Donor coordination and the “giver’s dilemma”

This year we’ve dealt with some particularly intense manifestations of what one might call the “giver’s dilemma.”

Imagine that two donors, Alice and Bob, are both considering supporting a charity whose room for more funding is $X, and each is willing to give the full$X to close that gap. If Alice finds out about Bob’s plans, her incentive is to give nothing to the charity, since she knows Bob will fill its funding gap. Conversely, if Bob finds out about Alice’s funding plans, his incentive is to give nothing to the charity and perhaps support another instead. This creates a problematic situation in which neither Alice nor Bob has the incentive to be honest with the other about his/her giving plans and preferences – and each has the incentive to try to wait out the other’s decision.

In this stylized example, our picture of the ideal resolution is for Alice and Bob to be honest with each other and agree to “split” the funding gap: each should give $X*(1/2). In a world where this behavior is expected, people can be honest with each other about their plans and ensure relatively “fair” resolutions rather than withholding information and trying to anticipate each other’s behavior. In real life, the situation can be more complicated. This year, in finalizing our funding targets for top charities, we faced the following three questions. In each case, we are actively trying to avoid behavior that amounts to “waiting out” or “canceling out” donors who are using our research, and trying to preserve the incentive of each donor to give as planned and share information about his/her giving. 1. Should Good Ventures wait until the end of giving season to decide on the size of its grants? (Note: we resolved this question last year and stuck to the same resolution this year, but it remains a puzzling question and a good example for this topic.) Once we arrive at a “target amount” we would like to see each charity raise, we could wait until the end of giving season before recommending grants to Good Ventures. Since Good Ventures is giving a great deal (a bit more than we expect all individual donors combined to give over the course of the year), this would probably ensure that each charity hits the exact target we set for it, regardless of what individual donors decide. We think this would be a very problematic approach. It would render individuals’ decisions about which charities to support effectively meaningless: each dollar given to one of our top charities would be fungible with dollars given to other top charities, and anyone who gave to one of our top charities would effectively/implicitly be supporting “GiveWell’s overall allocation” rather than the charity of their choice. This, in turn, would give individuals the ongoing incentive to try to wait out Good Ventures, and would potentially reduce both giving and reporting of giving. Instead, we and Good Ventures agreed that the Good Ventures grants should be finalized and announced before the start of giving season. Our recommendations have taken into account what we expect individuals to do, but our projections are approximate and non-binding, and any given individual decision has real impact on the total raised by the recipient charity. It’s true that in an indirect, approximate sense, supporting a given charity now may change our projections of what will happen next year and therefore cause Good Ventures to give less to the target charity next year. However, this is a very indirect and rough link, since situations change dramatically from year to year; our ability to predict donor behavior is quite limited; and we generally aim for round numbers with Good Ventures grants. 2. Should we post updates on how giving is going, so that donors can adjust their behavior accordingly? Posting too many updates would risk a similar problem to the one described above. We don’t want a situation in which each donor’s gift to charity X causes another donor to give less to the charity; this would create an incentive for donors to try to wait each other out. We do intend to post a notice when a given charity appears sure to hit its “maximum” – the most we think it could absorb before hitting a point of seriously diminishing returns. This will help donors avoid supporting a charity to go well over what we think is a reasonable amount of funding, without allowing donors to “cancel each other out” at a more granular level. In other words, if you give to a charity, you can expect that your donation will raise the total amount the charity takes in, by the amount of your donation – unless the charity ends up easily hitting its maximum, in which case your donation will be partly offset by the behavior of other donors who agree with us about where the maximum ought to be. We think this is a reasonable practical position. Note that we will also be doing an update in early 2015 on each charity’s room for more funding situation. 2015 donors may take this information into account and may therefore do some degree of “canceling out” 2014 donors. Our basic position is that “canceling out” can be a good thing when it is a result of charities’ running out of room for more funding, and is acceptable/inevitable over relatively long periods of time, but that we should avoid situations in which some GiveWell-influenced donors are “canceling out” others at a more granular level and targeting an exact amount of total money moved. 3. How should we handle major donors who have privately told us about their giving plans and disagree with us on ideal funding targets? A donor recently told us of their intention to give$1 million to SCI. We and the donor disagree about what the right “maximum” for SCI is: we put it at $6.3 million, while the donor – who is particularly excited about SCI relative to our other top charities – would rather see SCI as close as possible to the very upper end of its range, meaning they would put the maximum at$8.3 million. (This donor is relying on our analysis of room for more funding but has slightly different values.)

If we set SCI’s total target at $6.3 million, and took into account our knowledge of this donor’s plans, we would recommend a very small amount of giving – perhaps none at all – this giving season, since we believe SCI will hit$6.3 million between the $3 million from Good Ventures,$1 million from this donor, and other sources of funding that we detailed in our previous post. The end result would be that SCI raised about $6.3 million, while the donor gave$1 million. On the other hand, if the donor had not shared their plans with us, and we set the total target at $6.3 million, we would recommend$1 million more in support to SCI this giving season; the donor could wait for the end of giving season before making the gift. The end result would be that SCI raised about $7.3 million, while the donor gave$1 million.

We discussed what to do about this situation at length, and ended up pursuing a version of the “split the difference” heuristic. Since an informational advantage on our part would lead to SCI’s raising a total of ~$6.3 million, while an informational advantage on the donor’s part would lead to SCI’s raising a total of ~$7.3 million (in both cases including $1 million from the donor), we have set the total target at$6.8 million – the midpoint of those two figures. This is a compromise between “making no use of the information the donor has given us” (which we feel would be unfair to our audience) and “making full use of the information the donor has given us” (which we feel would be antagonistic to the donor and would give an incentive to not share plans with us in the future). We’ve discussed this resolution with the donor and agreed to it as representing good-faith behavior on both sides.

