Characteristically thoughtful post, @NickLaing! I did a double take when I realized "CoGi" was referring to Coefficient Giving. For the record, our preferred shortenings are Coefficient or CG (or cG, there's active debate on the capitalization in our internal slack - said internal slack perhaps being a very small part of the broader problem you're writing about).
This is great, nice post! Wanted to flag that there's some potentially useful older discussion of related phenomena that I didn't see linked in your post (though possible I missed) in this 2011 GiveWell blog post (title: Why we can't take expected valuate estimates literally even when they're unbiased). I think this is an issue for us at Coefficient Giving too and is one reason among many why I continue to find worldview diversification attractive.
This is great, nice post! Wanted to flag that there's some potentially useful older discussion of related phenomena that I didn't see linked in your post (though possible I missed) in this 2011 GiveWell blog post (title: Why we can't take expected valuate estimates literally even when they're unbiased). I think this is an issue for us at Coefficient Giving too and is one reason among many why I continue to find worldview diversification attractive.
A quick update on this: Good Ventures is now open to supporting work that Open Phil recommends on digital minds/AI moral patienthood. We're still figuring out where that work should slot in (including whether we’d open a public call for applications) and will update people working in the field when we do. Additionally, Good Ventures are now open to considering a wider range of recommendations in right-of-center AI policy and a couple other smaller areas (e.g. in macrostrategy/futurism), though those will be evaluated on a case-by-case basis for now. We’ll hopefully develop clearer parameters for GV interest over time (and share more when we have those). In practice, given our increasing work with other donors, we don’t think any of this is a huge update; we’d like to continue to hear about and expect to be able to direct funding to the most promising opportunities whether or not they are a fit for Good Ventures.
On the housing piece: we have a long internal report on the valuation question that we didn't think was particularly relevant to external folks so we haven't published it, but will see about doing so later this year. Fn 7 and the text around it of this grant writeup explain the basic math of a previous version of that valuation calc, though our recent version is a lot more complex.
If you're asking about the bar math, the general logic is explained here and the move to a 2,100x bar is mentioned here.
On R&D, the 70x number comes from Matt Clancy's report (and I think we may have made some modest internal revisions but I don't think they change the bottom line much). You're right that that implies we need ~30x leverage to clear our bar. We sometimes think that is possible directly through strategic project selection - e.g., we fund direct R&D on neglected and important global health problems, and sometimes (in the case of this portfolio) through policy/advocacy. I agree 30x leverage presents a high bar and I think it's totally reasonable to be skeptical about whether we can clear it, but we think we sometimes can.
Thanks Ozzie, you’re definitely allowed to ask questions like this! We won’t always be able to answer but we welcome questions and critiques of our work.
Our innovation policy work is generally based on the assumption that long-run health and income gains are ultimately attributable to R&D. For example, Matt Clancy estimated in this report that general funding for scientific research ranged from 50-330x in our framework, depending on the model and assumptions about downside risks from scientific research. In practice we currently internally use a value of average scientific research funding of 70x when evaluating our innovation policy work. Of course, 70x is well below our bar (currently ~2,100x), and so the premise of the program is not to directly fund additional scientific research, but instead to make grants that we think are sufficiently likely to increase the effective size of R&D effort by raising its efficiency or productivity or level enough to clear the bar. Moreover, while most of our giving in this program flows to grantees in high-income countries operating on the research frontier, the ultimate case is based on global impact: we assume research like this eventually benefits everyone, though with multi-decade lags (which in practice lead us to discount the benefits substantially, as discussed in Matt’s paper above and this report by Tom Davidson).
Our innovation policy work so far has cleared our internal bar for impact, and one reason we are excited to expand into this space is because we’ve found more opportunities that we think are above the bar than Good Ventures’ previous budget covered.
