# 42

Suppose that Alice is trying to figure out how to do the most good with her donation of \$1,000 this giving season, and can spend various kinds of resources to improve her decision. Unfortunately, many investments that could improve the decision quality would impose costs that are large relative to her donation: spending hundreds of hours (whether her own, those of charity staff, or of hired evaluators) investigating opportunities will cost more than her donation amount. She will also be limited in the projects she can fund: whereas a large funder can attract proposals for new projects, and fund a new position or startup organization, she seems to be limited to contributing to existing public opportunities.

One solution to this problem would be for Alice to work with Bob, a large donor, to construct a 'donor lottery.' Alice donates her \$1,000 to Bob's donor-advised fund, or DAF. Then Alice and Bob consult a random number generator to determine how to recommend donation allocations to the DAF. For example, they might plan that with 1/100 probability Alice would get to recommend the allocation of \$100,000 from the DAF, while with 99/100 probability Bob allocates the DAF without input from Alice.

In the event Alice 'wins' the donor lottery, then she can pursue a number of costly options to improve her decision quality, offset against a 100x larger donation, while the rest of the time she can avoid research costs. Effectively, the donor lottery cuts her research costs by 99% while opening up additional opportunities. If \$100,000 is not enough for some opportunities (e.g. if \$1,000,000 is needed to fund a new startup organization) she could repeat the process with still-larger donor Carol.

I have previously proposed that small EA donors who wish to capture economies of scale consider this method. On the whole, I think that it would be better for many donors giving less than \$100,000 per annum to use donor lotteries to consolidate into a smaller number of medium-large donors and capture economies of scale.

However, insofar as significant effort is required to access a donor lottery (which must be made regardless of the outcome), and it represents a novel untried method, it becomes less attractive. This post announces a concrete implementation of the donor lottery concept, thanks to effective altruist Paul Christiano, who has agreed to back it with his DAF account at Vanguard Charitable, and the first use cases, in hopes of working out bugs and establishing a precedent for others.

The remainder of the post will cover frequently asked questions and implementation details for other effective altruists who may wish to use or provide donor lotteries.

Donor lottery announcement

[ETA: finalized pre-draw details are in Paul Christiano's January 11th (GMT) post. The draw for this first lottery has now been completed, with Tim-Telleen Lawton winning, although there will be other future lotteries, hopefully with multiple providers.]

The donor lottery will be randomized using the first 10 hexadecimal digits of the NIST public randomness beacon output, with the result at 12:00 pm PST, January 15, 2017. This will provide plenty of time for people to contribute in either the 2016 or 2017 tax year (tax deductions apply for the year of donation to the DAF).

Participants will transfer funds to Paul Christiano's DAF prior to the draw, with each \$1,000 purchasing 1% of the possible values of the randomness beacon. If one of a contributor's numbers are drawn Paul intends to recommend allocation of \$99,000 from his DAF at the winner's direction to qualifying nonprofits. The difference from \$100,000, a 1% 'haircut,' would be recommended by Paul to a charity of his choice as compensation for offering the lottery from his charitable funds.

However, as is generally the case with DAFs, and in similar structures such as the Long Now Foundation's Long Bets, Vanguard Charitable is legally free not to follow Paul's recommendation (although such a case would be quite exceptional), and no binding contract exists. The participants thus far are familiar with Paul and one another in many cases, and the lottery relies on the honor system and the implausibility of accepting the reputational costs of visible failure to follow through on a public commitment in order to recommend different charities to the DAF.

The DAF has a minimum contribution size of \$5,000, so those donating less than that will be coordinating to donate in groups, but with numbers assigned separately. [ETA: donations under \$5,000 can now be accepted through donation swapping with Paul, no pooling necessary.]

[ETA: per Paul's follow-up post, the draw numbers have been rescaled to ensure a first-round winner (for the total donated to the lottery), with the winner then having the choice of placing a second bet to reach \$100k]

 Contributor Amount (\$) Low # (in decimal) High # (in decimal) Probability Timothy Telleen-Lawton 5050 0 121632721144.5 11% Gregory Lewis 5000 121632721144.5 242061157921.5 11% Ajeya Cotra 2200 242061157921.5 295049670103.5 5% Rohin Shah 2800 295049670103.5 362489594699.5 6% Helen Toner 2500 362489594699.5 422703813087.5 5% Nicole Ross 500 422703813087.5 434746656765.5 1% Howie Lempel 5000 434746656765.5 555175093542.5 11% Rebecca Raible 2000 555175093542.5 603346468253.5 4% Pablo Stafforini 2000 603346468253.5 651517842964.5 4% Aaron Gertler 500 651517842964.5 663560686641.5 1% Brayden McLean 5000 663560686641.5 783989123418.5 11% Benjamin Hoffman 100 783989123418.5 786397692154.5 0.2% Catherine Olsson 500 786397692154.5 798440535832.5 1% Eric Herboso 500 798440535832.5 810483379509.5 1% Ian David Moss 2500 810483379509.5 870697597898.5 5% Glenn Willen 500 870697597898.5 882740441576.5 1% Jacob Steinhardt 4000 882740441576.5 979083190997.5 9% Brandon Reinhart 5000 979083190997.5 1099511627775 11%

