How many EA billionaires five years from now?

(Addendum 2023-02-04: Given that Sam Bankman-Fried and Gary Wang appear to have committed serious fraud at FTX, and as a consequence have lost most of their fortunes, it would appear EA is at minus two billionaires since I wrote this. I haven't thought much on how this changes the forecast, but here's a rough guess. First, as I mention in a footnote, the model predicts the number of additional billionaires, meaning the net change in EA billionaires from mid-2022 to mid-2027. So the original prediction of 3.5 additional ones would now need 5.5 new billionaires to come true. That seems improbable. I now think ~2 additional billionaires (90% CI: -1 to 7) between now and mid-2027 seems about right, enough to make it ±0 since August 2022. That is because (1) I'm less optimistic about EAs' money-making abilities, (2) I'm less optimistic about EA membership growth and (3) I think it's more likely now than before that wealthy EA donors stay anonymous or dissociate from EA. But time will tell.)

Dwarkesh Patel argues that "there will be many more effective altruist billionaires". He gives three reasons for thinking so:

1. People who seek glory will be drawn to ambitious and prestigious effective altruist projects. One such project is making a ton of money in order to donate it to effective causes.
2. Effective altruist wealth creation is a kind of default choice for "young, risk-neutral, ambitious, pro-social tech nerds", i.e. people who are likelier than usual to become very wealthy. Effective altruists are more risk-tolerant by default, since you don't get diminishing returns on larger donations the same way you do on increased personal consumption.
3. These early-stage businesses will be able to recruit talented effective altruists, who will be unusually aligned with the business's objectives. That's because if the business is successful, even if you as an employee don't cash out personally, you're still having an impact (either because the business's profits are channelled to good causes, as with FTX, or because the business's mission is itself good, as with Wave).

The post itself is kind of fuzzy on what "many" means or which time period it's concerned with, but in a follow-up comment Patel mentions having made an even-odds bet to the effect that there'll be ≥10 new effective altruist billionaires in the next five years. He also created a Manifold Markets question which puts the probability at 38% as I write this. (A similar question on whether there'll be ≥1 new, non-crypto, non-inheritance effective altruist billionaire in 2031 is currently at 79% which seems noticeably more pessimistic.) I commend Patel for putting his money where his mouth is!

Summary

With (I believe) moderate assumptions and a simple model, I predict 3.5 additional effective altruist billionaires in 2027. With more optimistic assumptions, I predict 6.0 additional billionaires. ≥10 additional effective altruist billionaires in the next five years seems improbable. I present these results and the assumptions that produced them and then speculate haphazardly.

Assumptions

If we want to predict how many effective altruist billionaires there will be in 2027, we should attend to base rates.

As far as I know, there are five or six effective altruists billionaires right now, depending on how you count. They are Jaan Tallinn (Skype), Dustin Moskovitz (Facebook), Sam Bankman-Fried (FTX), Gary Wang (FTX) and one unknown person doing earning to give. We could also count Cari Tuna (Dustin Moskovitz's wife and cofounder of Open Philanthropy). It's possible that someone else from FTX is also an effective altruist and a billionaire. Of these, as far as I know only Sam Bankman-Fried and Gary Wang were effective altruists prior to becoming billionaires (the others never had the chance, since effective altruism wasn't a thing when they made their fortunes).

William MacAskill writes:

Effective altruism has done very well at raising potential funding for our top causes. This was true two years ago: GiveWell was moving hundreds of millions of dollars per year; Open Philanthropy had potential assets of $14 billion from Dustin Moskovitz and Cari Tuna. But the last two years have changed the situation considerably, even compared to that. The primary update comes from the success of FTX: Sam Bankman-Fried has an estimated net worth of $24 billion (though bear in mind the difficulty of valuing crypto assets, and their volatility), and intends to give essentially all of it away. The other EA-aligned FTX early employees add considerably to that total.

There are other prospective major donors, too. Jaan Tallinn, the cofounder of Skype, is an active EA donor. At least one person earning to give (and not related to FTX) has a net worth of over a billion; a number of others are on track to give hundreds of millions in their lifetime. Among Giving Pledge signatories, there are around ten who are at least somewhat sympathetic to either effective altruism or longtermism. And there are a number of other successful entrepreneurs who take EA or longtermism seriously, and who could increase the total aligned funding by a lot. So, while FTX's rapid growth is obviously unusual, it doesn't seem like a several-orders-of-magnitude sort of fluke to me, and I think it would be a mistake to think of it as a "black swan" sort of event, in terms of EA-aligned funding.

Let's make the following assumptions:

• Suppose right now there are about 9,500 engaged effective altruists; let's say we're 90% certain the number is somewhere between 6,500 and 14,000.^[1]
• Suppose we're 90% certain that P(billionaire|effective altruist) is in the 0.0177% to 0.0526% range.^[2]
  • 0.0177% is the estimate I get for P(billionaire|Ivy League graduate) (ranging from 0.0082% for Brown University to 0.0535% for Harvard University). 0.0526% is the base rate you get if there are 5 effective altruist billionaires and 9,500 effective altruists (5 ÷ 9,500 = 0.0526%).
  • This means our model predicts ~2.9 billionaires right now, an underestimate, which makes perfect sense given that we used the 0.0526% figure as an upper bound. You can take this as evidence that the model is conservative, but I take it as evidence that we're lucky to have five billionaires in our midst. (The model gives ≥5 billionaires in 2022 a probability of ~9%.)
• Suppose we're 90% certain that effective altruism will grow by somewhere between -5% and +40% each year.^[3]
  • This involves updating upward slightly from the current rate (est. median of +14% → median of +18%) due to recent media exposure; I'm not sure if said exposure warrants a larger update.
  • This implies we'll add, in expectation, roughly another 12,000 effective altruists from now until 2027, but with a ~31% probability of ≥19,000 (a tripling).

