GiveWell money moved in 2015: a review of my forecast and some future predictions

This post as­sumes some fa­mil­iar­ity with the work of char­ity eval­u­a­tor GiveWell, but here’s a quick back­grounder for those not fa­mil­iar: GiveWell lists a small set of top char­i­ties to donate to, and pub­lishes de­tailed re­views of each. The list for 2015 com­prises Against Malaria Foun­da­tion (malaria bed­net dis­tri­bu­tion), GiveDirectly (un­con­di­tional cash trans­fers), De­worm the World Ini­ti­a­tive (de­worm­ing), and Schis­to­so­mi­a­sis Con­trol Ini­ti­a­tive (de­worm­ing). GiveWell’s recom­men­da­tions have ac­quired the sta­tus of a gold stan­dard for con­ven­tional, unattached donors, and GiveWell has a par­tic­u­larly strong rep­u­ta­tion among many of the peo­ple who read this site and iden­tify as effec­tive al­tru­ists. GiveWell also works closely with Good Ven­tures, the pri­vate foun­da­tion of Face­book co-founder Dustin Moskovitz and his wife Cari Tuna. Good Ven­tures has made sig­nifi­cant dona­tions to GiveWell’s top char­i­ties, mostly at GiveWell’s sug­ges­tion, with the amount reach­ing $44.4 mil­lion (plus $1 mil­lion to stand­out char­i­ties, plus a sep­a­rate three-year $25 mil­lion grant to GiveDirectly) in 2015.

Two days ago, GiveWell pub­lished up­dated (and fi­nal) met­rics for money moved and web traf­fic in 2015. The key high­light: they moved al­most $40 mil­lion from donors out­side of Good Ven­tures. In a De­cem­ber 2015 post, I had made fore­casts on GiveWell’s non-Good Ven­tures money moved in 2015. In this post, I’ll look at the ac­tual met­rics, com­pare with my pre­dic­tions, and iden­tify where I was wrong.

Table of contents

0. TL;DR

  • GiveWell’s es­ti­mate of money moved, $39.73 mil­lion, falls be­tween the 97.5th per­centile and 99th per­centile of my dis­tri­bu­tion for how much money GiveWell would move to its top char­i­ties.

  • I erred in two big ways: I un­der­es­ti­mated the ex­tent to which $1 mil­lion+ dona­tions would dom­i­nate the land­scape, and I un­der­es­ti­mated the amount of money that would be moved to GiveDirectly. I be­lieve these er­rors are re­lated: I think many of the large dona­tions went to GiveDirectly (GiveWell does not re­lease data on in­di­vi­d­ual large dona­tions).

  • I also erred some­what by un­der­es­ti­mat­ing the sheer in­crease in num­ber of donors (though this was not a driv­ing fac­tor be­hind my un­der­es­ti­ma­tion of the to­tal amount of money moved). Re­lat­edly, I over­es­ti­mated the role of re­ten­tion/​up­sell (which was neg­ligible from 2014 to 2015) and (prob­a­bly) un­der­es­ti­mated the role of in­creased dis­cov­ery and brand aware­ness of GiveWell in 2015.

  • At a math­e­mat­i­cal level, I erred by failing to ap­ply the fact that when adding skewed dis­tri­bu­tions, the me­dian of the sum can be sig­nifi­cantly greater than the sum of the me­di­ans. Thus, even though many of my finer pre­dic­tions were roughly on tar­get, my over­all es­ti­ma­tion was way off-base. I hope to cor­rect that er­ror when mak­ing fu­ture pre­dic­tions.

  • In this post, I out­line some rea­sons why GiveDirectly might be a more at­trac­tive propo­si­tion for ul­tra high net-worth in­di­vi­d­u­als (UHNWIs) than I had ex­pected.

  • I dis­cuss why I be­lieve most met­rics will flatline in 2016, but I think there will still be growth be­cause even one or two ex­cep­tions to the flatlin­ing trend can drive a de­cent amount of growth. I ex­pect that the June 2016 recom­men­da­tion re­fresh, and the re­sponse to it, will yield more in­sight into how 2016 will shape up.

  • I con­sider im­pli­ca­tions for po­ten­tial donors, non­mon­e­tary con­trib­u­tors, and those vet­ting GiveWell as an or­ga­ni­za­tion.

  • I time-limited my­self while writ­ing the post, so it is not as pol­ished as I might make it with more time.

1. Analysis

1.1. Fore­cast and com­par­i­son with reality

In my fore­cast, I in­cluded a dis­tri­bu­tion of my es­ti­mates for to­tal money moved in 2015, as well as money moved by char­ity.

Here is the dis­tri­bu­tion by per­centile:

  • 1st per­centile: $8.8 million

  • 2.5th per­centile: $11.6 million

  • 10th per­centile: $14.7 million

  • 25th per­centile: $16.2 million

  • 50th per­centile: $18.7 million

  • 75th per­centile: $24.5 million

  • 90th per­centile: $28.6 million

  • 97.5th per­centile: $33.1 million

  • 99th per­centile: $45 million

Here is the dis­tri­bu­tion for top char­i­ties (50% prob­a­bil­ity for each):

  • Against Malaria Foun­da­tion: $8 mil­lion - $17 million

  • GiveDirectly: $1 mil­lion - $4 million

  • Schis­to­so­mi­a­sis Con­trol Ini­ti­a­tive: $1 mil­lion - $4 million

  • De­worm the World Ini­ti­a­tive: $250,000 - $4 million

The ac­tual amount of money moved, $39.73 mil­lion, fell some­where be­tween the 97.5th and 99th per­centile in my dis­tri­bu­tion. Here are the amounts of money moved by char­ity:

  • Against Malaria Foun­da­tion: $15.45 mil­lion, within my 50% in­ter­val, and 56.2% of the Ex­e­cu­tion Level 1 fund­ing gap iden­ti­fied by GiveWell.

  • GiveDirectly: $19.37 mil­lion, well out­side my 50% in­ter­val, and 78.1% of the Ex­e­cu­tion Level 1 fund­ing gap iden­ti­fied by GiveWell.

  • Schis­to­so­mi­a­sis Con­trol Ini­ti­a­tive: $2.67 mil­lion, within my 50% in­ter­val, and 54.2% of the Ex­e­cu­tion Level 1 fund­ing gap iden­ti­fied by GiveWell.

