Certificates of Impact [Paul Christiano; Link]

post by RyanCarey · 2014-11-11T16:58:13.337Z · LW · GW · Legacy · 9 comments

Paul proposes that we could create a market for certificates of impact. The certificates would be created whenever someone does something that has a positive impact in the world.

9 comments

Comments sorted by top scores.

comment by IlyaShpitser · 2014-11-11T21:03:37.957Z · LW(p) · GW(p)

I sometimes wonder if Zack Wienersmith secretly follows LW as a topic mine.

comment by SanguineEmpiricist · 2014-11-20T02:14:05.644Z · LW(p) · GW(p)

What about removing bad? This tendency towards overt positive forms and nothing talking about removing bad/evil is suspicious.

comment by solipsist · 2014-11-12T00:02:43.402Z · LW(p) · GW(p)

Problems of wealthy possible futures will be valued over problems of poor possible futures. The future philanthropy Hedonism Bots for People Who Can Afford Only 7 Hedonism Bots will pay billions for a Certificate of slightly improving the Orgasmotron. The future philanthropy Zombie Control Initiative can barely afford to give a holder of the Certificate of Inventing the Vaccine against the Z-virus one of the last remaining can of spam. Current market prices of those two certificates will reflect their future value.

EDIT: I now think the thesis of this comment is incorrect

Replies from: solipsist, paulfchristiano
comment by solipsist · 2014-11-12T00:26:36.793Z · LW(p) · GW(p)

I did my math wrong. The present value of the billions paid for a Certificate of slightly improving the Orgasmotron is less than the present value of the can of spam, because you have to value these securities against the rest of your portfolio. You will use a high discount rate for the hedonism world, and a negative discount rate for the zombie apocalypse world. (At least that's how I think of it)

comment by paulfchristiano · 2014-11-12T00:32:06.172Z · LW(p) · GW(p)

I don't think this is true. Suppose world A is rich and world B is poor. People want resources more in world B, so they are not willing to pay as many resources for a certificate. But on the flip side, a philanthropist is happy to accept fewer resources. That is, a resource that pays $1 (real) in world B, costs more than a resource that pays $1 (real) in world A, because it's catastrophe insurance. This seems to exactly offset the decreased willingness to pay of people in world B.

(Of course, wealthier worlds might also devote a larger share of their resources to philanthropy, and so for that reason might be better to support, though by a much smaller factor. This proposal allows such worlds to reap benefits from their increased philanthropic spending, while poorer worlds can't do so. We could prevent this inequality by leveling down, but I don't see why we should.)

comment by MaximumLiberty · 2014-11-11T19:07:22.033Z · LW(p) · GW(p)

I don't really want to create an account on yet another website, so I'll comment here. Anyone with a login there should feel free to copy and cross post.

It is a thought-provoking idea, but there seem to be serious problems with the model underling the proposal. The model fails to distinguish between flows and stocks.

Then at equilibrium, the price of certificates of impact on X is equal to the marginal cost of achieving an impact on X. Any deviation is an arbitrage opportunity: if you can do X more cheaply, then you can sell the resulting certificates for a profit; if you are doing X more expensively, then you could save money by buying certificates instead.

This is a faulty analogy to product markets, and is true only if certificates are consumed. They are not.

The better analogy would be to the market for money, where the demand for money is greatly affected by the public's desire to hold cash. We see hyperinflation (i.e. a rapid decline in the price of money) where the public believes that money is becoming worthless. When that happens, what used to be extra-marginal stock becomes infra-marginal flow. A similar crisis of confidence could occur for certificates of impact. In fact, that is exactly what has happened to carbon credit schemes.

One way to make the model look more like a product model is to expire the certificates. They are only good for a year, or they all expire on December 31, or something like that.

A related point is how to validate that the activity claimed to have an impact actually occurred. This is a systemic risk similar to the problem of assigning credit ratings to collateralized debt obligations. Someone has to say whether or not the activity occurred. If the person saying so is revealed to have exercised insufficient diligence, then the market has a crisis of confidence.

Replies from: paulfchristiano
comment by paulfchristiano · 2014-11-11T22:34:56.282Z · LW(p) · GW(p)

This is a faulty analogy to product markets, and is true only if certificates are consumed. They are not.

I don't quite understand this. Are you objecting to the quoted argument?

If X is being bought and sold for $1, and you can make X for less than $1, then you can make X and sell it, at a profit. Where does this assume that X is being consumed vs. saved?

Similarly, if X is being bought and sold for $1, you would be better off buying a unit for $1 than making it yourself for a higher price. This is true regardless of what you plan to do with X.

The better analogy would be to the market for money, where the demand for money is greatly affected by the public's desire to hold cash.

Yes, the demand for certificates is entirely determined by funders' desire to hold them. This analogy doesn't seem to undermine the proposal. If philanthropists decided they didn't value certificates any more, the price would drop. But a philanthropist shouldn't even be troubled by being stuck with a certificate they can't sell, it's just the same as having made a grant. The money is gone, but the good is done.

A related point is how to validate that the activity claimed to have an impact actually occurred. This is a systemic risk similar to the problem of assigning credit ratings to collateralized debt obligations. Someone has to say whether or not the activity occurred. If the person saying so is revealed to have exercised insufficient diligence, then the market has a crisis of confidence.

