Do we want the market to take on some of the risk? Probably no. Most current funds are ok with risks, as long as the odds are net positive. Scott seems to think that transferring risk to to investors is a bad externality of impacts market.
This is bad both because we don’t want people to lose all their money, and because this might create moral hazard on the part of final oracular funders to recoup some of people’s losses if they seem like an especially pitiful case.
I have not thought at lot about this, but I think I agree with Scott here.
So what we want is a way of tracking and rewarding good predictions. But we already have a solution for that, i.e. prediction markets. Granted that there are still design details to be worked out regarding prediction markets, but
It seems like an easier problem
There are already several experiments to learn from
In a prediction market, good predictors provide information and get rewarded for this. Funders can use the prediction market to guide what they fund. If they choose to fund something risky, some of the risk falls on the predictors (we can’t remove this completely, and I don’t think we want to either) but most of the risk falls on the funder, which I think is what we want.
This also solves some other problems.
Funders can just not fund things with obvious large downside risk.
Investor can’t capture almost all the value by just being fast at buying obvious opportunities.
Although I don’t think 2 will be a big problem even in a impact market. If it is an obvious good idea the people who wants to run the projects can just not go though the markets and instead reach out to a funder directly.
predictor = person who invest in prediction market investor = person who invest in impacts market funder = person or institution who want’s to fund altruism, with no expectation of financial return.
I’m I missing something that we want from a impacts market that we can’t get from a prediction market?
(I’ve written about impact purchase before. However the thing I want out of such a system is completely removed from Scott’s suggestion, and vise versa, so should probably not count as the same type of system.)
I think there were already prediction markets for future grants at the time when I did my first research into impact markets. Maybe they still exist.
The risk and prediction aspects that you mention factor somewhat into my value proposition for impact markets, but I’m not sure how important they are or generally how I feel about them.
My/our (impactmarkets.io) thinking is rather what I mentioned in the comment above, that we want to (1) radically reduce the time cost borne by funders, (2) expand the hiring pool for funders, and (3) enlist all the big networks of angel investors and impact investment funds in the search for the best funding opportunities. My thinking about the relative emphasis changes from time to time here as I become aware of new considerations. The risk angle could also make the top list, but of course only with fully informed investors who know exactly what they’re doing.
The biggest factor in my mind is currently the third one, followed by the first one. A quick Guesstimate model says that the third factor could 20x (5–85x) the accessible funding opportunities (access to thousands more social circles worldwide of people who are great at networking), and if hits-based funders currently aim for funding 10% successful projects, that translates to a reduction of about 10x in time cost.
I’m having a hard time guessing how the hiring pool might change. It will probably get vastly greater once grantmakers have to know only the priorities research and don’t need to be great judges of character and project-specific skill of others too, but funders might choose not to scale to absorb all of them.
I don’t know enough to exclude that this could be done with prediction markets, but to get angel investors and impact investment funds only prediction markets, they would have to involve real money, and that causes lots of legal hurdles. But even if we clear those, they would only put their money in them if they can expect the risky equivalent of a riskless 10–30% APY. I’m a bit hazy on how the initial liquidity gets onto new prediction markets that people create (e.g., on Polymarket or Augur) and how the initial value of Yes and No are set, so it’ll take me a lot more learning before I can repeat the profitability math for prediction markets. Also I don’t know yet how the markets would be resolved.
It seems all weird when I try to play through how this could work: You could establish the norm that researchers create markets on Augur that resolve positively if any of some set of retro funders endorse them. Then the researchers buy Yes. Then their impact investor friends buy Yes. Then the researchers sell out of some of their positions again to pay their rent etc. Then, if the project is successful, the retro funder comes in, buys just the right amount of No, and resolves the market to Yes. And by “just the right amount” I mean the sort of amount that just incentivizes investors to keep investing into projects like that without wasting money beyond that point. But that’s such a weird use of prediction markets… (On second thought, that wouldn’t work at all because there would not be enough liquidity on the Yes side for the researcher and the investor to buy it, right?)
