Ben_Todd comments on EA Funds - An update from CEA - Effective Altruism Forum

You are viewing a comment permalink. View the original post to see all comments and the full post content.

Comments (21)

You are viewing a single comment's thread. Show more comments above.

Comment author: Ben_Todd 08 August 2018 06:04:37AM 4 points [-]

I agree that would be ideal, but it doesn't seem like a high priority feature. The risk-free 1yr interest rate is about 2% at the minute (in treasuries), so even if the money is delayed for a whole year, we're only talking about a gain of 2%, and probably more like 1% after transaction costs.

You could invest in the stock market instead, but the expected return is still probably only 1-5% per year (as I argue here: https://80000hours.org/2015/10/common-investing-mistakes-in-the-effective-altruism-community/). Plus, then you have a major risk of losing lots of the money, which will probably be pretty hard to explain to many of the users, the press etc.

I expect the staff time spent adding and managing this feature could yield much more than a couple of percent growth to the impact of the funds in many other ways (e.g. the features Marek lists above).

Comment author: vollmer 08 August 2018 03:27:57PM *  4 points [-]

The sources you quote seem to suggest more like 5% in real annual returns (or 7% nominal), and you wrote "2-7% nominal returns". If you're investing $2m, that would be $40k-$140k per year. I'd expect this to cost maybe one week of staff time per year, so it might easily be worth the cost. (Mission hedging and more diversification would push this up further; fees and risk aversion would push it down. Overall I don't expect these factors to be very strong though.)

To me it seems that the difficulty of explaining this to the users is the stronger reason against implementing this. (Unless the users themselves can choose but that would cost more staff time again.)

Comment author: Ben_Todd 09 August 2018 03:20:14AM 2 points [-]

Sorry, I was thinking 1-5% of real returns rather than nominal, though I agree for these purposes nominal might be more relevant.

There's a lot of room for different figures depending on how mean reverting you think valuations are. I think we should expect them to be somewhat mean reverting, so my best guess is more like 3% real.

I was also partly thinking that valuations are much higher again since I wrote that post, so I think that article is a bit optimistic today. http://www.multpl.com/shiller-pe/

With costs, I'm more thinking about the set-up cost. I wouldn't be surprised if this took several person-months, which could be used to add features that would add much larger proportional gains.

I also guess the on-going costs would be quite a bit more than one week per year, due to the reasons Rob lists.

And then there are also the costs of explaining this feature to users, which seem pretty significant - e.g. even if you write up a clear explanation of why you're doing this and speak to a bunch of people about it in-person, you'll still end up with lots of misunderstandings.

Comment author: RyanCarey 08 August 2018 06:30:30AM 3 points [-]

Agreed. You could get a higher effective ROI by mission-hedging -- investing AI-risk funds in things like Google. But even then, the returns seem like a pretty second-order issue.

Comment author: kbog  (EA Profile) 08 August 2018 05:06:31PM *  0 points [-]

The benefits from mission hedging are far lower. If your stocks go up 10%, you get 10% more money. If Google stock goes up 10%, then money spent on AI risk goes up in importance by, I don't know, 0.5%, so you now have 10.5% more impact.

Comment author: kbog  (EA Profile) 08 August 2018 05:08:20PM *  1 point [-]

The S&P500 has returned an average of >9% per year for the last 90 years.

How can you say "past returns are no guarantee of future returns" in response to this fact, then turn around and estimate future returns based on a snapshot of what yield and buyback are "right now"? The latter is much less reliable.

It takes about a minute to put money into a fund and another minute to sell it. Seems pretty simple to me.

Comment author: Robert_Wiblin 08 August 2018 09:29:35PM 10 points [-]

The effort required to set up a non-profit trading account, go through KYC, make it secure, teach everyone in the org how it works, and do the necessary legal, budgetary and accounting compliance each year make this much more than a few minute job. Things that are easy for individuals are often less straightforward for organisations.