[ETA: added Howie Lempel December 7th, 2016; Rebecca Raible and Pablo Stafforini December 12th; Aaron Gertler December 16th; Brayden McLean and Benjamin Hoffman December 24th; Catherine Olsson Dec. 26th; Eric Herboso and Ian David Moss Dec. 31st; Glenn Willen Jan. 2; Jacob Steinhardt Jan. 4th; Brandon Reinhart Jan 8th.]

Others who wish to participate, please contact Paul F Christiano (at firstnamemiddleinitiallastname at gmail) for details regarding funds transfer to Vanguard Charitable, etc.

I would also hope that these early adopters might inspire other donor lotteries, both other providers and users.

Q: Step-by-step, how do I make use of a donor lottery as a small donor?

A:

[ETA: one lesson from the first lottery was that the minimum donation size for DAF contribution was a hassle, and donation swapping (donation to a charity the lottery sponsor was otherwise donating to, so they can reduce their donation to it accordingly and recommend channeling the funds to the winner's recommendation) was more efficient.]

1. Find a larger donor with a DAF willing to provide a donor lottery, such as Paul Christiano.
2. Determine the donation amount you would like a chance at, and agree to terms, e.g. "I donate \$1,000 to this DAF in exchange for a 1% chance at allocating \$99,000 from it."
3. Jointly announce the terms, and the public randomness source that will be used to determine the outcome, e.g. the NIST beacon, as discussed below.
4. Ensure copies of the announcement are secured, e.g. by tweeting a hash of the post.
5. Donate to the DAF.
6. The public randomness source provides its result.
7. If you 'won' the randomization, take steps to gather information or pursue donation opportunities that make sense as a large donor.
2. Talk to charities and experts who can afford to spend more time advising you on the larger donation.
3. Call for proposals, or request them from particular organizations and experts, for 'chunky' grants or projjects with minimum scale.
4. Spend more time thinking about and studying the options.
5. Hire research support, auditors, and other means of evaluation.
8. Make your recommendations to the DAF.
9. Your large donations go through.

Q: This idea seems novel. Is there an established precedent?

A: The Long Now Foundation has operated a closely related service, Long Bets, since 2002. This allows two people to make bets on future events, with the stakes going to the winner's charity of choice. The Long Bets rules page states:

There is no maximum amount. The bet money, treated as a donation to the The Long Now Foundation, must be paid at the time the Bet is made, and is tax-deductible immediately...

The charity designated should be IRS-approved if in the US. Foreign charities also are eligible. The Long Now Foundation cannot legally guarantee that the winnings will go to the winner's preferred charity. It is highly motivated to do so, of course, since how a bet is resolved is in the public eye.

This provides an established precedent of donations conditioned on uncertain events. Like this, and like DAFs, a donor lottery would not involve a legally binding contract, but such mechanisms can be quite reliable in practice.

Q: How can I be confident that the large donation will be forthcoming if I 'win'?

A: The terms of the donor lottery can be announced publicly, committing the backing of reputation. Money donated to a DAF can only be passed on to legally recognized charities. Using a trusted donor's DAF, and one for whom the amount is relatively small (so that reputation matters more) would also provide additional assurances. The announcements should be copied to secure and unchangeable public storage (some participants in this lottery will be tweeting hashes).

Ultimately, however, this setup relies on the honor system and reputation.

Q: Where can we find a trusted public random number generator?

A: The National Institute Standards and Technology in the United States offers a public randomness beacon, which "generates full-entropy bit-strings and posts them in blocks of 512 bits every 60 seconds. Each such value is sequence-numbered, time-stamped and signed, and includes the hash of the previous value to chain the sequence of values together and prevent even the source to retroactively change an output package without being detected."

To randomize, the parties can publicly state a particular minute (e.g. 5:04 pm, December 8th, 2016) for the determining output value. Then one can take the first 10 hexadecimal digits (which can show numbers from 0 to 1,099,511,627,775) and use them to cash out a probability. For example, a 1% chance could be implemented as "the first 10 digits lie in the range from 0 to 10,995,116,116,278."