Maybe someone with deeper knowledge of prediction markets can come up with a system that works well and has all the advantages though!
Update: Since there are maybe people here with more knowledge of prediction markets, it’s probably more productive to phrase this as a question: Would it be possible to create an Augur-like system where:
altruistic funders can provide funds, a “prize pool” of sorts, with only a one-time effort,
researchers can create markets at a low cost that try to predict whether any of a set of funders will endorse their research project and its results,
investors can bet money on such endorsements,
researchers can sell into those buys from investors to fund their project,
funders can eventually resolve the markets with their endorsements and will thereby reward investor and researcher to just the right extent, and
funders have no costs if markets don’t resolve or somehow resolve to no?
An alternative that we’ve been toying with are reverse charity fundraisers of sorts. You do your thing, and when you’re done, you publish it, and then there’s a reward button where anyone can reward you for it. “Your thing” can be doing research, funding research, copyediting research, etc.
I love the simplicity of it, but there are a few worries that we have when it comes to incentives for collaboration when participants have different levels of social influence. Still, it’s a very promising model in my mind.
Although after I wrote this post I updated towards this not being a good idea for anyone to get their income from this system long term. But I still think it would be a good alternative to other funding systems for new EAs. Retrospective funding have definite disadvantages, but for someone with out enough reputation, it may be better than the available alternatives.
Indeed! I think this transition from impact markets to other sources of funding can happen quite naturally. A new, unknown researcher may enjoy the confidence in her abilities of some close friends but has little to show that would convince major funders that she can do high-quality research. But once she has used impact markets to fund her first few high-quality pieces of research, she will have a good track record to show, and can plausibly access other sources of funding. Then she can choose between them freely, is not dependent on impact markets alone anymore.
What do we want from impacts markets?
Mostly, we want the prediction power of markets.
Do we want the market to take on some of the risk? Probably no. Most current funds are ok with risks, as long as the odds are net positive. Scott seems to think that transferring risk to to investors is a bad externality of impacts market.
I have not thought at lot about this, but I think I agree with Scott here.
So what we want is a way of tracking and rewarding good predictions. But we already have a solution for that, i.e. prediction markets. Granted that there are still design details to be worked out regarding prediction markets, but
It seems like an easier problem
There are already several experiments to learn from
In a prediction market, good predictors provide information and get rewarded for this. Funders can use the prediction market to guide what they fund. If they choose to fund something risky, some of the risk falls on the predictors (we can’t remove this completely, and I don’t think we want to either) but most of the risk falls on the funder, which I think is what we want.
This also solves some other problems.
Funders can just not fund things with obvious large downside risk.
Investor can’t capture almost all the value by just being fast at buying obvious opportunities.
Although I don’t think 2 will be a big problem even in a impact market. If it is an obvious good idea the people who wants to run the projects can just not go though the markets and instead reach out to a funder directly.
predictor = person who invest in prediction market
investor = person who invest in impacts market
funder = person or institution who want’s to fund altruism, with no expectation of financial return.
I’m I missing something that we want from a impacts market that we can’t get from a prediction market?
(I’ve written about impact purchase before. However the thing I want out of such a system is completely removed from Scott’s suggestion, and vise versa, so should probably not count as the same type of system.)
I think there were already prediction markets for future grants at the time when I did my first research into impact markets. Maybe they still exist.
The risk and prediction aspects that you mention factor somewhat into my value proposition for impact markets, but I’m not sure how important they are or generally how I feel about them.
My/our (impactmarkets.io) thinking is rather what I mentioned in the comment above, that we want to (1) radically reduce the time cost borne by funders, (2) expand the hiring pool for funders, and (3) enlist all the big networks of angel investors and impact investment funds in the search for the best funding opportunities. My thinking about the relative emphasis changes from time to time here as I become aware of new considerations. The risk angle could also make the top list, but of course only with fully informed investors who know exactly what they’re doing.
