Thanks -- I said "several years out" to reduce the risk of missing the effect of a grant in an analysis, rather than out of any reason to be concerned about retention. So having good results on a shorter timeframe would address what I was looking for.

Regarding the special note on upskilling grants -- are data available about the fraction of upskilling-grant recipients who are doing impactful work in the relevant subject area several years out?

I think that's contingent on the percentage of people who receive payments, and the ability to predict one's likelihood of receipt. If GD gives money to the same people every flood season, then I would be much more concerned about this than if everyone in the flood zone knows they have a 5% chance of receiving money in any year their area was flooded.

If the question is whether the beneficiaries may be more likely to stay / underinvest / not take action once identified as conditional beneficiaries shortly before the flood -- it didn't sound like getting the payment was conditional on being in the flood zone when the flood actually hit. If you were pre-registered to location X earlier in the season, and location X was selected as a beneficiary site, it sounds like you got paid. If that's true, one could argue for the opposite effect -- evacuation can be pricey, last-minute flood defenses require resources, etc. So getting them money a few days ahead of the storm might enable better risk-mitigation measures. I'm thinking of the people in the US who didn't leave before Hurricane Katrina due to lack of funds. 

<<NOTE: This is all based on a hypothetical>>

The circumstances under which I would find ethical uncertainty are fairly narrow (I can't speak for the rest of society). The following responses don't capture all the potential nuances but give a sense of where I would likely land in most scenarios:

How about if grantees had already donated their grants to GiveDirectly "earning to give" style and couldn't afford to repay?

Then I'd be okay with the debtor going after GiveDirectly itself as a subsequent transferee (cf. 11 USC 550). We just cannot have a society where fraudsters can successfully accomplish their fraudulent ends merely by routing monies through innocent third-party grantors.

And what if GiveDirectly had already disbursed all the funds but GiveDirectly's recipients mostly hadn't spent them yet?

I'd still be OK with the debtor going after GiveDirectly, which probably could return other funds in its possession. Money is fungible. If not, it could always declare bankruptcy and recapitalize itself from future donations ^[1] if it really were unable to presently repay. That's uncomfortable, but the alternative -- which is awfully close to ratifying SBF's fraud -- is worse to me.

As a practical matter, recoupments from individual beneficiaries would never happen in your hypothetical. Filing suits against individual GD beneficiaries in the US, and then trying to enforce those judgments in a foreign country would be extremely cost-ineffective. This is also where my ethical intuitions start to shift due to a concept we lawyers call detrimental reliance. The beneficiaries would have relied on the existence of that money, and would likely be worse off for having had and have lost than never having had at all. Due to that particular reliance interest, I would cast their moral claim more as the right not to be harmed by SBF's fraud than the right to retain a benefit flowing from SBF's fraud.

In that case, you have two groups with strong moral claims to not getting messed over (the already-paid GD beneficiaries and SBF's victims). The facts that would likely tip me in favor of the GD beneficiaries include their (1) utter inability to evaluate the bona fides of the money source, (2) the absence of any viable alternative method of redressing the harm caused by them getting tangled up in SBF's web, (3) their poverty; and (4) their status as individual persons rather than corporations. 

In contrast, others in the chain like GD would at least knew that the money was from crypto and could have taken legal advice about their risk exposure. And bankruptcy with recapitalization is an option for GD -- that step would mean they could serve fewer people than if they were allowed to keep the money free and clear, but not fewer than if SBF had never existed at all. ^[2]Therefore, the harm to GD here in your hypo is not of the same nature as the harm to the beneficiaries.

1. ^{^}

   In the hypo, suppose GD's sellable assets and cash on hand were $25MM. At the risk of simplification, GD could declare bankruptcy and its creditors aren't likely to get more than $25MM. Often, this can be paid in part out of future revenue -- which is what the Boy Scouts did in the US if I recall correctly. Or some donor could cut a $25MM check for the creditors to go away.

2. ^{^}

   I don't give a lot of weight to the interests of those who would have counterfactually been future beneficiaries if GD were allowed to keep the money, but will not be beneficiaries due to the need to return it. In my book, that's predominately an interest in receiving a benefit from SBF's fraud, which is not a legitimate interest in my book.

My own concerns on this thread have been centered around the offered salary not being enough for a fairly basic standard of living in some people's life circumstances (e.g., people with financial responsibility for dependents, people in the US with significant medical issues), rather than viewing it as a discount from "market rate." For instance,my initial suggestion in response to the salary range was that "covering special financial needs of people so they can take high impact / low salary roles" could be a mixed meta/object-level funder purpose to promote diversity and inclusion within EA.

Unlike the class of people who are being asked to consider a heavy discount from "market rate," people in this category are drawing dead at the outset -- no amount of engagement with the pre-founding process will change (e.g.) how much childcare costs. So it's hard to justify asking them to go through a process that would be known as a dead end for them if salary info were disclosed upfront. Unfortunately, neither childcare nor medical providers can be paid in impact. So one concern with a delayed-disclosure approach is that I don't think it is particularly respectful of the time, energy, and emotions of people in those situations. 

