What is the expected value of creating a GiveWell top charity?

Ear­lier this year, GiveWell Ex­per­i­men­tal pre­dicted that we, the team be­hind Char­ity Science Health (CSH), have a 15% chance of be­com­ing a GiveWell top char­ity by giv­ing sea­son 2019. That sounds pretty cool. But what does it mean for our plans? How good is it to be a top char­ity? And how does hav­ing a 15% chance of achiev­ing top char­ity sta­tus com­pare to other things we could be do­ing with our time?

While it would take a lot of time to treat this ques­tion rigor­ously, a rough ini­tial sketch seems very pru­dent for guid­ing our staff plans as we en­ter 2017 and de­cide whether to con­tinue as-is, scale up, or scale down, with a keen mind to what we could be do­ing in­stead. An ini­tial anal­y­sis and back of the en­velope calcu­la­tion seemed like the best way to in­form our in­tu­itions on this topic, so we set out to make one.

Our es­ti­mate is based on as­sum­ing that GiveWell top char­i­ties have im­pact ac­cord­ing to four fac­tors: (1) fund­ing from GiveWell (both from Good Ven­tures and di­rected by the in­ter­est of GiveWell donors, which will likely add up to $108M a year or more), (2) fund­ing from non-GiveWell sources (e.g., other foun­da­tions), (3) how cost-effec­tive they are, and (4) the to­tal room for more fund­ing of the or­ga­ni­za­tion.

Fund­ing from GiveWell

Between 2011 and 2015, the four top GiveWell charites col­lec­tively re­ceived $166M, with $108M com­ing from just 2015. If we di­vide the $166M over five years evenly among all four char­i­ties, that’s $8.3M per year per char­ity. A 15% chance of us achiev­ing this is thus worth an ex­pected value of $1.3M a year, which di­vided among three full-time em­ployee equiv­a­lents (FTE) is $433K in ex­pected value per per­son.

How­ever, GiveWell is mov­ing a lot more money now than it used to, es­pe­cially with Good Ven­tures (though due to an un­usu­ally large 2015 grant from GiveWell to GiveDirectly, the trend may be ex­ag­ger­ated). Ad­di­tion­ally, there’s some­what of a power law even within GiveWell top char­i­ties. In 2015, GiveDirectly got ~50% of all the fund­ing, the Against Malaria Foun­da­tion (AMF) got 35% of the fund­ing, and SCI only got 3% of the fund­ing. While we have a 15% chance of be­ing a top char­ity, there’s still a good chance we’d end up on the lower half and get only 3%.

So let’s as­sume now that in­stead of it be­ing the 2011-2015 track record of $166M over five years (av­er­ag­ing to $33.2M per year), giv­ing in­stead con­tinues lin­early at roughly $80M a year. We’ll also as­sume that we’ll break up our 15% chance into a 5% chance of get­ting into the top three (along with AMF and GiveDirectly) and re­ceiv­ing roughly one third of the fund­ing and a 10% chance of get­ting ~5% of the fund­ing like De­worm the World and SCI. 5% chance of 33% of $80M/​yr plus a 10% chance of 5% of $80M/​yr is $1.32M/​yr + 400K/​yr = $1.72M/​yr or $344K per per­son per year.

Lastly, it seems quite plau­si­ble that the ad­di­tion of a new GiveWell top char­ity would bring in more over­all money to GiveWell top char­i­ties. At min­i­mum, an eighth top char­ity would likely lead Good Ven­tures to make an ad­di­tional in­cen­tive grant of $2.5M that they would not have oth­er­wise made. How­ever, it also seems plau­si­ble to me that a new char­ity, es­pe­cially one fo­cused on some­thing other than malaria or de­worm­ing, would bring ad­di­tional in­ter­est to GiveWell from new donors or old donors in­ter­ested in new ar­eas. If we thought this effect would, to­gether with the ad­di­tional Good Ven­tures grant, boost the to­tal money moved by an ad­di­tional $10M, this would boost our ex­pected of $344K per per­son per year to $364K per per­son per year.

Com­par­a­tive cost-effectiveness

Another ma­jor fac­tor is how cost effec­tive we will be, rel­a­tive to other GiveWell top char­i­ties. For ex­am­ple, if we take GiveWell’s 2016-2017 es­ti­mates liter­ally, AMF is ~4x as cost-effec­tive as GiveDirectly (though I ex­pect GiveWell would kindly ask us not to take these es­ti­mates liter­ally).