Implications

Given these (pretty generous, in my opinion) assumptions, the probability that we'll have ≥10 additional effective altruists in 2027 is roughly 14% (with 37% of ≥5 and 76% of ≥1). That's way less than Patel's proposed 50% (or the market's 38% as of my writing this). Instead, the model predicts we'll get 3 or 4 additional effective altruist billionaires in that time (which is not bad).^[4]

Even if we say with certainty that P(billionaire|effective altruist) is 0.0526%, ignoring the Ivy League base rate, the probability that we'll have ≥10 additional billionaires in 2027 is roughly 31%. Either way, 50% looks improbable to me: in order to get there, you'd need to assume both P(billionaire|effective altruist) = 0.0526% and a ~25% annual growth rate of engaged effective altruists (meaning we expect the number of effective altruists to triple in five years).

Here are the different scenarios (with median numbers reported):

I could see the recent success of FTX influencing things in either direction:

• Maybe it inspires people to try to do the same thing; this might cause P(billionaire|effective altruist) to increase in the near future.
• Maybe it (and the sense that effective altruism now has a lot of money) moves people away from earning to give and towards direct work; this might cause P(billionaire|effective altruist) to decrease in the near future.

What about Patel's three arguments? As far as I can tell, P(billionaire|effective altruist) is somewhat smaller than P(billionaire|Harvard graduate) in the wake of FTX, though Harvard graduates have some advantages over effective altruists, like being much older on average, better credentialed and perhaps more talented and well-connected. I think this is evidence that there really is something special about effective altruists. They (or I suppose I should say, we) seem to be unusually likely to be very wealthy, and perhaps also unusually likely to become very wealthy. But I also think it suggests that P(billionaire|effective altruist) will at some point regress towards the saner Ivy League mean.

I'll end with two more speculative thoughts.

First, I'm not sure there's more glory in effective altruist ventures than there is in traditional start-ups. Trying to think of various founders, of course the effective altruists are more famous within effective altruism and in adjacent spheres, but more broadly I don't really see a difference; I see successful non-effective-altruist start-up founders getting lots of prestige and attention. This matters because Patel's first (and to a lesser extent second and third) argument is premised on effective altruism being unusually prestigious.

Second, I think the distinction between "effective altruist → billionaire" and "billionaire → effective altruist" is important. I'd guess that, if we do get ≥10 additional effective altruist billionaires in the next five years, some of them will have become effective altruists only after getting rich (at least one-third, let's say). But that would not really vindicate Patel's arguments, which are all reasons to think that effective altruists will become billionaires, not the other way around.

That said, the number of billionaires is steadily increasing, and that's one reason to think there will be more effective altruist billionaires in the future. There will just be more billionaires period. In fact, if we extrapolate the trends a little more, we'll all be billionaires one day.

Appendix: Squiggle Model

Here is the Squiggle model used to produce the predictions:

numberOfEas2022 = 6.5k to 14k

// time in years after 2022
numberOfAdditionalEasAtTime(t) = {
    annualEaGrowthRate = -0.05 to 0.4
    numberOfEas2022 * ((annualEaGrowthRate + 1) ^ t)
}
numberOfAdditionalEasAtTime(t) = numberOfAdditionalEasAtTime(t) - numberOfEas2022

p_BillionaireIfIvyBaseRate = 0.000177
p_BillionaireIfEaBaseRate = 5.0 / 9.5k
p_BillionaireIfEa = p_BillionaireIfIvyBaseRate to p_BillionaireIfEaBaseRate

// time in years after 2022
numberOfAdditionalEaBillionairesAtTime(t) =
    numberOfAdditionalEasAtTime(t) * p_BillionaireIfEa

{
    // additional billionaires
    numberOfAdditionalEaBillionairesAtTime:
        numberOfAdditionalEaBillionairesAtTime,
    numberOfAdditionalEaBillionaires2027:
        numberOfAdditionalEaBillionairesAtTime(5),
    medianAdditionalEaBillionaires2027:
        numberOfAdditionalEaBillionairesAtTime(5) |> quantile(.5),
    p_10AdditionalEaBillionairesIn2027:
        1.0 - numberOfAdditionalEaBillionairesAtTime(5) |> cdf(10),

    // additional eas
    numberOfAdditionalEasAtTime: numberOfAdditionalEasAtTime,
    numberOfAdditionalEas2027: numberOfAdditionalEasAtTime(5),
    medianAdditionalEas2027: numberOfAdditionalEasAtTime(5) |> quantile(.5),
}

You can paste it directly into a Squiggle playground and play around with the parameters yourself. Note that you may see slightly different results as the precise numbers depend on the random Monte Carlo draws (I think).