GiveWell’s Top Charities Are Increasingly Hard to Beat

Link post

GiveWell re­views their “near-ter­mist hu­man-cen­tric OpenPhil grants” (i.e., crim­i­nal jus­tice re­form, im­mi­gra­tion policy, land use re­form, macroe­co­nomic sta­bi­liza­tion policy, and sci­en­tific re­search) grant­mak­ing and find that many of these grants have sub­stan­tial risk of failing to ex­ceed the cost-effec­tive­ness of GiveWell’s top char­i­ties.

It ap­pears that over the past sev­eral years, the es­ti­mated cost-effec­tive­ness of GiveWell’s top char­i­ties (as a class) has in­creased higher than ex­pected, whereas so far the es­ti­mated cost-effec­tive­ness of “near-ter­mist hu­man-cen­tric OpenPhil grants” (as a class) has not pro­duced as many hits at a similar or bet­ter level as ex­pected. There are also fur­ther notes around com­par­ing the ro­bust­ness of these es­ti­ma­tions and ad­di­tional con­sid­er­a­tions for why non-GiveWell near-ter­mist hu­man-cen­tric grant­mak­ing is valuable.

OpenPhil says they’re “plan­ning to write more at a later date about the cost-effec­tive­ness of [their] ‘long-ter­mist’ and an­i­mal-in­clu­sive grant­mak­ing and the im­pli­ca­tions for our fu­ture re­source al­lo­ca­tion” which I’m es­pe­cially ex­cited about see­ing next.