Other sources, from weather to market movements, are abundant.

Q: If I believe I would do more good per dollar with a smaller chance of a larger donation, why not just gamble or make risky investments?

A: In principle, as discussed in my blog posts on donor lotteries, these methods could also be used by small donors to take advantage of economies of scale in expectation. The advantage of the donor lottery via DAF is just in reducing overhead and transaction costs.

Q: Does it matter what other donor lottery participants would do with their funds?

A: Suppose you are one of 10 people each contribute \$10,000 to a pool, and one randomly gets to allocate all \$100,000. Should you care what charities the other participants would recommend? On average, you are not transferring any net dollars to the charities recommended by the other participants (90% of the time you pay out \$10,000 to one of those charities, and 10% of the time \$90,000 gets directed to your preference instead). However, the reason to participate in the first place would be if marginal returns to donations to a charity vary with the amount, and donors hope they are increasing.

If other donors would support charities which one thought were actively bad (even with the chance to think more carefully in cases where they 'win' the donor lottery), and do so more effectively per dollar with larger donations, then one could do harm by helping harmful donations.

For charities with positive or neutral effects, the worry would be that participants are mistaken about increasing returns to donation size. For example, a donor who gives \$10,000 to an organization with a budget of \$50,000 might well expect that the organization would much prefer \$10,000 with certainty over a 10% chance of \$100,000. Such diminishing returns can give reason for risk-aversion. If one is focused on a small cause or organization, where diminishing returns matter over small ranges, this is reason to worry about the chance of the area facing a financial shortfall due to chance lottery outcomes. One way to address that is by conducting a lottery among supporters.

Q: How can the lottery provider offset the risk their DAF takes?

A: While for donation amounts that are small relative to funding for one's causes, there is a good case for altruists being approximately risk-neutral (e.g. for small donors), diminishing returns mean that for large pools of money there is some cost to taking on risk for donors who are large relative to the causes they work on. So a lottery provider could arrange a 'haircut,' e.g. Alice donates \$1,000 to Bob's DAF, and with 1% probability gets to allocate \$99,000, rather than \$100,000. The difference serves to compensate for the risk that Bob's DAF will be depleted of cash to fund charitable opportunities he prioritizes.

The optimal 'haircut' would vary depending on alignment between Alice and Bob, Bob's bankroll, diminishing returns in the causes Bob supports (relative to the size of his donations), etc.

Q: How large an amount should I consider using a lottery to reach? Does this imply all EA donations should be merged?

A: There are both helpful economies of scale and harmful diseconomies of scale. For example, if all effective altruist donations were consolidated in a single giant foundation, then they would all be at risk of mishap together, e.g. if a catastrophe befell it. If one donation risked attracting controversy or straining management bottlenecks for the whole pool, the cost would be much greater than for several smaller pools. A single pool would also suffer from reduced diversity of approaches, and could miss out on local knowledge of particular opportunities (which should give donors reason to think that they have diminishing returns to larger donations, not increasing returns).

A donor lottery executed through a DAF also requires that donations go to registered charities, excluding such things as political contributions and informal help to particular individuals (the latter being a case where local knowledge is especially important).

I would suggest that small donors consider using lotteries to move into the range of \$50,000-\$500,000. This is most attractive if one would otherwise be donating to organizations or areas where one's funds are small relative to total funding, and fungible with other donors, but have hope of identifying or creating better opportunities as a larger donor. It is least attractive when one one would otherwise be a large component of a small funding pool, especially exploiting local knowledge, so that one expects diminishing returns over the relevant range.

Additional lottery stages to reach quantities in the millions or more face high risk of missing low-hanging fruit and suffering from diminishing returns. However, they might make sense in some circumstances if one expects strong neglected opportunities in the creation of new organizations or other foundation-like grants.

Q: What if my target donation amount is too low or high for a known donor lottery?

A: For the smallest donations, e.g. \$10 or \$100, I would suggest an informal randomization with a friend to start. If that randomization delivers a larger donation, then it could be funneled through a more substantial donor lottery.

If a friend is donating to a charity with budget that is large relative to their donation (e.g. GiveDirectly or AMF) they can easily offer a simple honor system donor lottery without much worry about diminishing returns.

If you would like to have a chance at a quite large donation, e.g. millions of dollars, I would recommend a multi-stage process. First reach a donation total in the hundreds of thousands, either directly or via a donor lottery. At that point, try to be connected with much larger donors

Q: What if I want to delay my donation over time after winning?