While our cur­rent es­ti­mates don’t place us to be as cost-effec­tive as AMF, we think we have a good shot of be­ing more cost-effec­tive than GiveDirectly (GD). If we could even be 2x GD (or 0.5x AMF) and shared 5050 with GD for money from GiveWell’s sources, we’d be effec­tively dou­bling the im­pact value of that share.

To make this more con­crete, GD got $19M in 2015 ex­clud­ing Good Ven­tures and AMF got $15M. If we ad­justed those num­bers to the literal cost effec­tive es­ti­mates and did not ad­just for diminish­ing marginal re­turns, we might be able to say some­thing roughly like the $19M to GD be­ing worth the same as $4.75M to AMF, or 1.9M in cost-effec­tive­ness ad­justed dol­lars. The to­tal among AMF and GD would then be 19.75M cost-effec­tive­ness ad­justed dol­lars.


If Char­ity Science Health was a 2x GD char­ity, and we still as­sumed that AMF got $15M but split the re­main­ing money 50-50 be­tween GD and CSH (So that GD got $9.5M and CSH got $9.5M), the new cost-effec­tive­ness ad­justed to­tal (rel­a­tive to AMF) would be 15M cost-effec­tive­ness ad­justed dol­lars de­rived from AMF’s im­pact, 2.375M cost-effec­tive­ness ad­justed dol­lars from GD and 4.75M cost-effec­tive­ness ad­justed dol­lars from CSH, for a to­tal of 22.125 cost-effec­tive­ness ad­justed dol­lars. Thus the ad­di­tion of CSH could be mod­eled as a net gain of 2.375M cost-effec­tive­ness ad­justed dol­lars, even though the to­tal amount of money donated has not changed.

The more cost-effec­tive the new top char­ity, the bet­ter. If CSH were in­stead the same as AMF and still split with GD 5050, the net gain would in­crease from 2.375M cost-effec­tive­ness ad­justed dol­lars to 7.125M cost-effec­tive­ness ad­justed dol­lars.

Over­all, while I think the con­cept of “cost-effec­tive­ness ad­justed dol­lars” is use­ful for sketch­ing out one po­ten­tial area of im­pact for CSH, I hes­ti­tate to take the no­tion liter­ally and think that a lot more work would be needed to sketch out the con­cept more rigor­ously. In prac­tice, how­ever, com­par­ing our in­ter­nal rough guess cost-effec­tive­ness es­ti­mate of Char­ity Science Health with a rough guess of the av­er­age cost-effec­tive­ness of GiveWell top char­i­ties from 2016-2017, the two num­bers seem to be roughly even, so I don’t ex­pect any prac­ti­cal benefit from this fac­tor.

Fund­ing from sources out­side GiveWell

While peo­ple in the EA move­ment may not think about it much, GiveWell and Good Ven­tures are not the biggest fun­ders of non-prof­its out there. A po­ten­tial GiveWell top char­ity could also hope to vie for the at­ten­tion of other fun­ders, like the Gates Foun­da­tion, the Lam­pert Fam­ily Foun­da­tion (fun­ders of New In­cen­tives), the Global In­no­va­tion Fund, YCom­bi­na­tor, and oth­ers. While the GiveWell stamp of ap­proval could cer­tainly help lead these fun­ders to a char­ity, the best char­i­ties could cer­tainly find fund­ing in­de­pen­dently from GiveWell’s en­dorse­ment.

A new GiveWell top char­ity could un­lock hun­dreds of thou­sands, if not mil­lions, of dol­lars from large fun­ders, and it’s quite likely the money coun­ter­fac­tu­ally would not have gone to as good of a char­ity (by GiveWell’s stan­dards).

Room for more funding

It doesn’t mat­ter if you can at­tract tens of mil­lions of dol­lars if you can­not effec­tively spend tens of mil­lions of dol­lars. For ex­am­ple, GiveWell de­ter­mined that De­worm the World could only make use of ~$15.5M in “ex­e­cu­tion level 1 and 2” fund­ing over 2017. We’re highly un­cer­tain about our own po­ten­tial for room for more fund­ing in fu­ture years, so this could re­ally cap our po­ten­tial for im­pact even if we oth­er­wise suc­ceed. If we were only able to take in $5M, our abil­ity to move $20M would not mat­ter.

