We recently wrote about three questions we faced that relate to donor coordination. This post is a continuation of that topic and may only appeal to donors who are particularly interested in this issue.

Over the past two weeks, we’ve been discussing the question: how should we allocate funds that donors give us for regranting to our top charities?

We plan to allocate these funds according to the recommended allocation explained in our previous post up to the point where one of our top charities reaches the maximum target from individuals we have set (explained here).

The question we face is: what should we do if we move more money than we expect, and one or more of our recommended charities reaches our maximum target from individuals before the end of giving season? In that case, how should we allocate funds given to us for regranting?

We see three options:

1. Give according to our recommended allocation even after charities hit their max targets. This options maximizes some donors’ agency. For example, if Alice (hypothetical donor) gave $10,000 to the Against Malaria Foundation (AMF) in early December, we would not take Alice’s gift into account when deciding how to allocate funds given to us for regranting. Were AMF to meet its maximum target from individuals, we would still allocate funds there and Alice’s gift would cause AMF to receive $10,000 more than it would have had she not donated. However, were we to keep giving to charities beyond the maximum we believe they can effectively use, we would be allocating funds suboptimally. Many of the donors who give to us for regranting do so because they want us to use our judgment about where additional funds will do the most good. If we mechanically follow our predetermined allocation, we would not follow these donors’ wishes in allocating their funds. We also believe it is likely that many donors who give to a specific charity would not want us to allocate funds beyond the maximum targets we have set for organizations. To continue the example, our guess is that Alice would often say, “I would rather donate to one of GiveWell’s other top charities once AMF has closed its funding gap.”
2. Give where we think it’s needed most. In the event that one or more of our top charities passes its maximum target, we would reallocate funds to the charities with remaining funding gaps. For example, if AMF were to receive $5 million before the end of giving season, we might choose to reallocate the funds given to us for regranting to the Schistosomiasis Control Initiative (SCI), the Deworm the World Initiative (DtWI) and GiveDirectly up to their maximum targets from individuals. The benefits here are clear: we would direct funds to charities that we believed to have more pressing gaps. The potential problem is that our doing so would arguably take agency away from other donors. For example, if Barbara gave $10,000 to AMF in early December, and AMF reaches its $5 million maximum target from individuals in late December, we might choose to allocate funds we hold to charities other than AMF. This means that Barbara’s gift did not effectively cause AMF to have $10,000 more; instead, it effectively caused the other charities to which we allocated funds to have $10,000 more. If Barbara wanted to support AMF and only AMF, our decision to reallocate the pool of funds over which we have discretion removed her ability to cause the charity of her choice to receive additional funding.
3. Compromise with donors who request it. In our previous post on this topic, we wrote about the compromise we reached with one donor who was planning to give to SCI. He believed SCI’s funding gap was larger than we did, and planned to give $1 million to SCI, so we compromised by meeting in the middle: we increased our target by half the size of his donation ($500,000). In this option, we would reach a similar arrangement with donors who have given or plan to give to our top charities, disagree with us about the charities’ maximum targets from individuals, and therefore want the charities to receive more funding even if it would cause them to go past the maximum targets we have set. We would do the same thing with these donors as we did with the SCI donor by adjusting our maximum targets by half the size of their donations.

Our tentative plan is to take option #3. Specifically:

• As a general rule, we are not planning to allocate funds to charities such that they would receive more than our maximum targets.
• Any donors who disagree with us about our targets should email us at info@givewell.org to let us know. If they tell us the size of their donation (and we can verify it), we will increase our targets by 50% of the size of their donation for purposes of allocating funds earmarked for regranting. This post is our announcement of this fact. We aren’t planning to announce this elsewhere.

We believe that this plan will result in relatively few donors emailing us and GiveWell largely reallocating funds given to us for regranting according to our best judgment if/when one or more of our top charities hits its maximum target from individuals. We have a few reasons for believing this is the right course of action:

• I corresponded with five donors who are long-time supporters of GiveWell top charities, none of whom give to GiveWell for reallocation. They encouraged us to follow our intended path and were not particularly concerned about the problem of removing donor agency (generally speaking — there was some disagreement). In particular, one noted that, over time, all donations are eventually fungible. That is, if donors give more to a charity than it can effectively use in year 1, it is more likely that we recommend it receive less money in year 2. (This is what happened to our recommendation of AMF from 2011 when we first recommended them to 2013 when we said that AMF had limited room for more funding.)
• Our impression from conversations with many people is that very few donors are concerned with issues of donor agency. We think that most people who land on our website and make a donation would rather we use our judgment to allocate funds optimally than allocate funds suboptimally to maximize the agency of their decision to support a specific charity.

Generally, we see a real distinction between (a) setting targets for charities based on incentives, how strong we think the charities are, etc. and (b) setting targets for charities based on how much we think they can productively absorb. When donors give differently from our recommended allocation, by default we interpret this as a disagreement with us on (a) but not necessarily on (b). As such, we are hesitant to “offset” donations on the basis of (a), but we are much less hesitant to do so on the basis of (b). We set “room for more funding” targets taking into account funds available from other sources, and this means that in some long-run sense we’re always creating the possibility of “offsetting” others’ donations.

It’s possible that we’ll change our mind based on reactions to this post.

We’re not yet sure. It’s possible that we will allocate surplus funds to GiveDirectly (which has the highest maximum target from individuals and is unlikely to reach it), but it’s also possible that we ultimately decide to allocate at least some funds to AMF (whose maximum target is more of a “soft” and approximate cap) or to standout charities. We have found that we need to go through a period of intense debate and reflection before we decide how to allocate funds in a given situation; that’s not something we’ve done yet for the hypothetical situation outlined here.

When do we plan to allocate funds given to us for regranting?

We are currently focused on funds that come in during giving season. Our accountant will close the books on December in mid-January, and we will grant funds out in mid-February.

How will we allocate funds that have already been donated for reallocation?

Some donors gave for regranting before we clarified the language on our donate page, and may have intended that their gift follow the allocation stated on our donate page. We plan to email these donors to tell them our plans and ask them whether they’d prefer that we allocate their funds (i.e., funds we have already received) according to our target allocation rather than according to the plan laid out in this post.

Are we concerned that this policy will incentivize donors to try and avoid this plan?

It is possible that this policy will cause some donors to (a) wait to give until after we’ve allocated the bulk of the funds we hold or (b) don’t report their gifts to us.

We think this will happen to some extent but will likely be relatively minimal. Our targets are based not only on funds given by GiveWell-influenced donors but on all funding an organization receives, so eventually we will learn about and incorporate their gifts into our targets.