cause of the large con­ser­vatism above?


The liter­a­ture sug­gests there is about a 1 % risk per year of a 10 % global agri­cul­tural short­fall due to catas­tro­phes such as a large vol­canic erup­tion, a medium as­ter­oid or comet im­pact, re­gional nu­clear war, abrupt cli­mate change, and ex­treme weather caus­ing mul­ti­ple bread­bas­ket failures. This short­fall has an ex­pected mor­tal­ity of about 500 mil­lion peo­ple. To pre­vent such mass star­va­tion, al­ter­nate foods can be de­ployed that uti­lize stored bio­mass. This study de­vel­oped a model with liter­a­ture val­ues for vari­ables and, where no val­ues ex­isted, used large er­ror bounds to rec­og­nize un­cer­tainty. Then Monte Carlo anal­y­sis was performed on three in­ter­ven­tions: plan­ning, re­search, and de­vel­op­ment. The re­sults show that even the up­per bound of USD 400 per life saved by these in­ter­ven­tions is far lower than what is typ­i­cally paid to save a life in a less-de­vel­oped coun­try. Fur­ther­more, ev­ery day of de­lay on the im­ple­men­ta­tion of these in­ter­ven­tions costs 100–40,000 ex­pected lives (num­ber of lives saved mul­ti­plied by the prob­a­bil­ity that al­ter­nate foods would be re­quired). Th­ese in­ter­ven­tions plus train­ing would save 1–300 mil­lion ex­pected lives. In gen­eral, these solu­tions would re­duce the pos­si­bil­ity of civ­i­liza­tion col­lapse, could as­sist in pro­vid­ing food out­side of catas­trophic situ­a­tions, and would re­sult in billions of dol­lars per year of re­turn.


The value
is not of type