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. This post uses some estimation I did in a recent post titled GiveWell money moved forecasts and implications.
In this post, I describe a donation commitment I am making. Part of the reason for making this commitment publicly is that I believe the structure of the commitment might be relevant to the donation decisions of others intending to donate to the same recipients. The other reason is that my decision process and the structure of my final commitment may themselves be of interest to others.
I believe that GiveWell's updated top charity list represents one of the best attempts to provide a list of top places to donate to for the conventional, unattached donor. I do have some reservations about the way they have structured their recommendations, and in particular I think that they have missed out on a number of things in their evaluation that are pretty important to me as a donor (this isn't the place to go into them, but you can see my comment on GiveWell's December open thread for a summary of my concerns).
This is probably the first year in my life when I can make a reasonably significant donation to a GiveWell top-recommended charity without it noticeably affecting my finances; previously when I made such donations I could see the effect on my bank accounts for months and even years to come. Despite this, I am not making any upfront donation to a GiveWell top-recommended charity. Part of the reason is that I simply lack the time to investigate these, and I feel that GiveWell's reviews don't address many of the questions that I would like answers to (so I cannot rely on them exclusively). More importantly, however, I expect that even after thoroughly answering the questions I have, I am still quite likely to decide simply to not donate to a GiveWell top charity. Therefore, at this point in my life I don't believe it's rational for me to either spend time investigating things or make a donation based on expectation.
Despite my reservations, I believe that both GiveWell and its top charities represent important and positive developments in their respective areas, and I would like to see them continue to thrive. Fortunately, I believe that both GiveWell and its top charities are in no danger of death, and will probably raise quite a bit of money, more than what I consider necessary for their core survival and ability to progress. But what if I'm wrong? What if they fail to raise enough money? I would then find their funding gap important to fill, and might actually donate. Pondering this more, I thought that it might make sense to structure a donation commitment that is explicitly tied to the amount of money that GiveWell moves from donors other than Good Ventures to its top charities.
The donation commitment
My strategy is to structure my donation commitment so that it is zero for very small and very large amounts of money raised, and is an upside-down quadratic between a lower cutoff and an upper cutoff. Last year, GiveWell's top charities raised somewhere between $12.7 million and $13.0 million (in current United States dollars) for its top charities from donors excluding Good Ventures. Let's say the corresponding figure they report for this year is $X (again, in current United States dollars). [NOTE: "current" here means the literal dollar amount, without making any adjustments for inflation. In practice, this doesn't matter over the short time horizon under consideration. Also, in practice, I'll defer to GiveWell's numbers even if they start doing inflation adjustments]
If they report this number X before April 1, 2016, and X is between 10 million and 15 million, denote:
Y = (X - 10,000,000)(15,000,000 - X)/(1,000,000,000)
Subject to some personal financial conditions (that I don't want to elaborate on here, but that are very unlikely to be binding), if Y > 0, I will donate $Y to GiveWell's current top-rated charity, Against Malaria Foundation, subject to it still being the top-rated charity. I will complete this donation by April 15, 2016. I will call Y my liability. (To be a little more specific, my personal financial conditions will place a cap on the maximum amount of money I can safely donate. In practice, this cap is likely to be much more than the maximum that can ever arise from this formula. However, if I have unforeseen financial problems, then it could be less. If the cap turns out to be smaller than $Y, I'll pay the minimum of the two).
My liability range
My liability is maximized if GiveWell raises exactly in the middle of the 10-15 million range, i.e., at 12.5 million. At this level, my liability computes to $6,250. On the other hand, close to the endpoints of 10 million and 15 million dollars, my liability is closer to zero.
The low liability for small amounts of money moved signifies that at that level, GiveWell seems to have messed up something, and such messing up should not be rewarded too much. At that level, the more the money moved, the more money I will grant. However, beyond $12.5 million (which is roughly the amount raised last year) I start viewing additional growth in money moved as more optional, so I grant less if the money moved increases.
The upper limit of $6,250 is not a trivial amount for me, but it is not an amount that will destroy my finances (note that the personal financial conditions that I haven't elaborated on basically guarantee that in the event that my finances have taken a turn in a way that this amount will significantly hurt me, I'll probably be paying less anyway).
My probability estimate for a liability of zero is 88%. In other words, I believe there is only a 12% chance I'll need to pay out any money at all. Most of the remaining distribution falls in the $15 million+ range. I described my estimates for GiveWell money moved in a previous post. The overall expected value of my outlay, based on my current estimates, is about $300.
Incentive effects on other donors
Most potential donors to GiveWell-recommended charities would be unaware of this blog post! For those who are aware, however, this information does not significantly affect the marginal value of their money.