We also think our housing policy work clears our internal bar for impact. Our current internal valuation on a marginal housing unit in a highly constrained metro area in the US is just over $400k (so a grant would be above the bar if we think it causes a new unit in expectation for $200). A relatively small part of the case here is again based on innovation - there is some research indicating that increasing the density of people in innovative cities increases the rate of innovation. But our internal valuation for new housing units also incorporates a few other paths to impact. For example, increasing the density of productive cities also raises the incomes of movers and other residents, and reduces the overall carbon footprint of the housing stock. Collectively, we think these benefits are large enough to make a lot of grants related to housing policy clear our bar, given the leverage that advocacy can sometimes bring.
We’ll be evaluating new sub-areas as we go to make sure that they are also generally above our impact bar, but we suspect that the same logic of large potential importance will mean that policy changes that make even modest improvements to areas like clinical trial regulation, energy permitting, etc. could be highly impactful.
In terms of scale, while this is a significant expansion of Open Phil’s overall work in the space, it’s a modest expansion of Good Ventures’ (from ~$15M to ~$20M/year). The remaining funding is coming from other donors. As we wrote in our annual review last week:
One implication of our growing work with other donors is that it’s increasingly incorrect to think about Open Philanthropy as a single unified funder making top-down decisions. Increasingly, our resources come from different partners who are devoted to different causes and have different preferences and limitations for their giving. Their philanthropic dollars are not fungible, and we would be doing them a disservice if we treated them as if they were... it’s clearly less true than in the past (not that it was ever perfectly true) that the distribution of grants we advise across causes reflects our leadership’s unconstrained recommendations.
I think this is a complicated question - it's always been the case that individual OP staff had to submit grants to an overall review process and were not totally unilateral decision makers. As I said in my post above, they (and I) will now face somewhat more constraints. I think staff would differ in terms of how costly they would assess the new constraints as being. But it's true this was a GV rather than OP decision; it wasn't a place where GV was deferring to OP to weigh the costs and benefits.
Just flagging that I think "OP [is] open to funding XYZ areas if a new funder appears who wants to partner with them to do so" accurately describes the status quo. In the post above we (twice!) invited outreach from other funders interested in the some of these spaces, and we're planning to do a lot more work to try to find other funders for some of this work in the coming months.
Characteristically thoughtful post, @NickLaing! I did a double take when I realized "CoGi" was referring to Coefficient Giving. For the record, our preferred shortenings are Coefficient or CG (or cG, there's active debate on the capitalization in our internal slack - said internal slack perhaps being a very small part of the broader problem you're writing about).
This is great, nice post! Wanted to flag that there's some potentially useful older discussion of related phenomena that I didn't see linked in your post (though possible I missed) in this 2011 GiveWell blog post (title: Why we can't take expected valuate estimates literally even when they're unbiased). I think this is an issue for us at Coefficient Giving too and is one reason among many why I continue to find worldview diversification attractive.
This is great, nice post! Wanted to flag that there's some potentially useful older discussion of related phenomena that I didn't see linked in your post (though possible I missed) in this 2011 GiveWell blog post (title: Why we can't take expected valuate estimates literally even when they're unbiased). I think this is an issue for us at Coefficient Giving too and is one reason among many why I continue to find worldview diversification attractive.
Just wanted to say that I enjoyed reading this and the section starting with "Online:" and your concluding question really resonated with me.
A quick update on this: Good Ventures is now open to supporting work that Open Phil recommends on digital minds/AI moral patienthood. We're still figuring out where that work should slot in (including whether we’d open a public call for applications) and will update people working in the field when we do. Additionally, Good Ventures are now open to considering a wider range of recommendations in right-of-center AI policy and a couple other smaller areas (e.g. in macrostrategy/futurism), though those will be evaluated on a case-by-case basis for now. We’ll hopefully develop clearer parameters for GV interest over time (and share more when we have those). In practice, given our increasing work with other donors, we don’t think any of this is a huge update; we’d like to continue to hear about and expect to be able to direct funding to the most promising opportunities whether or not they are a fit for Good Ventures.
Thanks Nick.