  • De­worm the World Ini­ti­a­tive: $1.08 mil­lion, within my 50% in­ter­val, and 33.7% of the Ex­e­cu­tion Level 2 fund­ing gap iden­ti­fied by GiveWell (GiveWell did not leave any Ex­e­cu­tion Level 1 fund­ing gap for the or­ga­ni­za­tion)

In other words, in three of the four cases, the ac­tual amount of money moved was within the mid­dle 50% of my es­ti­mate. But the fourth case was very far off. In fact, if GiveDirectly had stayed at the mid­point of my pre­dicted in­ter­val for it, the to­tal money moved would have been safely in my 50% in­ter­val. To quite an ex­tent, this and other er­rors in my es­ti­mates boil down to my failing to keep in mind that the sum of the me­di­ans is not the me­dian of the sums, and that, in fact, for not-very-tightly-cor­re­lated dis­tri­bu­tions skewed to the right, the me­dian of the sum can be sig­nifi­cantly greater than the sum of the me­di­ans (ba­si­cally, if even one of the val­ues is a lit­tle bit more to the right of the me­dian, the sum cor­re­spond­ingly be­comes higher than the me­dian).

The up­shot: I was way off on the whole, but al­most the en­tirety of the ex­pla­na­tion for me be­ing way off is re­lated to GiveDirectly at­tract­ing sig­nifi­cantly more money than I had an­ti­ci­pated. Without that, my es­ti­mates would have been mod­er­ate un­der­es­ti­mates, and Elie Hassen­feld’s Oc­to­ber es­ti­mate of mov­ing a lit­tle over $20 mil­lion to top char­i­ties would have been spot on.

Caveat 1: Note that the per­centages of fund­ing gap met are po­ten­tially in­ac­cu­rate since they do not ex­clude dona­tions made prior to GiveWell’s re­lease of its year-end recom­men­da­tions, whereas the fund­ing gaps iden­ti­fied by GiveWell took into ac­count dona­tions made already. Still, they offer some sort of rough es­ti­mate of how much the money moved to or­ga­ni­za­tions com­pared with GiveWell’s iden­ti­fied fund­ing gaps for them.

Caveat 2: Both GiveDirectly and AMF re­ceived some money with un­known at­tri­bu­tion, and GiveWell es­ti­mates that it in­fluenced a to­tal of $7-10 mil­lion of these dona­tions (about $5.5 mil­lion to AMF and the rest to GiveDirectly). If these ad­di­tional grants are counted, then money moved to each of AMF and GiveDirectly would ex­ceed $20 mil­lion. For sim­plic­ity of anal­y­sis, I (like GiveWell) ig­nore this money. I note that mak­ing this at­tri­bu­tion would make two of the four 50% es­ti­mates wrong, ar­guably mak­ing me perfectly cal­ibrated (how­ever, the fact that both cases where I was wrong, I erred on the side of un­der­es­ti­mat­ing, sig­nals im­perfect cal­ibra­tion). On the other hand, it would push the to­tal money moved to the 99th per­centile of my dis­tri­bu­tion.

I had also made pre­dic­tions both re­gard­ing the num­ber of donors and the amount of money moved per donor:

We see that 2013 was a sig­nifi­cant in­crease year for num­ber of donors, whereas 2014 was an year when the num­ber of donors didn’t in­crease much (there was a lot of churn) but the av­er­age money moved per donor in­creased sig­nifi­cantly. I be­lieve that the situ­a­tion for 2015 will look pos­i­tive on both fronts: an in­crease in the num­ber of donors, and an in­crease in the money moved per donor. My me­dian es­ti­mates for the growth on both fronts is around 20%. Ex­plic­itly, my 50% in­ter­val for the num­ber of donors in 2015 is 9,500-12,500. My 50% in­ter­val es­ti­mate for the money moved per donor is $1,500-2,000.

In both cases, the ac­tual value ex­ceeded my 50% in­ter­val. The ac­tual num­ber of donors in 2015 was 14,287, and the to­tal money moved was $39.73 mil­lion, or a to­tal of about $2,780 per donor. The former es­ti­mate was off due to a sig­nifi­cant in­crease in the num­ber of donors at the low end of the range. The lat­ter es­ti­mate was off be­cause dona­tions of over a mil­lion dol­lars in­creased sig­nifi­cantly. In fact, 2015 was the first year where over half the to­tal money moved was ex­plained by dona­tions of over a mil­lion dol­lars. Here’s the table from the GiveWell blog post.

GiveWell money moved by donor size (2015)

In fact, if all the peo­ple who donated $100,000 or more had donated just $100,000 each, the to­tal amount of money moved would have been at the lower end­point of my 50% in­ter­val.

GiveWell’s full met­rics re­port (PDF, 16 pages) con­tains con­sid­er­ably more de­tail on their money moved. I’ll dis­cuss that more later in this post, but none of those ad­di­tional met­rics were part of my quan­ti­ta­tive fore­casts.

There are two ways of pars­ing the se­vere er­ror in my es­ti­mate of money moved:

  • I un­der­es­ti­mated the pop­u­lar­ity of GiveDirectly with in­di­vi­d­ual donors.

  • I un­der­es­ti­mated the sheer num­ber of peo­ple (8 in to­tal) who would donate $1 mil­lion or more.

Although GiveWell’s re­port does not break down the char­ity dis­tri­bu­tion for $1 mil­lion+ donors, I ex­pect that a sig­nifi­cant share of the $21.3 mil­lion moved from $1 mil­lion+ donors went to GiveDirectly. How­ever, I’ll ex­plore the two as­pects sep­a­rately first, then com­bine them and spec­u­late based on that.

The multi-mil­lion donors: who could they be?

What does it take to be able to donate over a mil­lion dol­lars in a sin­gle year? What kind of net worth or an­nual in­come are we talk­ing about? GiveWell doesn’t dis­cuss $1 mil­lion+ donors in depth, prob­a­bly to pro­tect their pri­vacy. But we can in­fer a few things just with ar­ith­metic. We can de­mar­cate two ex­tremes:

  • Stable $1 mil­lion+/​year donors (mostly ul­tra high net-worth in­di­vi­d­u­als): Th­ese are peo­ple with over $1 mil­lion in an­nual in­come (in­clud­ing in­ter­est in­come). For those rely­ing pri­mar­ily on in­ter­est in­come, this would im­ply at least $20 mil­lion in net worth, or the pos­ses­sion of some rapidly ap­pre­ci­at­ing as­sets. In other words, sta­ble $1 mil­lion+/​year donors aren’t just mil­lion­aires, they are at least one or­der of mag­ni­tude more than mil­lion­aires. More­over, they have at least $1 mil­lion to give off in liquid as­sets. So we’re not talk­ing of purely pa­per mil­lion­aires such as early em­ploy­ees at a com­pany like Uber. But we might be talk­ing of ven­ture cap­i­tal­ists, peo­ple work­ing in fi­nance, top CEOs, or peo­ple who were early em­ploy­ees or in­vestors in a com­pany that has IPO’d and who were able to cash out the stock. Most of these peo­ple would prob­a­bly qual­ify as ul­tra high-net worth in­di­vi­d­u­als, the cut­off for which is $30 mil­lion in as­sets ex­clud­ing one’s pri­mary res­i­dence.