A: Some of the opportunities for improved decision-making as a larger donor take time. If funds will be sitting in the DAF for some time, compounding due to investment returns, you may want to make an arrangement with the donor providing the lottery as to the treatment of investment returns, e.g. the stakes growing in line with returns for the DAF as a whole.

Q: Are there any interesting alternatives to a random number generator?

A: Instead of using the results of a public random number generator, one could instead 'bet' (in the style of Long Bets, but with less of a haircut) on predictions that are relevant to charitable allocation, or the relative predictive skill of the parties. For instance, a payout might occur conditional on a particular election result, or scientific benchmark. This could be used to insure against special needs (e.g. perhaps you care more about donations if certain technologies are developed, or if certain problems become more severe), and to make it more likely that better decision-makers allocate the donations.

Q: Why not instead allocate the donations of a group of small donors using randomization?

A: Another way to capture some of these benefits is to team up with a few friends donating on the same scale, and use randomization (weighted by donation amount) to pick one of you to allocate the donations of the group (using the honor system). It may be significantly more difficult to buy chances of large donations, but this is a useful alternative.

Q:  Won't donor lotteries reduce the amount of practice people get thinking about charity?

A: Thinking carefully about one's donations can be a helpful educational exercise, and build important skills for a would-be effective altruist, as well as participate in various social processes. Giving Games let people allocate money as a way to increase their interest in donating effectively and in investigating their options. This may be a reason not to allocate one's entire donation budget to a lottery, and to donate a smaller amount to one's best guess, for practice, while other funds go into a lottery for higher expected quality.

If the indirect benefits such as practice and signaling are sufficiently great, and economies of scale sufficiently low, this could be a reason to avoid donor lotteries entirely.

Q:  My question isn't answered above, or I would like to participate in matchmaking for a donor lottery.

# Reactions

Sorted by Click to highlight new comments since:

Cool. I'm in with \$2k.

I'd like to contribute \$1k. Would you like to coordinate together so we can meet the \$5k threshold?

Edit: After further consideration, I decided to instead donate \$500 to the donor lottery while increasing my direct donations elsewhere.

I have been put in touch with other donors that are each contributing less than \$5k, but you can just team up with us. Email me at MyFrstName at MyLastName, followed by the most common domain extension.

Ideally there should be a better procedure for doing this; the associated trivial inconvenience may be discouraging some people from joining.

Note that this is now being implemented by donation swapping, so small donors don't have to put in any extra work.

Thanks for pushing this Paul and Carl.

I'm quite interested in backing one of these, but without a DAF have always been a bit too lazy to work out how I would structure it. Depending on the response to this I'I'll think about it again

Looks like Tim Telleen-Lawton won, as the first ten digits of the beacon at noon PST were 0CF7565C0F=55689239567. Congratulations to Tim, and to all of the early adopters.

Seems great to reduce the costs of analysis, and to increase the capacity for EAs to fund projects that have a minimum funding size.

I like that you've encouraged folks to take different slices from the same random number. For the benefit of others reading, this reduces Paul's risk and increases the amount of donations that he can match. e.g. if you have a real random variable ranging from 0-100 inclusive, then one person takes 0-1, one person takes 1-2.5, and so on, filling up as much as you can!

What happens if the \$100K isn't reached in time?

Paul has already backstopped the \$100k from his own funds in the DAF. He is not dependent on the contributors for the amount of the payout.

Does he control all the unclaimed numbers?

Yes.

How are funds kept within the DAF? If it's not being held in an aggressive investment then we'll want to make the donation as late as possible before the draw.

Also: is there no type of financial instrument that will offer extremely high short term variance without reducing expected value?

The donations will be invested in equities (some US, some international, some emerging). Returns between the donation and the drawing seem like a very minor consideration, maybe you are looking at 0.5% in expectation, or more like 0.1% for investors with access to leverage.

There are high variance financial instruments in the derivatives space, but you need to have the right financial accounts, EV is negative, and generally there are various kinds of hassle. But it's an option.

For some assets you could donate the upside while selling on the downside to take a loss, so it could actually be more tax-efficient. This probably requires a year of foresight though + rules out some options.

I think that in general lottery-ticket assets tend to be pretty expensive. Also, if the volatility is correlated with the market, then there are further problems from correlations between your returns and other donors' returns.

If you had the right financial accounts, you might just short something, because that's very high variance and anticorrelated with other donors. edit: actually as Michael points out, it has the same variance (though it has more extreme downside risks)

Going short on an asset has the same variance as going long, but with opposite expected value (actually slightly lower because of borrowing costs).

Oh good point, shorting has more extreme downside risks but the same variance. I wonder if shorting causes you to reach the point where you double or lose your pot any quicker than going long.

Also very negative expected value though.