The same problem is faced by someone making a grant or offering a prize, and it should be addressed in the same way. For example, GiveWell can do exactly the same thing they currently do, but now it is valuing certificates rather than recommending donations. Do you think there is a material distinction between these cases?

Someone might purchase a certificate based on GiveWell's recommendations and later discover that they are bogus. But exactly the same thing could happen today, with someone donating based on GiveWell's recommendations and later discovering that they are bogus.

Replies from: MaximumLiberty
comment by MaximumLiberty · 2014-11-12T23:38:47.687Z · LW(p) · GW(p)

I don't quite understand this. Are you objecting to the quoted argument?

Yes, but I should have been more specific, and I think I probably focused overly much on the sentence about deviation and arbitrage. (I'm blaming jet lag.)

I agree that, if someone can produce certificates at less than what they perceive as a stable-ish price, they will produce more than the price will eventually fall. What I disagree with is the idea (familiar from product markets) that deviation will induce arbitrage (i.e. that an increase in price will induce greater supply of new activities that result in certificates). This is generally untrue in the short term for markets where there are large stocks being held. Money is a classic example; during some periods, precious metals and diamonds were as well. The stocks act as a shock absorber.

The mechanism is that a price increase will induce holders of stocks to sell them. It won't induce new production until the holders of stock willing to sell at a slightly higher price are exhausted. Of course, this is conditioned on certificates being like these things where people hold them as stocks for later resale, rather than "consuming" them. If the certificates time out, then that would be the equivalent of them being consumed. Likewise if your prediction is right that there are a group of end-purchasers who never resell them.

(In financial markets, market-making intermediaries hold large stocks of financial instruments. I would imagine that a certificates market would need a similar group of market-makers to be liquid. Ideally, this can be a good thing, because the intermediaries can smooth out short term price fluctuations and send a more stable signal to organizations. But, on top of the intermediaries, you have the organizations that hold certificates for the long term.)

Additionally, I was focusing on risk to the system. One systemic risk is that if a holder of certificates runs into financial trouble (which happens to non-profits), they may be forced to dump a large number of certificates on the market.

The validity issue is similar. I'm not concerned if a specific certificate turned out to be false. What would be a systemic risk is the perception that many certificates are false. That would lead holders of certificates to dump them on the market, so that they can reinvest their philanthropic capital into something worthy. (Remember that trustees have a fiduciary duty to protect the finances of the charity, so they can't just shrug and live with it like you or I might.)

I think there is a material distinction between the situations you describe. GiveWell stands behind a very few efforts at one time. If I want to give, I can easily do a bit of research myself about the organizations that they endorse. If I understand the certificates proposal proposal, the core point is to encompass just about anyone who wants to claim that they did something good. Presumably, that would give rise to organizations that validate those claims. But that seems a quite different undertaking from what GiveWell does today: it would require an organizational structure for mass production. If money is involved, then you get commercial pressures on standards; if money is not involved, you still get political and social pressures. And, if it turns out that the certificate of impact "ratings agencies" were sloppy or used lax standards, then the market suddenly comes to believe that it is holding certificates that don't represent the activities that the holders were organized to promote. If I was on the board of trustees, I'd dump the certificates and redeploy what philanthropic capital I can salvage. When I do that, and others see the price falling, they might follow suit. And then there is the impact on the organizations that had planned activities on the assumption that they could sell the resulting certificates on the market to cover their costs.

One way to eliminate all of these risks is to make it so that people who buy to hold can't resell. Then it is much, much more like a grant.

Replies from: paulfchristiano
comment by paulfchristiano · 2014-11-13T00:35:25.417Z · LW(p) · GW(p)

What I disagree with is the idea (familiar from product markets) that deviation will induce arbitrage (i.e. that an increase in price will induce greater supply of new activities that result in certificates). This is generally untrue in the short term for markets where there are large stocks being held. Money is a classic example; during some periods, precious metals and diamonds were as well. The stocks act as a shock absorber.

Why is it important that deviations in price induce production? It seems like the current behavior is the efficient one.

The validity issue is similar. I'm not concerned if a specific certificate turned out to be false. What would be a systemic risk is the perception that many certificates are false. That would lead holders of certificates to dump them on the market, so that they can reinvest their philanthropic capital into something worthy. (Remember that trustees have a fiduciary duty to protect the finances of the charity, so they can't just shrug and live with it like you or I might.)

Again, this seems like a feature to me, as far as it goes. The risk of collapse should be factored into the price of a certificate or the desirability of making a grant. And if philanthropists think a fall is not justified, they should be happy to continue holding (and buying more aggressively as prices fall); I suspect if anything that the prices are a (very slightly) better medium for figuring out whether everything is bogus, than a public debate without prices.

Without leverage or diminishing marginal returns for certificate-holders, I don't see why a systemic risk is any worse than a bunch of idiosyncratic risks. The only novel characteristic seems to be price signals letting a crisis of confidence spread faster, which seems like a special case of letting information spread efficiently and encouraging donors to act on transparent information rather than any defensible guess. I don't yet see the reason to think this is bad on net.

Independently of the system used, if people no longer thought charities were doing good work, it would be hard for them to raise money. If people were right, that would be good, if they were wrong, it would be bad. So the question (in this case) seems to be whether we think the system is better or worse than the status quo for getting the right answer.