  • One-off $1 mil­lion+/​year donors (mostly high net-worth but not ul­tra high net-worth in­di­vi­d­u­als): Th­ese are peo­ple who are mak­ing a one-time dona­tion of that amount. The net worth of peo­ple here only needs to be mod­er­ately over $1 mil­lion for such a dona­tion to be the­o­ret­i­cally pos­si­ble. For peo­ple who fall in this cat­e­gory and do not fall in the pre­vi­ous cat­e­gory, they must have ei­ther in­her­ited money or been in the work­force long enough to save up enough money, or hit it big with some spe­cific in­vest­ments (whether as em­ploy­ees vest­ing shares or by buy­ing fi­nan­cial as­sets or real es­tate). Th­ese peo­ple would qual­ify as high net-worth in­di­vi­d­u­als but not meet the thresh­old for be­ing ul­tra high net-worth (with the ex­cep­tion of peo­ple with fi­nan­cial as­sets that are cur­rently locked down but po­ten­tially worth a lot).

Where are these peo­ple? One thing I re­al­ized is that I know al­most none of these peo­ple per­son­ally, and even the ones in this list I am pos­si­bly in­di­rectly ac­quainted with (friends of friends) I rarely talk to. It also seems like these peo­ple do not post reg­u­larly on var­i­ous on­line fora, whether it’s the GiveWell com­ments sec­tion or Face­book groups. On the other hand, it’s pos­si­ble that they do, and just don’t let on that they have so much money. Either way, since the to­tal num­ber of such peo­ple is pretty small, we can’t re­ally draw rigor­ous statis­ti­cal con­clu­sions from whether any one of them hap­pens to have been spot­ted. The point is that most of the meth­ods I used to iden­tify a trend won’t work on this pop­u­la­tion be­cause it is:

  • quan­ti­ta­tively small enough to be a round­ing er­ror in any kind of statis­ti­cal com­pila­tion (the to­tal num­ber of ul­tra high net-worth in­di­vi­d­u­als in the whole world is around 100,000),

  • out­side my friend and ac­quain­tance cir­cle, and

  • likely to keep a low pro­file on pub­lic fora.

Thus, my in­abil­ity to pre­dict an ex­plo­sion in dona­tions by this group in 2015, amidst var­i­ous other am­biva­lent sig­nals (such as flatlin­ing or de­clin­ing in­ter­est in GiveWell on other di­men­sions) is ex­cus­able.

That still doesn’t ex­plain why there were so many donors in this group, and why this amount in­creased sig­nifi­cantly in 2015.

1.2. What’s with GiveDirectly? And more hy­pothe­ses on UHNWIs

A lot of dona­tion money went to GiveDirectly, de­spite the fact that the char­ity that GiveWell pushed at year-end was, pretty un­am­bigu­ously, the Against Malaria Foun­da­tion.

  • GiveWell ex­plic­itly sug­gested to donors to donate all their money at the mar­gin to the Against Malaria Foun­da­tion.

  • The 2015 post sum­ma­riz­ing staff dona­tions sug­gested a strong prefer­ence for the Against Malaria Foun­da­tion rel­a­tive to other GiveWell-recom­mended char­i­ties. Holden and Elie, the two GiveWell co-founders and co-ex­ec­u­tive di­rec­tors, donated 100% to AMF. None of the staff al­lo­cated money pri­mar­ily to GiveDirectly; the ones who did donate to GiveDirectly usu­ally donated 5% or 10%. This was in sharp con­trast with the 2013 staff dona­tions post. In 2013, some staff mem­bers had al­lo­cated all their money to GiveDirectly.

  • One of the hand­ful of posts GiveWell pub­lished in De­cem­ber 2015 (the peak of giv­ing sea­son) pro­vided sug­ges­tions on how to do a fundraiser for AMF. There was no similar post for GiveWell’s other top char­i­ties. This was also a rel­a­tively un­usual post com­ing from GiveWell. They haven’t done a similar post for years; the most re­cent similar posts I could find were from 2010, back when GiveWell recom­mended VillageReach.

  • Giv­ing What We Can, one of the or­ga­ni­za­tions in­volved in recom­mend­ing char­i­ties (us­ing GiveWell’s recom­men­da­tions as an in­put) and rais­ing funds for them, does not list GiveDirectly in its top char­i­ties. Its 2015 top char­i­ties blog post only men­tioned GiveDirectly in pass­ing as a po­ten­tial op­tion for peo­ple con­cerned about long-run room for more fund­ing.

Thus, I ex­pect that a lot of the GiveDirectly-fo­cused donors were not GiveWell reg­u­lars. I also sus­pect, as noted above, that many of these donors were from the $1 mil­lion+ group. Part of this is the ar­ith­metic of how dona­tions shook out. The other part is that I’m not aware of any “mass” move­ment in fa­vor of GiveDirectly, and so the best ex­pla­na­tion for how it raised so much money would be if the money came from a small group of peo­ple who are some­what out of the pub­lic eye.

So where did these folks come from, and how did they get to donate? I see two pos­si­ble ex­pla­na­tions:

  • De­mon­stra­tion effects: Ul­tra high net-worth in­di­vi­d­u­als (UHNWIs) are used to sales pitches from peo­ple hun­gry for their money. So, more so than oth­ers, they take cues from what other UHNWIs are do­ing to filter out the range of pos­si­ble ac­tions. Good Ven­tures’ Au­gust 2015 dona­tion of $25 mil­lion to GiveDirectly may have had a se­ri­ous im­pact on other po­ten­tial donors. See­ing some­body ac­tu­ally put down that much money le­gi­t­imizes a dona­tion op­por­tu­nity. Note that al­though GV made grants to other GiveWell-recom­mended char­i­ties in 2015, in to­tal GD still re­ceived way more money than any other top char­ity. More­over, it may take a UHNWI a sig­nifi­cant amount of time to de­cide to donate $1 mil­lion+, so the fact that the GD dona­tion oc­curred in Au­gust may have meant am­ple time to get through the pro­cess.