In all of the above cases, we’re largely improvising and looking for pragmatic solutions. We haven’t worked out a detailed framework for laying out principles and responding to different possible situations regarding the “giver’s dilemma.” We’d love thoughts from our audience on these matters.

• Maxwell Fritz on December 2, 2014 at 2:28 pm said:

Appreciate the transparency.

What would you have done if the donation completely filled your actual 6.3M funding target, and the donor then insisted that it be kept private, and wasn’t amenable to any half-measure compromises?

I would really hope you would have disclosed it anyway. Yes, there are downsides, but the alternative (making something a top charity when it has no room for funding) is worse.

This particular compromise is imperfect but ultimately fine. I really don’t like GiveWell changing its target in bold due to a donor opinion of what the target should be. But I get that it’s a situation without policies in place. Meeting in the middle, revealing the conflict, and asking for help with future approach was an okay line to take.

• Peter Hurford on December 2, 2014 at 4:02 pm said:

…I’m glad that GiveWell is having the problem of moving too much money.

• Colin Rust on December 2, 2014 at 7:34 pm said:

Makes sense to me.

Like Peter Hurford said, these are great problems to have.

• David Barry on December 3, 2014 at 6:55 am said:

I have four ideas (1a, 1b, 2a, 2b) for a fair adjustment to the funding target in response to the situation in question 3. In increasing order of my personal preference:

Idea 1a:

If GiveWell didn’t know about this donor, they would try to move $3.3mn to SCI (I’m including Good Ventures in this total since that makes sense to me). The donor will donate$1mn.

So, $3.3mn says that the funding target is$6.3mn, and $1mn says that the target should be$8.3mn. Take a weighted average and the new target is $6.77mn. Idea 1b: But in the above case, GiveWell will now try to move$3.77mn instead of $3.3mn, the weighted average funding target should be re-calculated. Which will lead to a new amount moved by GiveWell, so you re-calculate again…. Solve for the equilibrium, and you get a new funding target of$6.72mn.

These are close to what happened in reality, but both of these give compromises that might be undesirable in some circumstances. e.g., suppose that the donor thinks the funding target should be $25mn rather than$8.3mn. Then the donor’s $1mn increases the target by$4.35mn (idea 1a) or $2.68mn (idea 1b). Idea 2a: If the donor didn’t tell GiveWell about their planned donation, then they could wait until GiveWell’s funding target was reached, and then add a full$1mn.

GiveWell’s $3.3mn says that the target should be$6.3mn, and the donor’s $1mn could get it to a maximum of$7.3mn. Take a weighted average, and the compromise target is $6.53mn. Idea 2b: Iterate idea 2a as was done in idea 1b. The target becomes$6.52mn.

I’d like to think that if I had a million dollars to give away, I’d be satisfied with a real increase of $220k to my preferred charity with the remainder funging elsewhere. But I’m not in that situation, so I don’t know! I put a spreadsheet on Google Docs here if anyone wants to play with the numbers. (I’m not sure how best to generalise to having multiple large donors with differing preferred funding targets.) • Rob Bensinger on December 4, 2014 at 1:19 am said: It’s important for GiveWell to preserve its reputation for transparency; but it’s more important for GiveWell to preserve its reputation for keeping its word. So if GiveWell agrees to keep a donation secret, then learns that the donation has a major impact on a top charity’s funding target, it should keep its promise and sacrifice some transparency. (And I say this as an independent small donor who would be left in the dark in this circumstance.) This is a good reason not to make many explicit non-disclosure agreements in the first place. The goal here is to move billions of dollars to highly effective charities over the coming decades; preserving GiveWell’s credibility is a means to that end. So how much you agree not to disclose should depend on the value you expect from big vs. small donors over the long haul, and (in either case) how useful it is for GiveWell as an organization to know a lot about big donors’ behavior. • Ian Turner on December 6, 2014 at 12:26 pm said: All this talk about waiting seems to be mainly oriented with respect to the December giving season. Does it follow that a donor who wants to be sure his or her donations are not subject to a substitution effect would want to give in the spring? • Michelle Hutchinson on December 30, 2014 at 1:34 pm said: As Peter says, what great problems to have! I think it would be great to foster an even more collaborative attitude than the one described above. The ‘ideal’ described for Alice and Bob seems to be the best situation conditional on them trusting their own judgement for donations more than the other’s judgement. If that wasn’t the case, they presumably wouldn’t mind which of them gave to the original charity – the important thing would be to make sure that the charity got$X, and that the remaining $X was given to the next most effective cause. Perhaps in a sense it’s trivially true that everyone trusts their own judgement more than other people’s, since that’s the judgement you have to act on (even if the judgement says ‘follow person y’s advice). But I would have expected the ideal situation between Alice and Bob to be for one of them to donate to the original charity, and the other to donate to the organisation they together come up with as the next best. Assuming they have similar values (which admittedly is a strong assumption!) pooling their information seems likely to come up with a better choice than individually giving to separate second-choices. (Though this doesn’t preclude splitting the second donation if they decide that is best.) In practical terms, this might well simply end up with Alice and Bob ‘splitting the difference’ (giving$X to the original and $X/2 to each of their preferred second choices). But even if that’s the case, the collaborative attitude could have beneficial results: for example, if money could be saved by Alice giving the full$X to the original charity and Bob splitting his 2 ways between her and his choice (since that involves fewer transactions in total).

• One solution would be for GiveWell to set a target for each category for donations moved through its website. It could track these with the familiar thermometer or other graphic depiction of how much was still needed. From a psychological standpoint, seeing how much is still needed can powerfully move donors to either focus on one which needs just a bit more to push it over the top or perhaps the one which still has farther to go. It essentially creates a race.