The biggest factor in my mind is currently the third one, followed by the first one. A quick Guesstimate model says that the third factor could 20x (5–85x) the accessible funding opportunities (access to thousands more social circles worldwide of people who are great at networking), and if hits-based funders currently aim for funding 10% successful projects, that translates to a reduction of about 10x in time cost.
I’m having a hard time guessing how the hiring pool might change. It will probably get vastly greater once grantmakers have to know only the priorities research and don’t need to be great judges of character and project-specific skill of others too, but funders might choose not to scale to absorb all of them.
I don’t know enough to exclude that this could be done with prediction markets, but to get angel investors and impact investment funds only prediction markets, they would have to involve real money, and that causes lots of legal hurdles. But even if we clear those, they would only put their money in them if they can expect the risky equivalent of a riskless 10–30% APY. I’m a bit hazy on how the initial liquidity gets onto new prediction markets that people create (e.g., on Polymarket or Augur) and how the initial value of Yes and No are set, so it’ll take me a lot more learning before I can repeat the profitability math for prediction markets. Also I don’t know yet how the markets would be resolved.
It seems all weird when I try to play through how this could work: You could establish the norm that researchers create markets on Augur that resolve positively if any of some set of retro funders endorse them. Then the researchers buy Yes. Then their impact investor friends buy Yes. Then the researchers sell out of some of their positions again to pay their rent etc. Then, if the project is successful, the retro funder comes in, buys just the right amount of No, and resolves the market to Yes. And by “just the right amount” I mean the sort of amount that just incentivizes investors to keep investing into projects like that without wasting money beyond that point. But that’s such a weird use of prediction markets… (On second thought, that wouldn’t work at all because there would not be enough liquidity on the Yes side for the researcher and the investor to buy it, right?)Maybe someone with deeper knowledge of prediction markets can come up with a system that works well and has all the advantages though!
Update: Since there are maybe people here with more knowledge of prediction markets, it’s probably more productive to phrase this as a question: Would it be possible to create an Augur-like system where:
altruistic funders can provide funds, a “prize pool” of sorts, with only a one-time effort,
researchers can create markets at a low cost that try to predict whether any of a set of funders will endorse their research project and its results,
investors can bet money on such endorsements,
researchers can sell into those buys from investors to fund their project,
funders can eventually resolve the markets with their endorsements and will thereby reward investor and researcher to just the right extent, and
funders have no costs if markets don’t resolve or somehow resolve to no?
An alternative that we’ve been toying with are reverse charity fundraisers of sorts. You do your thing, and when you’re done, you publish it, and then there’s a reward button where anyone can reward you for it. “Your thing” can be doing research, funding research, copyediting research, etc.
I love the simplicity of it, but there are a few worries that we have when it comes to incentives for collaboration when participants have different levels of social influence. Still, it’s a very promising model in my mind.
I would very much want there to be a “money after the project” funding system for smaller projects in EA.
https://forum.effectivealtruism.org/posts/7iptwuSyzDzxsEY5z/the-case-for-impact-purchase-or-part-1
Although after I wrote this post I updated towards this not being a good idea for anyone to get their income from this system long term. But I still think it would be a good alternative to other funding systems for new EAs. Retrospective funding have definite disadvantages, but for someone with out enough reputation, it may be better than the available alternatives.
Indeed! I think this transition from impact markets to other sources of funding can happen quite naturally. A new, unknown researcher may enjoy the confidence in her abilities of some close friends but has little to show that would convince major funders that she can do high-quality research. But once she has used impact markets to fund her first few high-quality pieces of research, she will have a good track record to show, and can plausibly access other sources of funding. Then she can choose between them freely, is not dependent on impact markets alone anymore.
Someone who is exited about impact markets should do Goal Factoring on your preferred version of impact markets.