  • Greater ap­peal of the GiveDirectly model to UHNWIs: Per­haps due to its scal­a­bil­ity or its nov­elty, or its con­cep­tual sim­plic­ity and ro­bust­ness to many differ­ent as­sump­tion sets, and its mes­sage of fi­nan­cial em­pow­er­ment, GiveDirectly may res­onate with big-money folks more than top­ics in global health (a good prima fa­cie set of rea­sons was ar­tic­u­lated by Holden in a 2012 blog post). It wouldn’t be sur­pris­ing if peo­ple who have man­aged to cash in on suc­cess more (i.e., their suc­cess has been rec­og­nized in the form of money) would have a more pos­i­tive view of fi­nan­cial em­pow­er­ment and the abil­ity of more money to solve prob­lems. Also, as men­tioned above, UHNWIs may be par­tic­u­larly skep­ti­cal of weak ev­i­dence bases. With limited time to perform eval­u­a­tion but also a lot more re­spon­si­bil­ity to spend money wisely (be­cause of the larger sums in­volved) they may grav­i­tate to in­ter­ven­tions that have es­sen­tially un­limited room for more fund­ing and are ex­tremely ro­bust with min­i­mal need for com­plex chains of rea­son­ing. I should note that I have en­coun­tered this mind­set in a few (non-UHNWI) peo­ple, but I don’t have di­rect ev­i­dence that UHNWIs are much more prone to this mind­set. So there is a bit of spec­u­la­tion here, but it is plau­si­ble.

One in­di­ca­tor of the spe­cific in­ter­est of UHNWIs in GiveDirectly is found in the GiveWell blog post dis­cussing how it was Good Ven­tures’ in­sis­tence that led to the $25 mil­lion grant to GiveDirectly:

Our ex­pe­riences with GiveDirectly over the last few months demon­strate some po­ten­tial benefits of some­times be­ing an ac­tive fun­der. In this case, it was only af­ter Good Ven­tures en­couraged us to ask our top char­i­ties about how they could use ad­di­tional large amounts of fund­ing that we had ex­ten­sive dis­cus­sions with GiveDirectly about pos­si­ble growth plans. Our im­pres­sion is that GiveDirectly’s vi­sion for how it could use ad­di­tional funds to grow as an or­ga­ni­za­tion sharp­ened over the course of these dis­cus­sions as it be­came clear that Good Ven­tures was in­ter­ested in mak­ing a sub­stan­tial grant.

Ul­ti­mately, these dis­cus­sions re­vealed po­ten­tially high-im­pact giv­ing op­por­tu­ni­ties that we would not have been aware of if we had not pur­sued the ques­tion of how GiveDirectly could use ad­di­tional fund­ing to ac­cel­er­ate its growth rel­a­tively ac­tively. We had pre­vi­ous dis­cus­sions with GiveDirectly’s lead­er­ship where we ex­plic­itly asked them a) whether there was any­thing else we could do to sup­port their growth and b) whether it would po­ten­tially be valuable to provide ad­di­tional fund­ing for their fundrais­ing team. In both cases, GiveDirectly an­swered that it did not see promis­ing op­por­tu­ni­ties for ad­di­tional fund­ing. Our im­pres­sion is that our per­sis­tent en­courage­ment to think more deeply about how it would use sig­nifi­cantly more fund­ing, cou­pled with a strong in­ter­est from Good Ven­tures in pro­vid­ing such fund­ing, mo­ti­vated GiveDirectly to re­flect on its needs and make the fund­ing re­quest de­tailed in this post. (Like­wise, we would not have pur­sued these con­ver­sa­tions with GiveDirectly as ac­tively in the ab­sence of Good Ven­tures’ per­sis­tent en­courage­ment to do so.)

Notably, how­ever, Good Ven­tures also prompted GiveWell to con­duct similar in­ves­ti­ga­tions into AMF, so it’s not as if they were solely in­ter­ested in GiveDirectly.

UPDATE: Part of a com­ment by Cather­ine Hol­lan­der on the GiveWell met­rics blog post is roughly in line with what I wrote above. It also em­pha­sizes some­thing I had in my mind when writ­ing the post but ne­glected to men­tion: word of mouth from past big donors played an im­por­tant role. This makes sense: UHNWIs are likely to listen to their friends who are also UHNWIs and have demon­strated their se­ri­ous­ness by pre­vi­ously donat­ing to GiveWell-recom­mended char­i­ties. Even a cou­ple of UHNWIs who had donated in the past can use their net­work of friends to get their friends in­ter­ested in GiveWell. As an ex­am­ple, word of mouth from Dustin Moskovitz (whose money funds Good Ven­tures) could have led a few of the early Face­book em­ploy­ees and in­vestors (who are friends with him) to donate to GiveWell-recom­mended char­i­ties.

That said, for GiveWell’s very largest donors, we be­lieve the in­crease in dona­tions was due to our long-term ap­proach of build­ing a high-qual­ity re­search product and rep­u­ta­tion rather than any short-term changes to our donor out­reach strat­egy. In al­most all cases, the ma­jor donors we saw in 2015 had heard about us from an­other in­di­vi­d­ual who spent a lot of time talk­ing with us over the years, or had heard of us prior to 2015. Our im­pres­sion is that the in­crease in ma­jor dona­tions in 2015 was a re­flec­tion of our long-term ap­proach, rather than a new out­reach strat­egy.

1.3. In­cor­rect pre­dic­tion of over­all donor growth, and a deeper dive into dis­cov­ery, con­ver­sion, and re­ten­tion/​upsell

The other di­rec­tion in which my over­all model was off was that I failed to pre­dict the over­all in­crease in the num­ber of donors. This isn’t ex­pli­ca­ble in the ways out­lined above. The over­all num­ber of donors is some­thing that the var­i­ous statis­tics I had col­lected should have been able to in­form with rea­son­able rigor. I think my failure here was mostly just quan­ti­ta­tive and not qual­i­ta­tive.

In my ear­lier post, I iden­ti­fied three challenges for mov­ing money:

  • Dis­cov­ery and brand awareness

  • Conversion

  • Re­ten­tion and upsell

I looked at a num­ber of met­rics, rang­ing from Wikipe­dia pageviews to search in­ter­est, and con­cluded that in­ter­est in GiveWell was ei­ther flatlin­ing or grow­ing mod­estly. I didn’t see a surge in in­ter­est. Also, based on a num­ber of sur­veys I con­ducted, I be­lieved that de­spite in­creas­ing brand aware­ness, most peo­ple weren’t con­vinced of the value of donat­ing to GiveWell. I con­cluded that con­ver­sions would there­fore rise only mod­estly. I wasn’t broadly off (99%+ of donors re­main un­in­ter­ested in GiveWell, just as they used to be) but my in­stincts weren’t fine-tuned enough to pre­dict just how many ex­cep­tions there would be.

Let’s take a some­what closer look at the long-term donor re­ten­tion figures re­leased by GiveWell in its full met­rics re­port (PDF, 16 pages). On pages 8 and 9, they in­clude three ta­bles on donor re­ten­tion. Here are the find­ings in the first table:

  • 5,686 donors gave in 2014 and not in 2015, with the amount be­ing about $4.097 mil­lion.