On the housing piece: we have a long internal report on the valuation question that we didn't think was particularly relevant to external folks so we haven't published it, but will see about doing so later this year. Fn 7 and the text around it of this grant writeup explain the basic math of a previous version of that valuation calc, though our recent version is a lot more complex.
If you're asking about the bar math, the general logic is explained here and the move to a 2,100x bar is mentioned here.
On R&D, the 70x number comes from Matt Clancy's report (and I think we may have made some modest internal revisions but I don't think they change the bottom line much). You're right that that implies we need ~30x leverage to clear our bar. We sometimes think that is possible directly through strategic project selection - e.g., we fund direct R&D on neglected and important global health problems, and sometimes (in the case of this portfolio) through policy/advocacy. I agree 30x leverage presents a high bar and I think it's totally reasonable to be skeptical about whether we can clear it, but we think we sometimes can.
Thanks Ozzie, you’re definitely allowed to ask questions like this! We won’t always be able to answer but we welcome questions and critiques of our work.
Our innovation policy work is generally based on the assumption that long-run health and income gains are ultimately attributable to R&D. For example, Matt Clancy estimated in this report that general funding for scientific research ranged from 50-330x in our framework, depending on the model and assumptions about downside risks from scientific research. In practice we currently internally use a value of average scientific research funding of 70x when evaluating our innovation policy work. Of course, 70x is well below our bar (currently ~2,100x), and so the premise of the program is not to directly fund additional scientific research, but instead to make grants that we think are sufficiently likely to increase the effective size of R&D effort by raising its efficiency or productivity or level enough to clear the bar. Moreover, while most of our giving in this program flows to grantees in high-income countries operating on the research frontier, the ultimate case is based on global impact: we assume research like this eventually benefits everyone, though with multi-decade lags (which in practice lead us to discount the benefits substantially, as discussed in Matt’s paper above and this report by Tom Davidson).
Our innovation policy work so far has cleared our internal bar for impact, and one reason we are excited to expand into this space is because we’ve found more opportunities that we think are above the bar than Good Ventures’ previous budget covered.
We also think our housing policy work clears our internal bar for impact. Our current internal valuation on a marginal housing unit in a highly constrained metro area in the US is just over $400k (so a grant would be above the bar if we think it causes a new unit in expectation for $200). A relatively small part of the case here is again based on innovation - there is some research indicating that increasing the density of people in innovative cities increases the rate of innovation. But our internal valuation for new housing units also incorporates a few other paths to impact. For example, increasing the density of productive cities also raises the incomes of movers and other residents, and reduces the overall carbon footprint of the housing stock. Collectively, we think these benefits are large enough to make a lot of grants related to housing policy clear our bar, given the leverage that advocacy can sometimes bring.
We’ll be evaluating new sub-areas as we go to make sure that they are also generally above our impact bar, but we suspect that the same logic of large potential importance will mean that policy changes that make even modest improvements to areas like clinical trial regulation, energy permitting, etc. could be highly impactful.
In terms of scale, while this is a significant expansion of Open Phil’s overall work in the space, it’s a modest expansion of Good Ventures’ (from ~$15M to ~$20M/year). The remaining funding is coming from other donors. As we wrote in our annual review last week:
FWIW I certainly agree with "non-trivial"; "huge" is a judgment call IMO. We'll see!
I think this is a complicated question - it's always been the case that individual OP staff had to submit grants to an overall review process and were not totally unilateral decision makers. As I said in my post above, they (and I) will now face somewhat more constraints. I think staff would differ in terms of how costly they would assess the new constraints as being. But it's true this was a GV rather than OP decision; it wasn't a place where GV was deferring to OP to weigh the costs and benefits.
Just flagging that I think "OP [is] open to funding XYZ areas if a new funder appears who wants to partner with them to do so" accurately describes the status quo. In the post above we (twice!) invited outreach from other funders interested in the some of these spaces, and we're planning to do a lot more work to try to find other funders for some of this work in the coming months.