[anonymous]1

I think this is a very good idea. Unfortunately, I don't really know any of you, and I don't think it's worth the time to thoroughly research your reputations and characters, so I'm not going to contribute.

However, I would be interested in a registered charitable organization whose sole purpose is to run a donation lottery annually. In fact, I would donate to the operations of such a charity if the necessary safeguards and/or reputation were in place. Seems like an easy "bolt-on" project for GiveWell, no?

If anyone else would like to see a permanent donor lottery from GiveWell, let me know how much you're willing to contribute to start it (via private message if you prefer). I'll total the amounts in a few weeks and present to GiveWell. Maybe it will pique their interest.

Thanks Mac, I agree that's the general direction this should go given success with this initial trial (and the lessons from actually doing it), and hope there is a good option for you in future (the current lottery is closed for the last few days pre-draw). That organization may or may not be GiveWell, although they are certainly aware of it.

6 of the 18 participants are GiveWell employees (including Open Philanthropy Project staff), one is a former GiveWell employee, and one is an advisor to the Open Philanthropy Project (as are Paul Christiano and myself).

Holden Karnofsky wrote favorably about it in the GiveWell blog post on staff members' personal donations, along with Tim, Ajeya, and Helen (the latter 3 participated).

We also (with the consent of the participants) rescaled the win probabilities and payout recommendation to guarantee a winner (who can then make a second bet to get from \$45,600 to \$100,000), so as to ensure we can learn from that process too before trying a bigger institutional version. See Paul's recent follow-up post.

So I'm hopeful something along the lines you suggest can be managed in future (although this round is closed now).

This seems like a reasonable concern, and longer term building good institutions for donor lotteries seems valuable.

However, I suspect there may be more overheads (and possible legal complications) associated with trying to run it as part of an existing charity. In the immediate, I wonder if there are enough people who you do trust who might give character references which would work for this? (You implied trust in GiveWell, and I believe Paul and Carl are fairly well known to several GiveWell staff; on the other hand you might think that the institutional reputation of GiveWell is more valuable than the individual reputations of people who work there, and so be more inclined to trust a project it backs not because you know more about it, but because it has more at stake.)

[anonymous]0

However, I suspect there may be more overheads (and possible legal complications) associated with trying to run it as part of an existing charity

Given the current level of interest, the informal, small, and disconnected donor lotteries may be more efficient for the reasons you mentioned. My hunch is that donor lotteries could quickly grow to a non-trivial size, at which point I believe the economies of scale achieved by an institution would dominate.

you might think that the institutional reputation of GiveWell is more valuable than the individual reputations of people who work there

Yes.

Lottery entrant Ben Hoffman has a new explainer post on donor lotteries.

Amazingly written post. Thank you for sharing.

So I'm guessing the idea is that donors will do something a bit more complex than just throwing the money over to AMF?

If you were confident that additional investment won't let you pick something you favor over AMF, or (with larger or staged lotteries) fund a new project you would fund over AMF, then you believe you have constant returns to donation budget over the relevant range. So you wouldn't expect to do extra good.

Most of the reasons for increasing expected returns with donation budget (to take more time to make a better decision in a minority of world-states, to open up new opportunities, to make it worthwhile for charities to spend more time answering your questions) involve the possibility of something more complex.

How do you view this interacting with considerations of giving now vs later? An alternate plan would be invest my donations for e.g. 10 years, then those plus interest would be enough to make me a largeish donor for that year - presumably with lower transaction and coordination costs.

If there are relevant economies of scale then they would be worth capturing in whatever period you are donating. Restriction to reaching the desired scale through investment returns by the desired time periods would rule things out for most small donors and belief sets where it matters, so I think measures like donor lotteries or derivatives would still add value if one expects increasing returns.

Donating every few years from one's DAF in larger increments makes sense to me, but on a 10 year time scale you could miss out on the expiration of low-hanging fruit as funding in relevant causes becomes more crowded.,

Thanks Paul and Carl for getting this off the ground!

I unfortunately haven't been able to arrange to contribute tax-deductibly in time (I am outside of the US), but for anyone considering running future lotteries:

I think this is a great idea, and intend to contribute my annual donations - currently in the high 4-figures - through donation lotteries such as this if they are available in the future.

(If EAs end up committing a significant proportion of the \$100k (or even all of it), will Paul reduce his 'haircut' or not?)

If \$50,000 gets contributed, it reduces my risk by a factor of 2, so I could halve the fee (and if \$100k got contributed I could reduce it to zero). I'll probably do that.

It looks like the total will be around \$50k, so I'm going to reduce the cut to 0.5%.

Curated and popular this week
Relevant opportunities