  • 3,690 donors gave in 2014 and 2015, with the amount donated grow­ing from $11.881 mil­lion to $16.057 mil­lion.

  • 11,584 donors gave in 2015 and not in 2014, with the amount donated be­ing $27.545 mil­lion.

I had pre­vi­ously ex­pected most of GiveWell’s growth to come from re­ten­tion and up­sell, but the truth was differ­ent. For those who donated in 2014, the to­tal amount they donated in 2015 was only a lit­tle bit higher than be­fore: an in­crease from $15.979 mil­lion to $16.057 mil­lion, or about 5%. More­over, new donors ac­counted for well over half the dona­tion amount.

The failure to pre­dict the sig­nifi­cant amount of money moved by new donors re­lates to the failure I doc­u­mented ear­lier, namely, a failure to iden­tify big dona­tion amounts and money moved to GiveDirectly. But there were two other er­rors I made:

  • I over­pre­dicted the mag­ni­tude of re­ten­tion and up­sell. I had not made an ac­tual nu­mer­i­cal es­ti­mate, but I be­lieve that had I been asked, I would have pre­dicted more sig­nifi­cant growth than 5% from the 2014 donor set.

  • I un­der­pre­dicted the sheer num­ber of new donors, es­pe­cially the ones at the low­est end of dona­tion amounts (just as I un­der­pre­dicted dona­tions at the high­est end). This was partly based on my in­spec­tion of the web met­rics. I saw a flatlin­ing of pageviews of GiveWell’s Wikipe­dia page, and con­cluded that in­ter­est in GiveWell from new donors was limited. I ex­pected that to the ex­tent that GiveWell would see growth, it would be from a com­bi­na­tion of con­vert­ing peo­ple who had already been in­ter­ested in the past to fi­nally donate, and up­sel­ling past donors to donate more, par­tic­u­larly with the in­creased im­por­tance given to AMF. it is pos­si­ble that I was right, and that growth stemmed largely from con­ver­sion, rather than dis­cov­ery and up­sell. But I sus­pect that I sim­ply un­der­es­ti­mated the role of dis­cov­ery, and that there may have been some other mys­te­ri­ous fac­tors be­hind the lack of growth in in­ter­est in GiveWell’s Wikipe­dia page (for in­stance, a less skep­ti­cal set of new donors who don’t at­tempt that hard to cross-check or­ga­ni­za­tions across mul­ti­ple sources).

UPDATE: Cather­ine Hol­lan­der, in a com­ment on the GiveWell met­rics blog post, high­lights the role played by me­dia pub­lic­ity to at­tract­ing new, small donors.

We’d guess that sub­stan­tial me­dia and gen­eral at­ten­tion to GiveWell’s work prob­a­bly drove the in­crease we saw in smaller donors. We’ve seen this in past years where we had rel­a­tively strong me­dia cov­er­age, and in 2015, GiveWell re­ceived a sig­nifi­cant amount of cov­er­age, in­clud­ing in new books such as William MacAskill’s Do­ing Good Bet­ter, Peter Singer’s The Most Good You Can Do, and Larissa MacFar­quhar’s Strangers Drown­ing.

1.4. Com­par­i­son with 2014 data, and re­la­tion with the 2015 Effec­tive Altru­ism Sur­vey (added July 30, 2016)

On July 29, 2016, Chris Cundy pub­lished a post link­ing to a PDF of the 2015 Effec­tive Altru­ism Sur­vey Re­sults. This sur­vey cov­ers dona­tion in­for­ma­tion for 2014, mostly for self-iden­ti­fied effec­tive al­tru­ists. The rea­son the 2015 sur­vey con­tains 2014 dona­tion data is, as the PDF ex­plains:

We asked about 2014 dona­tions, as the sur­vey started be­fore the end of 2015. Similarly, last year [refer­ring to the 2014 EA Sur­vey Re­sults] we asked about 2013 dona­tions.

I also looked at the ac­tual sur­vey ques­tions to ver­ify that both the ques­tions on dona­tions (amount donated, and char­i­ties donated to) ex­plic­itly re­stricted to 2014. The two ques­tions, ex­plic­itly, were:

  • Over 2014 roughly how much did you donate?

  • Over 2014, which char­i­ties did you donate to?

The post pro­vides in­ter­est­ing ad­di­tional in­for­ma­tion in a few ways:

  • It sam­ples from a some­what differ­ent pop­u­la­tion, namely, mostly self-iden­ti­fied effec­tive al­tru­ists and oth­ers who have come across the term or idea. With that said, I think that the set of peo­ple who took the sur­vey and also donate to GiveWell char­i­ties are not that differ­ent from typ­i­cal GiveWell donor.

  • It has a num­ber of other ques­tions on peo­ple’s be­hav­ior and pri­ori­ties (for in­stance, whether they eat meat, and the rel­a­tive pri­ori­ties of var­i­ous cause ar­eas), that pro­vides more con­text to their dona­tion choices.

  • It in­cludes the ac­tual num­ber of donors to each char­ity (in­clud­ing GiveWell-recom­mended char­i­ties, and oth­ers that have some name recog­ni­tion in the EA com­mu­nity). This is in con­trast to GiveWell’s re­port, which in­cludes the to­tal money moved but not the num­ber of donors by char­ity. The EA Sur­vey not re­port­ing the to­tal money moved by char­ity is partly a re­sult of the way the sur­vey was de­signed: it asked a ques­tion about how much money was donated in 2014 in to­tal, and an­other ques­tion about the char­i­ties donated to, but did not ask about the amount donated per char­ity. In par­tic­u­lar, for peo­ple who donated to more than one char­ity, the sur­vey data won’t be enough to figure out how much went to each char­ity.

Since the year is off by one, how­ever, we’ll need to also pull up the 2014 money moved num­bers by char­ity for GiveWell. Here’s the table from the 2014 money moved post:

And for good mea­sure, here is a table from the Wikipe­dia page of money moved ev­ery year to the top char­i­ties from donors ex­clud­ing Good Ven­tures:

Or­ga­ni­za­tion Money moved in year 2011[73] Money moved in year 2012[74] Money moved in year 2013[75] Money moved in year 2014[76] Money moved in 2015[10][77] To­tal
Against Malaria Foun­da­tion 1,810,237 4,579,514 2,490,588 4,434,478 15,445,609 28,760,426
GiveDirectly 86,146 729,359 3,482,865 4,061,487 19,364,385 27,724,242
Schis­to­so­mi­a­sis Con­trol Ini­ti­a­tive 510,480 861,548 1,440,184 3,340,403 2,657,389 8,810,004
De­worm the World Ini­ti­a­tive -- -- 642,836 878,044 1,080,068 2,601,348

As you can see, in 2014, the to­tal money moved to AMF, GiveDirectly, and SCI was all in the same bal­l­park, whereas DtWI got much less money. A key fact for con­text around AMF: GiveWell re­turned AMF to the list of top char­i­ties at the end of 2014, af­ter tem­porar­ily tak­ing it off the list due to con­cerns around room for more fund­ing. There­fore, for most of the year, AMF wasn’t ac­tu­ally in the top list. That prob­a­bly ex­plains why it did not raise that much more money than the oth­ers.

