Allocate Donation Election Funds by Proportional Representation
I found this year’s Donation Election to be an intriguing experiment in charitable giving. However, I was disappointed to see funds allocated in a very concentrated way, with most of the funds given to the top few most plausible candidates. I believe a proportional approach would work better.
For those who don’t know, proportional representation is a method of voting where the seats in parliament a party gets (which would correspond to the funds a charity gets in the donation election) are delegated proportionally to the number of votes it receives (or in a donation election, the amount of money put in by people who choose this charity as their top choice). Proportional Representation has two big advantages over traditional first-past-the-post and other more concentrated voting systems.
First, it removes all tactical voting. In a first-past-the-post system, votes to third parties are usually wasted as these candidates have little chance of winning the election. This is less of an issue in ranked-choice voting, the system used in this year’s donation election, but it is still an issue. For instance, consider an Instant-Runoff-Voting election for 1 candidate (which would correspond to if all donations were given to the candidate with the most votes) with third parties that are polling at the following rates:
Center-left party (first-place votes): 40%
center-right party (first-place votes): 25%
Far-right party (first-place votes): 35%
Head-to-head polling:
center-left vs. far-right: 55%-45%
center-left vs. center-right: 45%-55%
center-right vs. far-right: 60%-45%
Since the far-right party is likely to lose the head-to-head matchup, far-right voters are incentivized to switch their first-place vote to the center-right party in order to beat the center-left. Strategic voting like this turns out to be a broad weakness of all non-proportional electoral systems. And indeed, in a donation election, we don’t want there to be an incentive for people to strategically misstate their choice.
Second, it allows for more expressive representation. In a more concentrated election system, voters are often faced with more of a binary choice: left-wing or right-wing, animal welfare charity or global health charity. This deprives voters of a more expressive vote that better accounts for the minutiae in their views. In a proportional voting system, as long as there is one party that represents a voter’s views very well, this is never an issue. The voter can just cast their vote for that one party and not have to make the hard decision where it feels like they are voting for the lesser of two evils. In the case of the Donation Election, as long as there is one charity that each voter thinks is doing great work and can use the extra funding, they need not worry about comparing the viability of different charities which are not their top choice.
I believe this is a broadly-agreeable common-sense change that would be beneficial to implement for next year’s donation election.
I wholeheartedly agree, and think we need to look elsewhere to apply this model.
Donor Lotteries unhealthily exhibit winner-take-all dynamics, centralizing rather than distributing power. If this individual makes a bad decision, then the impact of that money evaporates—it’s a very risky proposition.
A more robust solution would be to proportionally distribute the funds to everyone who joins, based on the amount they put in. This would democratize funding ability throughout the EA ecosystem and lead to a much healthier funding ecosystem.
As an extra benefit, the Proportional Representation can be implemented in a decentralized way—and people can opt into it on a case-by-case basis, without the approval of any central authority.
To implement this, donors should simply give to the place they believe to be most deserving. You might be concerned that this is antidemocratic! But in fact it’s entirely democratic, and you’re just taking responsibility for paying out the proportion of the total that you represent. Others can in parallel take responsibility for their shares. Thus in a distributed manner you reach deeply democratic outcomes.
You may lose the information-gathering function though—in @RedStateBlueState’s model, we learn what proportion of the Forum user base voted for Charity X (and what portion of those users’ votes it received). That would be difficult to get in your decentralized take—an org would have to determine ForumUser status, and figure out the share of each user’s vote it received.
Ah but the genius of it is that you still have all of the information—it’s just also distributed! Each person knows exactly who they voted for, and as a bonus they avoid having to entrust any of their data to a centralized system which could be controlled by any kind of nefarious types. (I heard that the US had issues with election integrity in 2020, and if a major nation can’t manage this, I think it’s really a bit much to expect a shoestring-budget org like CEA to manage it.)
One crux here is what the purpose of the donation election is. My take is that it is in significant part an information-gathering device, and that your PR proposal would reduce that value. Under at least pure PR, a rational voter will likely concentrate all their votes on a single org. We would learn almost nothing about voter preferences beyond their top choice.
IIRC, I suggested allocating monies using a variety of decision rules, of which PR-with-threshold would probably be one. Adding some PR element should reduce an information problem with the 2023 design—voters weren’t motivated to ponder the merits of any org unless they believed it might end up in the top three.
PR systems are more prone to give seats to extremist views—such as far-right racist parties. There’s no direct analogue here, but there might be reasons that allocating any money to outlier views might be undesirable. The usual approach in electoral systems is to set a threshold, commonly 5 percent, below which the party gets nothing.
If you do PR with a threshold like that, voters still have to guess whether their preferred charity will make the threshold. Do you think that poses some of the same downsides as IRV here?