Political spending is a classical example of a zero-sum conflict. If Dustin gives $1M to Clinton and Peter gives $1M to Trump, they might both be happier if all the money went to charity.
This suggests the following scheme: rather than having people contribute directly to Hillary or Trump, they give the money to me and choose a charity. If I receive $1 for Hillary and $1 for Trump, I cancel them out and instead send each dollar to the chosen charity. If there are any dollars left over for one candidate or the other, I send them to the candidate as usual. (This looks like it was nearly-implemented by Repledge, HT Claire Zabel for bringing that up recently. Also see Toby Ord on moral trade, of which this is a central example.)
The result is that we redirect a bunch of money from zero-sum conflict to charity, and everyone is happier. What’s not to love?
There are a bunch of problems with this scheme. For example:
- Each contributor might believe that they are spending dollars more intelligently than the people they are matched with, or that their candidate is better at spending dollars, or that their candidate is in fact preferable and hence will get more mileage out of advertising.
- The two candidates presumably have different values for money, so at least one of them would rationally prefer to not have donations matched.
- If I create a surplus for my preferred candidate, someone who mostly cares about charity may decide to “pledge” for the other candidate just to redirect my money to charity.
- The whole thing requires some effort, and it might be a scam, and maybe there is a mistake in the argument, and so on.
Each of these issues decreases the value of making political contributions through a matching scheme. So unless people think that charitable donations do a significant fraction of the good good of political contributions, they aren’t going to do it.
In addition to the direct problems, making a watered-down contribution sends a different social signal: it shows that you care about charity almost as much as the victory of your preferred political candidate. If you are deeply immersed in a certain kind of social circle then this may get you brownie points. But for most people making political donations, I suspect that’s the wrong signal to send.
Greasing the wheels
We might be able to improve the situation by using a little bit of extra money to grease the wheels. Suppose that I have $1M lying around that I was planning to send to GiveDirectly. Instead, I set it aside in a special matching fund, and adjust the contribution matching scheme so it works like this:
- When I get $1 for Clinton and $1 for Trump, I send them both to GiveDirectly, like before. (Except that now the money is going to my charity of choice, instead of the political contributors’ charities of choice.)
- If I have any unmatched contributions, I send them along to the candidate, like before.
- But now I sweeten the deal. If there are $X unmatched, currently on their way to Trump, I add in another $X out of my special matching fund, so Trump receives 2*$X. Similarly if there are $X unmatched dollars currently heading to Clinton, I double it to 2*$X.
- Whatever is left over in the matching fund I send to GiveDirectly, or at least I definitely don’t spend it on politics. If the matching fund runs out, then I just only give $1M to the winning candidate.
So what’s the calculus like for a Clinton supporter considering giving $1 to my scheme?
- If Trump comes out ahead, then giving an extra $1 to Clinton increases Clinton’s take by $2.
- If Clinton comes out ahead, then giving an extra $1 to Trump decreases Clinton’s take by $2.
Either way, they get a 2x multiplier (as long as the margin of victory is <$1M one way or the other; if it’s a blowout then we’re back to 1-for-1).
So we still have the concerns I listed before–there are some costs, and uncertainties about whether this really works, and possible asymmetries, and weird incentives. But now even if all I care about is politics, it’s pretty plausible that I should be giving to the contribution-matching-scheme. All it takes is for those frictions to be less than a factor of 2, and that sounds realistic.
(We may want to keep the totals private until the big reveal at the end, to address legitimate concerns about one donation influencing the amount that other donors give—either increasing donation by the opposition or substituting for donations on their side. By the same token, I probably want anyone with inside info about the totals to be barred from participating, since they are especially likely to join whatever side happens to be losing.)
How pricey is this?
Suppose that one candidate gets $A and the other gets $A+$B.
Then I’ve moved $2A dollars from politics to charity, and I need to spend $B of matching funds. So I come out ahead as long as $2A > $B, i.e. as long as the short candidate raised 1/3 as much as the long candidate.
For comparison, in the 2012 presidential election both candidates raised comparable amounts; in the 2016 election Clinton raised about twice as much. I 2016 is one of the most lopsided elections in recent memory, and that the majority of elections are well within the 1:3 ratio.
Of course it would be easy for a particular contribution matching scheme to get uneven adoption amongst donors, such that it is much more imbalanced than the population at large, but we could try to be careful about outreach to get a balanced audience.
Obviously I could adjust the 100% matching—I could have contributed 200% of the unmatched contributions, or 50%. Repledge’s proposal corresponds to 0%. Doubling seemed nice because 2 is a really hefty multiplier yet still small enough that you could easily turn an altruistic profit in a balanced fundraising race.
I could instead try to mitigate this problem by adjusting my matching formula to take into account the projected imbalance. For example, if I expect Clinton to raise twice as much as Trump, then I might do the following:
- If Trump raises $X and Clinton raises 2*$X, then I don’t pay any matching dollars to either candidate. (I still send along the $X unmatched dollars to Clinton though.)