It is there­fore par­tic­u­larly in­ter­est­ing that the num­ber of peo­ple in the sur­vey who donated to AMF was about 400, com­pared to 250 for SCI and 200 for GiveDirectly (see pages 9 and 10 of the sur­vey PDF). How did the char­i­ties raise such similar amounts of money, de­spite the num­ber of donors be­ing so differ­ent? It seems that even as far back as 2014, the dona­tion amounts, both as ob­served in the EA Sur­vey Re­sults and as mea­sured by GiveWell, were heav­ily skewed in the di­rec­tion of big donors. My just-so story for this is as fol­lows: most peo­ple donate at the end of the year, based on GiveWell’s lat­est re­search as re­leased at the end of the year. So, most GiveWell donors ended up donat­ing to AMF be­cause it was marked as a top recom­men­da­tion by GiveWell at year-end. Big­ger donors, on the other hand, have longer de­ci­sion pro­cesses. They may con­duct ad­di­tional vet­ting of char­i­ties or en­gage in dis­cus­sions with char­ity rep­re­sen­ta­tives. Since AMF was not on the top list for most of the year, they were more evenly dis­tributed be­tween the three char­i­ties, lead­ing to an ul­ti­mately more even split

Note that this is to­tally con­sis­tent with the fact that in 2015, AMF raised a lot more money than SCI. In 2015, AMF was in the top spot through­out the year, giv­ing big­ger donors enough time to se­lect, vet, and ar­range money for AMF.

1.5. Some qual­i­ta­tive pre­dic­tions of the tra­jec­tory for 2016

The sig­nifi­cant amount of money moved that was in­fluenced by GiveWell recom­men­da­tions puts GiveWell in new ter­ri­tory. In 2015, GiveWell hit nine-digit lev­els of money moved ($110.1 mil­lion, in­clud­ing the money donated by Good Ven­tures).

One main down­ward trend I ex­pect to see for 2016 is a smaller to­tal amount of money granted by Good Ven­tures to GiveWell top picks:

  • The $25 mil­lion Good Ven­tures grant to GiveDirectly (made in Au­gust 2015) was a three-year grant, with about $6-9 mil­lion of it in­tended to im­prove GiveDirectly’ fundrais­ing op­er­a­tion. Given GiveDirectly’s ap­par­ent suc­cess at rais­ing more funds, it seems like the grant did its work. I there­fore don’t ex­pect any similar mid-year grant to GiveDirectly, though I ex­pect that Good Ven­tures will still make its usual end-of-year grant (on the or­der of $10 mil­lion) to sup­port GiveDirectly’s work.

  • Similarly, I ex­pect that Good Ven­tures will con­tinue to make grants of similar mag­ni­tudes to those made this year, ex­cept to those char­i­ties that seem to have had trou­ble putting to good use the money given so far. And it’s quite likely that at least one of AMF, SCI, and DtWI will be deemed to have not yet suc­cess­fully used funds given so far.

The up­shot is that I ex­pect Good Ven­tures money to GiveWell top picks to not be more than $50 mil­lion this year (com­pared to $44.4 mil­lion to top picks ex­clud­ing stand­out char­i­ties and ex­clud­ing the one-off $25 mil­lion GiveDirectly grant).

If my the­ory re­gard­ing the mo­ti­va­tion of UHNWIs is cor­rect, then this dropoff in Good Ven­tures money pushed is likely to lead to a cor­re­spond­ing drop in UHNWI in­ter­est in GiveDirectly. On the other hand, since GiveDirectly’s fundrais­ing op­er­a­tions are still scal­ing up (us­ing the money from Good Ven­tures) there might be growth on that front. Re­cently, GiveDirectly launched a $10 mil­lion fundrais­ing effort to get match­ing funds for a long-term study of ba­sic in­come. Coin­ci­den­tally or not, there has been a surge of in­ter­est in ba­sic in­come among the in­tel­li­gentsia of late, par­tic­u­larly in the United States (full dis­clo­sure: I per­son­ally am not in fa­vor of ba­sic in­come in gen­eral, but GiveDirectly’s ex­per­i­ment might be valuable given the small amounts in­volved). GiveDirectly was there­fore able to get some me­dia cov­er­age of its fundrais­ing ini­ti­a­tive.

While money raised through these fundrais­ing efforts won’t be (for the most part) at­tributed to GiveWell, it’s pos­si­ble that some peo­ple will hear of GiveDirectly through fundrais­ing efforts like this, then sub­se­quently in­ves­ti­gate it and chance upon the GiveWell re­view, and later on donate based on the GiveWell re­view. This might lead to an ap­par­ent in­crease in the money moved by GiveWell, de­spite GiveWell not ini­ti­at­ing the move­ment of money but serv­ing more as a vet­ting and le­gi­t­imiz­ing mechanism.

Over­all, I ex­pect 2016 to largely be a year of stag­na­tion and slow­down in growth. In par­tic­u­lar, look­ing again at the three stages, here are my thoughts for 2016:

  • Dis­cov­ery and brand aware­ness: I be­lieve that brand aware­ness of GiveWell will flatline through all of 2016, and po­ten­tially fall. I know that this looks like I’m once again un­der­es­ti­mat­ing GiveWell’s growth pre­dic­tion, af­ter hav­ing done so in 2015. My take is that the sig­nifi­cant de­cline in its Wikipe­dia pageviews, the paucity of new ma­te­rial pub­lished to the GiveWell blog, and the lack of ex­ter­nal me­dia cov­er­age of GiveWell in 2016 (some­thing that differ­en­ti­ates it from 2015, when there was a lot of me­dia cov­er­age of GiveWell in the wake of Peter Singer’s and William MacAskill’s books) will lead to it not be­ing widely dis­cov­ered. How­ever, events later in the year might change the situ­a­tion.

  • Con­ver­sion: I be­lieve that GiveWell will have more difficulty con­vert­ing peo­ple this year than in pre­vi­ous years. Ex­clud­ing the UHNWI/​GiveDirectly phe­nomenon, I be­lieve its re­views of its top char­i­ties will point to less room for more fund­ing, and some char­i­ties may even (tem­porar­ily) exit the list. I ex­pect new re­cruits won’t re­ally find it that com­pel­ling or con­vinc­ing.