The margin by which a given donation affects my donation varies from -0.5% to 0.5%, with the +0.5% occurring at a donation level of $10 million, and the -0.5% occurring at a donation level of $15 million. The average of these different marginal effects is exactly zero, which follows mathematically from the fact that at both extremes my outlay is the same (i.e., zero). Moreover, the actual amount that GiveWell will report is uncertain and isn't known by the donor at the time of the donation. So the donor is really relying on a probability distribution over possible amounts of money raised, and must average over the probability distribution. Whatever the distribution, the average cannot be more than 0.5% of the donor's donation in magnitude, and most likely will be a lot less. [Note: I got the 0.5% by doing some calculus; I leave verification as an exercise for the interested reader.] In fact, if the donor estimates only a 12% chance of the money moved even falling in the relevant range, then for every $100 donated, the effect on my donation is at most 6 cents.
The informational effects are more significant, and I turn to those next.
What I'm signaling
By structuring this as a donation commitment (rather than making the donation decision after the fact), I hope to achieve a few goals:
- I'm indicating explicit confidence in GiveWell's work and its top charities, and my support of the work, by showcasing my willingness to donate a relatively nontrivial amount of money to them.
- I am providing clear information about the level of growth that I believe is important for GiveWell and its recipient organizations, below which I would be concerned enough to pitch in.
- By being clear upfront about the conditions, I can plan more clearly for the future fund outlay. It also leaves me with less wiggle room to change the conditions later.
- I am spreading the meme that the amount we donate to charities should be based on how much they'd otherwise raise, and that it's possible to do so in a way that does not create perverse incentives for other donors.
Parting thoughts
I am not suggesting that most donors to GiveWell-recommended charities structure their donations this way. If my goal was to add to the funds of a GiveWell-recommended charity, then donating to it would be the best way forward (as I explained at the beginning of the post, I haven't investigated GiveWell-recommended charities thoroughly enough, didn't find GiveWell's reviews complete enough, and believe that even if I reviewed them I still may not find them compelling). However, if the goal is explicitly to act as an insurance for the organization, making sure that they have enough money to survive and achieve a basic level of growth, then a donation commitment of this sort makes sense.
Also, although the above is an isolated donation commitment, I could imagine making a wide range of donation commitments where the amounts of money committed together add up to a substantial amount. As long as the conditions aren't very strongly correlated, it's highly unlikely that many of them would apply in the same year (this is, sort of, the idea behind insurance).
UPDATE: In a GiveWell blog post comment on December 23, Elie Hassenfeld wrote that GiveWell had, as of December 22, 2015, tracked $15.7 million in donations. Assuming no problems with this estimate, it already puts my obligation for the conditional donation at zero.
The reasoning here is asymmetric: you should also note that if Givewell raises more than $12.5 million, that means they're doing something really well, and should be rewarded. Conversely, if Givewell raises less than $12.5 million, growth is less optional because it will start cutting into more critical operations.
Moreover, it's important to remember that incentives for Givewell recommended charities are not necessarily incentives for Givewell and vice versa.
This is the same for a direct donation, though. Actually it's probably true for a greater extent if one made a concrete donation.
This amount depends on the quantity of money granted from Good Ventures, though. Also, shouldn't Givewell's information on funding gaps already be covering this issue?
Yes, although a robust donation or commitment would be better in both respects. Depends on what kind of donation strategy you want to compare this to.
It seems like your idea is built on the premises that charity markets are inefficient and that you can predict which way this inefficiency will go based on the amount of money the charity raises. I'm inclined to accept the first premise but not the second.
I can imagine the possibility that EA charities as a group would benefit from insurance against individual ones failing to meet their funding goals, but I think that problem would be best met either with a centralized fund or organization, or with people simply looking directly for underfunded causes rather than planning based on donation amounts. (Also what would you do with the money if you don't spend it on Givewell?)
Thank you, you make some really important points.
The point here is that there are two competing effects that go in opposite directions: (a) I don't want to spend money if I believe they are already raising enough to "survive", (b) I don't want to reward failure to raise money. The relative strength of these effects varies, with (a) getting stronger as the money moved goes up, versus (b) getting weaker. The upside-down quadratic is a first attempt at formalizing both intuitions. It's not necessarily the only formulation but it's good enough.
That is true holding the amount of money donated constant, i.e., a direct donation of $6,250 means more than a conditional commitment to donate $6,250. But I think a conditional commitment to donate up to $6,250 means more than a direct donation of $300 even if the EV of amount donated is $300.
This is taking GiveWell's information on funding gaps at prima facie value, but I've already indicated that I haven't had the time to review their top charity recommendations and found their reviews inadequate on these fronts. My estimate of funding gaps is more based on the idea "below what level would people think that GiveWell has failed?" Since GiveWell basically has Good Ventures money almost for free, it's the non-Good Ventures money moved that is actually of significance as far as judging GiveWell's growth and influence.
The risk-adjusted personal cost can be more than $300 too.