[very quick technical-ish thoughts on a necessary condition] I would guess funding retrospectively today is only useful if you think you're going to increase funding in the future. 

The purpose of funding is to enable people to complete projects -- more people are willing to do a project with more funding, and only really rare, super-duper value-aligned people will do a project with very little funding. 

Retrospectively funding people that are already doing the project ignores they've already signaled they're the type that is happy to do the project for low funding. But, like you said, such funding also signals to future researchers on the margin that their expected funding is higher than whatever it is right now. 

So, it's a way for orgs like OpenPhil to smooth researcher consumption over time if they expect they themselves will be getting richer in the future (and thus be able to increase funding for their projects). Retrospective funding should be set at the transfer required in expectation for the marginal researcher who brings in extra value equal to the cost of retrospective funding (MB = MC)? This argument is in expectation and is likely less effective for more risk-averse people -- if OpenPhil has the money now they should probably just increase upfront funding. 

This also relies on other things, like OpenPhil being seen as trustworthy enough to keep up the scheme of retrospective funding, of the signal of retrospective funding being sufficiently public, etc. 

(On the run now, but I see people linked some stuff in the comments i'll check out soon. If all of what I said is common knowledge already, sorry!!)

I was thinking about this problem a lot at the beginning of the summer, and I caught myself smiling a ton while reading this at the deja vu! I think this question of moral trade on political goods is fascinating, and I definitely agree with the premise: that spending in politics is much higher than is socially optimal because there is an ‘arms race’ element to political fundraising. Few thoughts:

• I think the question of political identity verification (i.e., the non-political altruist is happy to lie, pretend like they’re Democrat, and match with Republicans to get double the cash going to GiveWell) is pretty serious. The proposed fix you gave on your blog -- with half of the donations from both sides being locked into their political party -- incurs a huge efficiency loss with 50%. Maybe this is just the best we can do… But I think this problem should get a lot of attention. More on that later.
• Things like past political donations, proof of party affiliation, etc., may be useful especially in the beginning phase of the market when convincing people that the platform is legit is particularly important. How to use this in tandem with whatever incentive mechanism is used is worth thinking about, as requiring these types of documents also decreases market thickness.
• An important design question (as others have pointed out) is what ratio the political dollars should be matched (i.e. Democrats have raised double the amount as Republicans, so every 2 Democrat Dollars match with 1 Republican Dollars). You indicated this won’t be a problem in your blog because half of all political campaigns end up being reasonably close in total amounts raised. I think this is a bigger problem -- the last three elections (in which campaign spending has changed dramatically due to some rule changes) had very lopsided total amounts raised. That being said, you’ve done a lot of great work on this already in the write-up you put on your blog. One compromise which balances opacity and setting ρ optimally could just be letting ρ = the current ratio of dollars raised from both sides. There are still endogeneity concerns, but maybe it’s close enough -- I think it's also worth doing some simulations on past campaign histories to see how whether this rate is relatively stable. An alternate idea is to implement an auction scheme to decide the optimal ratio for matching. This may actually work perfectly and be reasonably easy to follow -- I’d need to think about it a bit more.
• When modeling this, I thought of political dollars for Democrats and Republicans as a positional good, similar to how you did. Fred Hirsch has done some work on this, and his model may be helpful.
• Finally, while trying to figure out the verification problem, I thought up a loonshot idea which may end up capturing even more benefit for the EA community in the longterm. Idea is to introduce a second market for altruists alongside the political de-escalation market. Here’s how it would work. Let’s say I think the Against Malaria Foundation is 100x more effective than the Rotary Foundation. I can then offer a 100:1 match (up to maybe $1000, because I’m not made of money) on changing people’s donations from the Rotary Foundation to the Against Malaria Foundation. This could, in general, be huge for the altruistic community -- it allows EAs to compel other altruists to donate as if they are also EAs. There are three important points relative to the political market, as well: 1) As long as the market is thick, this solves the verification problem with non-political altruists in the political market, because the apolitical altruist can now get a better multiplier on their donation in the altruists market. 2) I think the people most likely to use the platform (especially in the beginning) are highly altruistic people who are also political. This means they’re probably donating much more to altruistic causes than they are to political ones, so this market will probably get thick quicker. 3) The altruistic market can stand independently from the more tenuous political one, acting as a proof-of-concept. If it works, then we know the matching market works well in theory and that we don’t have the verification roadblock we had before. If it doesn’t work, then at least we didn’t lose something like 250k trying to make it work. (This last bit also works in the reverse; if the altruistic market works, it probably makes it easier to fundraise for the political market.) Of course, this problem is maybe even more complicated, so maybe this is just substituting a hard problem with an even harder one.

I'm a 4th year Math/Econ undergrad at UChicago also super interested in mechanism design and was thinking about working out moral trade markets for my thesis. I sent you an email to collaborate if you're interested!

FEAT (Forecasting EA Tournament)

The Delphi Tournament

There are two names that we're in between for the tournament. Please upvote the name that you like more.