  • Re­ten­tion and up­sell: I ex­pect to­tal money moved in 2016 by 2015 donors to grow by less than 5%, and to pos­si­bly fall.

Although my me­dian es­ti­mate for all three of these as­pects is near-zero, I ex­pect to be wrong about at least one of the three, and I ex­pect that, once that er­ror is ac­counted for, there will be de­cent growth (some­where be­tween 10% and 70%) in to­tal dona­tions from the sub-$100,000 and the $100,000-999,999 groups. As noted ear­lier in the post, one of the er­rors I made last year was in tac­itly equat­ing the sum of the me­di­ans with the me­dian of the sums. This time, I’m not do­ing that. I’m ex­plic­itly ac­knowl­edg­ing that de­spite stag­na­tion be­ing the me­dian sce­nario for each in­di­vi­d­ual case, over­all growth is likely to be pos­i­tive be­cause of the long right tails for each case (i.e., more room on the right than on the left)

Another rele­vant fac­tor to con­sider is the change in GiveWell’s recom­men­da­tion re­fresh fre­quency from an­nual to twice a year (June and Novem­ber). It will be in­ter­est­ing to see if they are able to push that change start­ing with 2016, and the effect that has on dona­tion pat­terns.

I will re­turn with more quan­ti­ta­tive es­ti­mates af­ter more data is available on the dona­tion trends and other met­rics of in­ter­est for 2016.


Im­pli­ca­tions for so­phis­ti­cated small (sub-$100,000) in­di­vi­d­ual donors

Back when GiveWell’s to­tal amount of money moved was small, there were many rea­sons to donate. Most im­por­tant was to set a norm in fa­vor of ev­i­dence-based dona­tion, and to en­courage the growth of GiveWell-recom­mended char­i­ties and the cred­i­bil­ity of GiveWell. In par­tic­u­lar, do­ing so made it more likely that char­i­ties would ac­tu­ally en­gage with GiveWell’s vet­ting pro­cess, lead­ing to greater scope for find­ing bet­ter places to donate money to. I be­lieve none of these rea­sons ap­ply with any­thing like the same force to­day (un­less you’re a UHNWI plunk­ing sev­eral mil­lion dol­lars).

There­fore, I be­lieve that in or­der to be con­vinced to donate to a GiveWell top char­ity, you should be con­vinced of both these things:

  • The char­ity is ob­jec­tively a very good use of marginal funds (set­ting aside any norm-set­ting or demon­stra­tion effects).

  • You ei­ther have a very good sense of how much money the char­ity would raise with­out you, or the case for marginal dona­tions to the char­ity is fairly ro­bust across the range of pos­si­ble val­ues of money moved by the char­ity. Note that if you be­lieve GiveWell’s quan­ti­ta­tive es­ti­mates of fund­ing gaps, and be­lieve that un­til a char­ity meets 100% of its Ex­e­cu­tion Level 1 fund­ing gap, the case for it doesn’t re­ally hit diminish­ing re­turns, then 2015 leaves you in good shape: none of the or­ga­ni­za­tions with a nonzero Ex­e­cu­tion Level 1 fund­ing gap closed it. Even GiveDirectly only got to clos­ing 78% of its fund­ing gap. On the other hand, if you are skep­ti­cal of GiveWell’s es­ti­mated fund­ing gaps, then this skep­ti­cism should lead you to re­think whether and how much to donate to GiveWell top picks.

One in­ter­est­ing thing to watch out for is how GiveWell’s 2016 re­views re­late to its 2015 es­ti­mates for fund­ing gaps. For or­ga­ni­za­tions that did not come close to meet­ing the Ex­e­cu­tion Level 1 fund­ing gap, were they able to put the money to use effec­tively? If not, then it is likely that GiveWell over­stated the fund­ing gap. If they did, then GiveWell circa 2016 would have vin­di­cated GiveWell circa 2015 (which doesn’t nec­es­sar­ily mean it’s right, but it at least means that the pas­sage of a year did not re­veal a glar­ing er­ror in the pre­vi­ous es­ti­ma­tion ex­er­cise).

Im­pli­ca­tions for peo­ple in­ter­ested in con­tribut­ing in non-mon­e­tary ways

I think there are three po­ten­tial les­sons here:

  • As­sum­ing you think money moved to GiveWell top picks is bet­ter al­lo­cated than coun­ter­fac­tual uses, spend­ing more effort on fundrais­ing for top picks might be worth­while. How­ever, cur­rent fundrais­ing strate­gies are miss­ing out on much of the money moved. For in­stance, most (pub­lic enough that I’m aware of them) fundraisers fo­cused on AMF, but more money was moved to GiveDirectly. This could also sug­gest that so­phis­ti­cated fundrais­ing efforts would re­quire peo­ple who have a bet­ter un­der­stand­ing of the way UHNWIs op­er­ate, and bet­ter con­nec­tions with them. I don’t in­tend to say cur­rent fundrais­ing efforts were not cost-effec­tive (they still ap­pear to have raised more money than spent on them) but rather that there ap­pear to be un­tapped seg­ments and poorly un­der­stood dona­tion dy­nam­ics that could be lev­er­aged more effec­tively. It is, how­ever, pos­si­ble that more se­cre­tive efforts to tap UHNWIs have already been suc­cess­ful and were the driv­ing fac­tor for the 2015 in­crease.

  • The in­creased level of money moved sug­gests that in­creased third-party scrutiny of GiveWell top picks, as well as the se­lec­tion pro­cess they used to ar­rive at those picks, may be more jus­tified than when GiveWell was mov­ing less than 10 mil­lion dol­lars a year. I’d like to see more peo­ple take in­de­pen­dent, fresh, hard looks at the prob­lems GiveWell is try­ing to solve. I’m not just talk­ing of re-vet­ting GiveWell top picks, but of go­ing through a similar pro­cess and iden­ti­fy­ing key points of di­ver­gence with them. Per­son­ally, I feel like this sort of fresh look is one of the things that has been most con­spicu­ously ab­sent in re­cent years, and it’s some­thing that I hope to en­courage more of.

  • There is also more of a case for spend­ing sig­nifi­cantly more re­sources on ex­plor­ing new char­ity op­por­tu­ni­ties, both to get a GiveWell recom­men­da­tion and to po­ten­tially com­pete for what ap­pears to be a larger pool of money that peo­ple are will­ing to spend on ev­i­dence-based in­ter­ven­tions in global health and global poverty. Char­ity En­trepreneur­ship is one of a hand­ful of groups try­ing to do that. In ad­di­tion, a case might be made for more fund­ing of efforts like In­no­va­tions for Poverty Ac­tion or the MIT Poverty Ac­tion Lab.