- If Clinton doubles Trump’s donations and then raises $B beyond that, then I give an extra $B to Clinton.
- If Clinton falls $B short of doubling Trump’s donations, then I give an extra $B to Trump.
As before, this means that donors are getting extra bang-for-their-buck regardless of what happens: Clinton donors always get 2x bang for their buck, while Trump backers always get 3x bang for their buck.
If I calibrate this correctly I could potentially make the whole scheme a lot cheaper—indeed, if I guess the ratio exactly right then it’s free. Probably the easiest way to set the ratio is to use the amounts donated in similar historical elections.
You could run the same scheme where you keep the money instead of giving it to charity, and it’s still possible you’d turn a profit. Or you could give half to charity and keep half for yourself. I think the version with everything going to a broadly acceptable charity is significantly more promising though.
If you try to keep the money yourself, people would feel like you were cheating them and generally being a jerk, and so wouldn’t go along with it. This might take the form of moral or proto-decision-theoretic reasoning; it might be because contributing through the scheme is considered a bad signal; it might take the form of public opposition and a regulatory crackdown; I don’t know.
But even if individuals aren’t actively motivated to give to charity, I think they are much less likely to coordinate to kill the charitable version. Stopping some opportunist from making a bunch of money is one thing. Stopping money from going to poor Kenyans is something else altogether.
I think there is some money lying on the ground that we could pick up by canceling out zero-sum political contributions.So far we haven’t managed to do it, but I don’t see an in principle obstruction. As far as I can tell, the FEC is fine with someone doing that.
This post discussed one obvious problem with contribution matching, and a possible solution. Like the other schemes I’ve blogged about here, I don’t think this is a high altruistic priority. But I do think it’s interesting to think about. And in contrast with most of the schemes that I talk about, if this looked like a good idea then it’s actually kind of feasible.
6 thoughts on “Repledge++”
I had this thought a while ago, and had the following possible amendment to fix the issue where people who value charity pledge money under false pretenses, to cancel out any actual advantage of either party:
If you get $1 pledged for party A and $1 pledged for party B, then you donate *half* of each dollar to charity, and send the other half along to each party (so GiveDirectly gets $1, A gets $0.50, and B gets $0.50).
Now suppose I’m a malicious charity-donor who doesn’t care about politics, and I know that Clinton is ahead by $100. In the original scheme, I “pledge” $100, causing $200 to go to charity; by doing this instead of donating directly I’ve moved $100 from Clinton to charity. In the new scheme, I could pledge my $100, thereby causing $100 (half of $200) to go to charity … which is not better than just giving my $100 to charity. Of course it also moves $50 from Clinton to Trump, so if I prefer Trump to Clinton then I’m incentivized to do it — but that’s just the system working exactly as intended.
This way people who don’t care about politics don’t screw it up. On the other hand, only half as much money gets sent to charity, so maybe that’s not a worthwhile sacrifice.
The first two problems in your list are almost the same. That is, I as a participant believe that both candidates have different marginal value of a contribution.
This can be solved by a market: I submit one dollar to you for the Clinton capaign, with the stipulation that it may be cancelled by e.g. 50 cent from the Trump capaign. Then see if there are any takers in the market.
I think that the third problem is much more significant. Suppose I want to make a charitable donation, and that I also lightly prefer Clinton. Then I can get a free lunch by using your market: I can give my money against malaria, while also diverting some Trump-money to some christian charity.
In other words, the scheme depends on participants beeing honest about their preferences between donating to charity vs donating to politics, while at the same time incentivizing participants to be dishonest about their preferences.
In principle it seems like the scheme can work even if such cancellation is going on. Suppose that you are a Trump supporter, concerned that someone may be donating to Clinton when really they just wanted the money to go to charity. This may make you annoyed at the system, but it doesn’t change the basic calculus: they are going to donate whether or not you do, and so if you don’t offset their donation then that’s still $2 going to Clinton (or $2 fewer going to Trump).
The key thing is that by giving to Trump you aren’t increasing the probability that a charity-lover donates to the Clinton side. This is why I suggested keeping the totals private until the final reveal, since that seems like it prevents any causal effect of giving to Trump.
Just discovered this blog and loving it.
When would the “big reveal” happen? I don’t know enough about how effective political funding is depending on timing, but would it make a difference whether the money is donated during the last week of a campaign, or in the middle?
Or would there be a series of reveals (ie: every month the amount is calculated and then distributed)?
(Though, this would have the problem of decreasing the proportion of money given to charity, assuming it’s not one candidate winning out over the other.)
How would you suggest implementing this? A smart contract?
Also, moral trade could be implemented in various other ways, like nullifying each other’s votes before an election as discussed in the paper by Toby Ord. How could we implement trust for such protocols such that nobody secretly violates the agreement?