While I agree that Lightcone-level salaries would be wildly inappropriate here (and don't think the EA salary scale should generally give very much weight to Bay Area tech salaries as a general rule), I don't think that is what Elizabeth was trying to say. I took her comment more as "if you're only able to offer what is roughly minimum-wage salary for a significant portion of the readership base, you need to be explicit about that, recognize that this is unfortunate, and recognize that it is a dealstopper for a lot of people." 

Yeah, it's a tough situation involving a lot of tradeoffs. I think we both see some issues with involvement of Western funders here, although the nature of the problems we're focusing on seem to be different.

To formalize it a bit, as I see it:

1. Suppose we come up with a potentially cost-effective intervention that requires significant HCP resources of a particular type (e.g., physician, mental-health therapist).
2. The supply of HCPs in LMICs is pretty much fixed in the short run. Therefore, by default our intervention will divert HCPs from whatever they would otherwise be doing. Not only do we need to count the loss of the value they would otherwise have been creating in their counterfactual job, we're also weakening the local health service and imposing our own judgment on the allocation of scarce HCP resources in the LMIC.
3. As an alternative, we could fund a counterfactual increase in HCPs in the LMIC by funding new HCP schools. If we fund enough new training, our contribution to increasing the HCP pool will equal the number of HCPs we are diverting from what they would otherwise be doing, at least largely solving the problem in paragraph 2 in theory. The number of HCPs deployed according to local priorities will be as high as if EA did not exist. However, this is expensive, and requires a significant amount of lead time. Thus, if we are going to take this approach, we have to start funding the new HCP schools years before we're ready to roll out the intervention.
4. Paragraph 3 is going to be a real challenge in any EA cost-effectiveness analysis where the alternate use of the money is that we can save an under-5 child's life from malaria for $5,000 without consuming HCP resources (e.g., through bednets). The cost-effectiveness will be even harder if much of the effect of increasing the HCP population results in increases to the number of HCPs doing relatively less effective work (e.g., therapists opening private practices catering to wealthier individuals with mild-to-moderate depression and anxiety) or working in wealthier countries.
5. That the new HCPs would exist counterfactually as a result of EA funding makes me more willing to require a period of service in exchange for free training (or else require the HCP to repay the entire cost of providing their education) if that's what it took to get funding for the new HCP schools. 
   1. In the counterfactual -- that the cost-effectiveness analysis didn't work out, and we didn't build the new schools -- these people never had the opportunity to become HCPs at all. Anyone who did not like the post-graduation restrictions could apply to a non-EA, higher-cost school instead, and choose a different career path if not accepted there.
   2. I'm a bit less concerned about the effects of restricting employment choices of HCPs who received free or subsidized training, as long as the conditions are not too tight. For instance, working for the nationalized health service outside of the relatively more economically advantaged areas might be enough for me. 
6. Conclusion: In my view, the avoidance of disruption to a country's existing supply of HCPs is a major factor in favor of interventions that do not require significant additional use of HCP resources. Pulling HCPs from other roles is problematic, and increasing the supply is both expensive and time-consuming (plus may not achieve the desired results). This will likely pose a particular challenge for interventions targeting severe mental health conditions.

It's not just "standard of living" but also security -- if you have advantages like a savings cushion, a supportive and well-resourced family, etc. you can tolerate certain risks on a near-minimum wage salary^[1] better than someone without those things. Those risks include major medical expenses, short/medium term disability, the need for major car repairs, the project not working out and being unemployed until you can find your next job, etc., etc. As you implied, there are also various factors that make you more prone to risks or less able to weather them -- like having moral responsibility for the well-being of one's child(ren), being older (and thus relatively more exposed to medical expenses and disability risks), and so on.

I think some of the disconnect here is between the sacrifice we're implicitly asking of would-be CE incubatees in developed countries in comparison to what other orgs are paying for fairly entry-level work. To the extent your point (4) is valid, it suggests to me that some orgs are significantly overpaying for those positions.

1. ^{^}

   Where I live, minimum wage is $15 per hour, so slightly more than $30K/year.

Agree that the comment was in response to a question about salaries at CE itself. However, I think the reference to "offering higher salaries" most naturally points back to the 40K-60K range identified in a preceding sentence. I wouldn't read it as necessarily applicable to a lower range.

The opacity about funding for people with higher needs is unfortunate. It's understandable that people would be hesitant to invest all that time -- especially if going through the incubation program -- without first getting some more clarity on that point. It also would suck for CE and other members of the cohort for someone in the incubation program to have to drop over this issue rather than drop much earlier.

Joey's comment was in respect to a $40K-$60K range, which is considerably different than $25K-$45K.

I totally get the economic realities of the situation, but I think it's unavoidable that low salaries are going to be a dealbreaker for some highly-qualified candidates. For example, where I live, childcare for one young child is ~ $25K, so this would be a non-starter for many people with a young child (unless, e.g., a co-parent/partner made enough money to largely support the household solo).