Im­pli­ca­tions for GiveWell as an or­ga­ni­za­tion, and the money it’s spending

Prima fa­cie, GiveWell seems to have a huge mul­ti­plier. It es­ti­mates that $1.1 mil­lion was spent in 2015 sup­port­ing GiveWell’s core op­er­a­tions (the re­main­ing money was spent on the Open Philan­thropy Pro­ject, which is in the pro­cess of be­com­ing a sep­a­rate or­ga­ni­za­tion). The or­ga­ni­za­tion moved 100 times that much money. With that huge a mul­ti­plier, it seems al­most pointless to quib­ble with whether the speci­fics of how GiveWell was spend­ing money were effi­cient in a nar­row sense. Even if GiveWell spent four times as much, it would still have an im­pres­sive mul­ti­plier on money moved.

This does not defini­tively im­ply that a donor should at the mar­gin donate to GiveWell. It’s not clear that more money in their hands trans­lates to more money moved, or bet­ter qual­ity re­search. Also, GiveWell’s re­vealed prefer­ence (with re­spect to redi­rect­ing re­sources to­ward the Open Philan­thropy Pro­ject) sug­gests that they don’t see a lot of room for more spend­ing on core GiveWell op­er­a­tions. More­over, po­ten­tial donors to GiveWell’s core op­er­a­tions don’t have much lev­er­age over GiveWell’s work: they don’t need your money, since they can always get more from other donors, with Good Ven­tures a good fal­lback op­tion.

The up­shot is that un­less you feel highly al­igned with the ob­ject level re­search or out­reach be­ing done by GiveWell, it’s prob­a­bly not a good idea to donate with them. But it’s also prob­a­bly not worth­while to feel con­cerned about the re­sources they are spend­ing on their op­er­a­tions since they move con­sid­er­ably more money, and they don’t have any short-run plans for sig­nifi­cant ex­pan­sion.

PS: Since I’m time-limit­ing my­self on writ­ing this post, the anal­y­sis isn’t as thor­ough as op­ti­mally pos­si­ble. Nonethe­less, I hope it is in­ter­est­ing to peo­ple.

PS2: GiveWell’s full met­rics re­port (PDF, 16 pages) con­tains many more de­tails, in­clud­ing sources of donors, web traf­fic, and sur­veys of donors that re­veal in­for­ma­tion on how GiveWell in­fluenced their de­ci­sion. I recom­mend read­ing it in its en­tirety. I ex­pect to refer­ence many of these sec­tions in my blog post later this year where I lay out quan­ti­ta­tive pre­dic­tions.

PS3: In re­sponse to some feed­back I re­ceived, I re­al­ized that I should clar­ify that my goal with this post was sim­ply to com­pare my re­sults against re­al­ity, not against al­ter­na­tive fore­casts made by oth­ers. The only al­ter­na­tive fore­cast I have available, that was made pub­lic, was Elie’s board meet­ing fore­cast of be­tween $20 mil­lion and $30 mil­lion in money moved. I had pre­dicted a lower amount of money moved, and ex­plained rea­sons in my origi­nal post, namely, that on the met­rics available to me for q3/​q4 at the time, which didn’t in­clude money moved, year-over-year growth had ap­peared to have gone down. I won’t quote the post all over again, but you can re-read the origi­nal post for ful­ler con­text. I think those rea­sons made sense given what I knew at the time, but I should prob­a­bly have given more cre­dence to a quasi-pub­lic es­ti­mate be­ing put out by GiveWell, in ad­di­tion to all the other broader points I made in this post (namely, the effects of deal­ing with a large num­ber of skewed dis­tri­bu­tions).

It’s im­por­tant to note, though, that even us­ing Elie’s es­ti­mate (or a sim­ple ex­trap­o­la­tion from year-over-year Q2 2015 growth, the most re­cent year-over-year growth num­ber available), would have caused a sig­nifi­cant un­der­es­ti­ma­tion in money moved. Yes, the es­ti­mates would have been a lit­tle bet­ter, and my (wrong in hind­sight) rea­sons for push­ing down rel­a­tive to GiveWell’s ex­ist­ing es­ti­mates would have been re­moved. But the fun­da­men­tal failure to pre­dict the huge amount of money moved by $1 mil­lion+ donors, or the large amount of money moved to GiveDirectly, would have per­sisted.

3. Updates

3.1. High net worth in­di­vi­d­ual outreach

On June 6, GiveWell re­leased au­dio of its March 21, 2016 board meet­ing. Near the end of the meet­ing, GiveWell rep­re­sen­ta­tives Holden Karnofsky and Elie Hassen­feld dis­cussed GiveWell’s out­reach efforts be­ing led by Cather­ine Hol­lan­der and Sean Con­ley. One of the key ar­eas of fo­cus men­tioned was one-on-one fol­lowups with all donors who had given $10,000 or more, a recog­ni­tion by GiveWell of the im­por­tance of high-value donors. They men­tioned that GiveWell had done this a long time ago but had stopped do­ing it con­sis­tently due to a lack of band­width; with the larger em­ployee count, they were able to re­sume. This is con­sis­tent with the March 2016 blog post high­light­ing why they do donor calls (such posts are usu­ally made only dur­ing Giv­ing Sea­son). All of this fits well with the ob­ser­va­tion made through­out this post of the im­por­tance of large donors to shap­ing both the amount and na­ture of GiveWell’s money moved.

3.2. Pre­dic­tions for 2016

On Novem­ber 15, 2016, I en­tered a num­ber of pre­dic­tions in Pre­dic­tionBook about end-of-year char­ity recom­men­da­tions, money moved, etc. You can ac­cess my en­tire set of pre­dic­tions at my user pro­file page, but the list in­cludes many pre­dic­tions not di­rectly re­lated to GiveWell money moved. Below, I list the pre­dic­tions most di­rectly re­lated to the themes of this post.

Pre­dic­tions on money moved (each pre­dic­tion ini­tially with 75% prob­a­bil­ity):

Another re­lated pre­dic­tion, not about the amount of money moved, but rather, about timing, was GiveWell will pub­lish a blog post an­nounc­ing its 2016 end-of-year top char­ity recom­men­da­tions be­fore Thurs­day, Novem­ber 24, 2016 (Thanks­giv­ing Day). My cur­rent prob­a­bil­ity es­ti­mate for this is 70%.

I will edit the prob­a­bil­ities around these pre­dic­tions as new in­for­ma­tion comes in, and in­clude com­ments ex­plain­ing the up­dates.

Dis­clo­sure: I am not fi­nan­cially af­fili­ated with GiveWell, but have some past and on­go­ing as­so­ci­a­tion with them. You can get a de­tailed ac­count of my past in­ter­ac­tion and af­fili­a­tion with GiveWell here.