Staff allocation

Right now, we have all our high-tal­ent “se­nior” staff al­lo­cated to run­ning CSH. How­ever, there could come a time when we could run CSH in a more “pas­sive” mode, con­tin­u­ing to have nearly as high im­pact, though with­out seek­ing as strongly to scale. This is similar to what CSH staff have already done with our prior pro­ject of Char­ity Science Outreach, drop­ping the team from ~5.5 FTE to ~1 FTE while still con­tin­u­ing to move a good amount of money.

Thus, in a hy­po­thet­i­cal fu­ture, per­haps CSH could spend $10M per year al­most in­definitely while the CSH “se­nior” staff move on to start a differ­ent char­ity. If we as­sumed that “pas­sive” mode lasted for a decade, the pre­sent value of that $100M over ten years would be worth $55.8M to­day (as­sum­ing a 6% in­ter­est rate), or $18.6M per cur­rent “se­nior” staff mem­ber (3 FTE), though we still would have to ad­just for the coun­ter­fac­tual value of the non-se­nior staff mem­bers.

Conclusion

A new GiveWell top char­ity will have im­pact by (a) at­tract­ing some new fund­ing through GiveWell that would not have oth­er­wise been at­tracted, (b) be­ing more cost-effec­tive than cur­rent GiveWell top char­i­ties, and (c) by at­tract­ing new fund­ing from non-GiveWell foun­da­tions. How­ever, it is capped by a large chance of failure, room for more fund­ing, staff time from peo­ple who could oth­er­wise be do­ing other great things.

When I try to model all these fac­tors out in the fol­low­ing Guessti­mate model, I get an es­ti­mate that the to­tal efforts of our team (in­clud­ing se­nior staff, non-se­nior staff, and vol­un­teers) are roughly equal to, in ex­pec­ta­tion, the work of a team of full-time equiv­a­lent peo­ple each earn­ing to give $400K a year to GiveWell top char­i­ties, with our 95% con­fi­dence in­ter­val rang­ing from $220K/​yr to $720K/​yr. This com­pares fa­vor­ably, as $400K donated per year is, I think, higher than what a typ­i­cally am­bi­tious and skil­led EA could be ex­pected to earn to give and is much higher than the earn­ing to give lev­els found in 80,000 Hours 2014 re­port.

How­ever, more work has to be done to re­fine this es­ti­mate. For ex­am­ple, this does not take into ac­count the tricky as­pect of dou­ble count­ing, since our im­pact would be par­tially in mov­ing the dona­tions of oth­ers. Also, this does not sep­a­rate out the fact that the coun­ter­fac­tu­als and in­di­vi­d­ual im­pact of all our staff and vol­un­teers are definitely not the same. Us­ing the same model, at­tribut­ing 100% of the im­pact to se­nior staff and 0% of the im­pact to any­one else, shifts the es­ti­mate to $4.1M per se­nior staff year ($550K/​yr to $25M/​yr).

But over­all, this makes found­ing a GiveWell top char­ity could be a very high earn­ing (though also some­what high var­i­ance) ca­reer choice. Hope­fully this es­ti­mate will con­tinue to be fur­ther re­fined as we learn more about our chances of suc­cess and failure!

-

Thanks to Joey Savoie, Justis Mills, and Mar­cus Davis for re­view­ing a draft of this es­say.

Up­date 19 Dec: Thanks to Vipul Naik and Owen Cot­ton-Bar­ratt for point­ing out prob­lems with the model. After tak­ing their feed­back into ac­count, my es­ti­mate for to­tal staff time (in­clud­ing vol­un­teers) was re­vised from a mean of $470K/​yr (95% in­ter­val: $150K/​yr to $1.4M/​yr) to a mean of $400K/​yr (95% in­ter­val: $220K/​yr to $720K/​yr). The es­ti­mate for se­nior staff time was re­vised from a mean of $4.1M/​yr (95% in­ter­val: $550K/​yr to $25M/​yr) to a mean of $3.7M/​yr (95% in­ter­val: $700K/​yr to $12M/​yr).

Thanks also to Vipul Naik for spot­ting sev­eral ty­pos.