6

Donation insurance

This post is from Linchuan Zhang, who asked me to post this on his behalf. You can contact him at email.linch@gmail.com to discuss more. He'll also read the comments here. You can also collaborate with him on the original document on HackPad here.

Need

It is generally understood within the effective altruism community that people make suboptimal decisions when they do not have financial security, Further, this lack of option value is sufficiently harmful that many prominent EAs argue for having a "runway," ie, enough personal savings to handle the risks that come along as a natural part of life. See, for instance, this 80000 Hours post. However, since it's often claimed (though not definitively proven/no consensus) that there's a significant "haste consideration" in donating now instead of later, the opportunity costs of holding large amounts of liquid money on hand is nontrivial. One possible solution is "a community insurance scheme", outlined in the bottom of the article above. However, this "scheme has a serious challenge – who decides whether someone really deserved the insurance payout, and who is freeloading – but is potentially much more efficient."

I think we can improve on this.

Proposed Solution

I propose the following: Specific donations are insured in a one-directional, two-party contract. 

Here's how it works: Person A is debating how much of her $18,000 to donate to the Against Malaria Foundation. She originally wants to save $9,000 in the bank (enough to last her around 6 months). Hearing this, Person B suggest that A only saves 4000 dollars in the bank, B is willing to insure donations to the AMF up to 6,000 for 12 months. B specifies a list of conditions that he's willing to insure A for (including termination of employment, voluntary or otherwise, significant physical or personal illness, death in a family, etc), wherein he'll give A $6,000 back within 30 days of any of the triggers happening. A agrees, and both the agreement and A's donation is publicly logged online.

If none of the conditions trigger in the next 12 months, then Person B managed to help Person A help the AMF without any cost to himself! Yay! 

If any of those conditions trigger in the next 12 months, A can choose to either i)not get insurance because she doesn't need it, or, ii)as is more likely the case, she will contact B (whether/what forms of proof are necessary depends on A/B's relationship with each other). B actually only has $4000 in savings, but he gets paid $3,000 after taxes/personal consumption per month. B waits until he gets paid, and then he gives A $6000, leaving 1,000 in the bank. A is glad to have the insurance. B still needs to tough out 1 month with low savings, but the statistical odds of both A and B being in financial trouble is far lower than A alone.  

One Caveat: The same donation cannot be insured by two different parties, to prevent potential abuse. 
Also, If A insures B, B should not insure A within the same time period. 

Benefits 

  • Almost certain to have greater total donations (at least among EAs). Donation insurers have an easy way to increase donations on expectation
  • Helps with community.
  • People can feel safer and are thus more likely to take risks.
  • People who donate a lot wouldn't be SOL or be forced to rely on the generosity of family, banks or friends. 

Potential Risks

  • Abuse : Risk of traditional abuse/moral hazard substantially lower than a traditional community insurance scheme (because you aren't actually gaining any additional money, you're just getting your donation money back). 
  • Counterparty Risks: It's possible that B insures A, but B also took a hit to his financial position. However, B can insure his donations too. The fact that these insurance policies are always 2-party and single-dimensional reduces this greatly.
  • Naive EV utilitarians may take advantage of this to claim donation insurance for superficial problems just so they can donate more...hopefully we don't have so many dicks in the movement though. 
  • Obviously if you believe in a "dishaste consideration", you should not promote easier ways to donate now as opposed to later.


Comments (7)

Comment author: Gleb_T  (EA Profile) 20 December 2015 11:16:04PM 1 point [-]

Nice idea! Two optimizations:

1) We can also set up a community pool to insure person A in case person B turns out to run into financial issues at the same time. A committee can take charge of this community pool and pay out person A or person B depending on the situation. Then, person A or B can pay money back into the community pool.

2) For the problem of dicks in the movement, we can have mediators - person C - that person B can turn to in case s/he suspects superficial problems.

Comment author: Linch 21 December 2015 11:16:51PM 1 point [-]

Thanks for the comment!

1) Please see my reply to kbog for some reasons why I think a donation pool is less feasible in the short run than 2-party donation insurance.

2) I think that's a possible solution. I also think the number of naive EV utilitarians is vastly overestimated (it's quite probable the number is under 10, and quite possible it's actually 0). Alternatively people can just trust fellow aspiring effective altruists well enough to not renege on the spirit of an agreement.

Comment author: Gleb_T  (EA Profile) 21 December 2015 11:35:22PM 0 points [-]

For 1, I read your comment below, makes sense to start small and expand :-)

For 2, I think you'd get more people involved if they overcome the emotional fears of naive EV utilitarians through some sort of mediator.

Comment author: Lila 26 December 2015 08:10:16PM 0 points [-]

Why don't people just get typical forms of insurance, like health, disability, and unemployment?

Comment author: Linch 01 January 2016 08:29:26PM *  1 point [-]

Because they're more expensive?

Standard EV calculations tells you that (given risk neutrality) insurance is a loser's game, assuming that you don't have special information that the insurers do not.

Comment author: kbog  (EA Profile) 21 December 2015 09:03:08PM *  0 points [-]

Well this would work but if people know each other well enough to trust each other, but why can't B just give friendly loans or assistance without worrying about official terms and contracts?

You're establishing small risk pools to manage risk, but you can do that MUCH more efficiently with a large, centralized risk pool.

Comment author: Linch 21 December 2015 09:35:42PM *  1 point [-]

Hi again kbog,

We've discussed this a bit on Facebook but I think I want to leave my thought process in the open.

I think it is very awkward to ask people for loans, and also very emotionally/"morally" fraught to plan for the possibility that you may need to ask other EAs for loans or assistance. Eventually we can probably help to combat this by creating social norms where it becomes more acceptable to do so, but I think it's a nontrivial cultural shift.

In contrast, suggesting a new and cool plan (with the donation serving as a VERY credible mitigation of potential adverse selection) might sidestep a lot of cultural and other emotional concerns. I was not thinking of very official terms and contracts (which can be expensive in terms of transaction costs), but just something more formalized than completely informal agreements between friends (eg. public pre-agreements, akin to public Bayesian bets or GWWC's Try Giving).

Getting people to precommit to helping each other seems like it can plausibly lead to more short run donations and lower average savings base than a very vague understanding that your friends can pitch in eventually.

(This all assumes at least a weak belief in some variant of the haste consideration, of course)

I agree that there are strong advantages of a very large centralized risk pool. There are some disadvantages though, which we've talked about in the PM and I will cover in more detail after I rewrite this post.

However, the biggest difference is that I think take-up of 2-party donation insurance by early adopters can happen NOW (in a sense just formalizing the "friendly loans or assistance" that some EAs already sort of have anyway). If there's enough takeup and evidence in favor of increased donations, this can morph into more advanced risk pooling systems if it's advantageous, and if Knightian uncertainty kicks in and it flops, the downside risk to our community or image is much lower than if a centralized "Effective altruism" organization is running it.

So I think a 2-party donation insurance scheme is both something we can have right NOW, something that I believe with over 90% confidence better than the entirely informal arrangements the community currently has, and something that can fail gracefully if it ends up failing.