Previously I’ve been a Research Fellow at the Forethought Foundation, where I worked on What We Owe The Future with Will MacAskill; an Applied Researcher at Founders Pledge; and a Program Analyst for UNDP.
Stephen Clare
Karma: 4,804
Hi Katherine, I lived and worked in Rwanda for my previous job, so please feel free to message me if you think I can be helpful. It’s a wonderful country. A few thoughts:
I would be careful not to generalize too much from Rwanda --> Africa, as Rwanda’s culture, history, geography, and economy are unique.
I would also just stay away from questions about the genocide or violence—these are super sensitive and there are v complex social and political norms around them.
Outside of the city, few people speak English, and in the city English-speakers are much more likely to be wealthier and better-educated. That will bias the sample of who you get to speak with and maybe pushes in favour of looking into animal stuff (do you speak French? Older people are more likely to speak French).
Is this a personal trip during which you’re hoping to do some interesting research, or are you going primarily for research purposes?
Edit: One other thing is that, depending on which question you investigate, the Rwandan government might not like this if they hear about it. I knew a team of volunteers who came to Rwanda to perform free vasectomies who got deported after they sent an unauthorized tweet. You’ll very likely be fine, but the govt’s wariness of outsiders poking around is something to be aware of.
The model looks great! I think it’s well-formulated and the data are well-researched, so it seems informative.
Substantive things:
-You might want to add pessimistic guesses for the cost of your advocacy. Intuitively, $100k for 5% attribution seems high when I consider travel costs, salaries, lobbying costs, etc. Generally when we assess policy change, we’ve considered the benefits to be the benefits of an org’s most successful campaigns, and the costs to be the org’s total costs because it’s inherently hard to predict in advance which campaigns will be successful (you can’t just use the costs for that campaign).
-Consider adding a note in the sheet explaining what you mean by “Percentage of the consumption decrease due to decreased smoking prevalence” (cell A9) and “Percentage of lives saved due to quitting” (A14). Not sure exactly what those are referring to at the moment.
Some aesthetic things:
-in cells B17:D20 your units are “millions of lives saved”, but the outputs are on the order of thousands of lives saved, so it’s a bit hard to read. You might want to change that to just “lives saved”
-you generally might want to snip the number of significant figures in the cells because it makes it harder to read. e.g. for the cells showing proportion of smokers per age group, snip it off at 2 sig figs
(I’ll come back with some more comments later—just wanted to give quick initial impressions!)
This is one of those findings that, once it’s laid out clearly, seems so simple and important that you wonder why no one did this before. So great science.
Is it right that the AI scenario is an extension in the Guesstimate model, and doesn’t connect to your extrapolation of cumulative emissions? To me it seems more likely than not that the rapid growth in the AI scenario would result in part from AI-driven technological progress in a swathe of economic sectors, including energy, and that this could substantially drive down carbon intensity.
Nice catch, thanks for the careful read Saulius. I think this is especially important because it means that moral weight considerations creep into our measure of AMF’s cost-efficiency even before we try to compare them to THL. GW currently assigns the same value to averting under-5 and age 5+ deaths (100 units), so that’s convenient. I’d guess the “Cost per outcome as good as” cell also factors in other benefits from reduced morbidity?
Thanks Jason! Looking forward to reading the new research.
I agree that this seems important. It also makes me worry about the equilibrium effects. If producer A switches to a more expensive system and producer B doesn’t, then I wonder how many consumers just end up buying more cheap eggs from B.
Will and Rob devote a decent chunk of time to climate change on this 80K podcast, which you might find interesting. One quote from Will stuck with me in particular:
I don’t want there to be this big battle between environmentalism and EA or other views, especially when it’s like it could go either way. It’s like elements of environmentalism which are like extremely in line with what a typical EA would think and then maybe there’s other elements that are less similar [...] For some reason it’s been the case that people are like, “Oh, well it’s not as important as AI”. It’s like an odd framing rather than, “Yes, you’ve had this amazing insight that future generations matter. We are taking these actions that are impacting negatively on future generations. This is something that could make the long run future worse for a whole bunch of different reasons. Is it the very most important thing on the margin to be funding?”
I agree that, as a community, we should make sure we’re up-to-date on climate change to avoid making mistakes or embarassing ourselves. I also think, at least in the past, the attitude towards climate work has been vaguely dismissive. That’s not helpful, though it seems to be changing (cf. the quote above). As others have mentioned, I suspect climate change is a gateway to EA for a lot of altruistic and long-term-friendly people (it was for me!).
As far as direct longtermist work, I’m not convinced that climate change is neglected by EAs. As you mention, climate change has been covered by orgs like 80K and Founders Pledge (disclaimer, I work there). The climate chapter in The Precipice is very good. And while you may be right that it’s a bit naive to just count all climate-related funding in the world when considering the neglectedness of this issue, I suspect that even if you just considered “useful” climate funding, e.g. advocacy for carbon taxes or funding for clean energy, the total would still dwarf the funding for some of the other major risks.
From a non-ex-risk perspective, I agree that more work could be done to compare climate work to work in global health and development. There’s a chance that, especially when considering the air pollution benefits of moving away from coal power, climate work could be competitive here. Hauke’s analysis, which you cite, has huge confidence intervals which at least suggest that the ranking is not obvious.
On the one hand, the great strength of EA is a willingness to prioritize among competing priorities and double down on those where we can have the biggest impact. On the other hand, we want to keep growing and welcoming more allies into the fold. It’s a tricky balancing act and the only way we’ll manage it is through self-reflection. So thanks for bringing that to the table in this post!
It seems to me that this conception of neglectedness doesn’t help much with cause prioritization. Every problem EAs think about is probably neglected in some global sense. As a civilization we should absolutely do more to fight climate change. I think working on effective climate change solutions is a great career choice; better than, like, 98% of other possible options. But a lot of other factors bear on what the absolute best use of marginal resources is.
Thanks for sharing that. It’s good to know that that’s how the message comes across. I agree we should avoid that kind of bait-and-switch which engages people under false pretences. Sam discusses this in a different context as the top comment on this post, so it’s an ongoing concern.
I’ll just speak on my own experience. I was focused on climate change throughout my undergrad and early career because I wanted to work on a really important problem and it seemed obvious that this meant I should work on climate change. Learning about EA was eye-opening because I realized (1) there are other important problems on the same scale as climate change, (2) there are frameworks to help me think about how to prioritize work among them, and (3) it might be even more useful for me to work on some of these other problems.
I personally don’t see climate change as some separate thing that people engage with before they switch to “EA stuff.” Climate change is EA stuff. It’s a major global problem that concerns future generations and threatens civilization. However, it is unique among plausible x-risks in that it’s also a widely-known problem that gets lots of attention from funders, voters, politicians, activists, and smart people who want to do altruistic work. Climate change might be the only thing that’s both an x-risk and a Popular Social Cause.
It would be nice for our climate change message to do at least two thing. First, help people like me, who are searching for the best thing to do with their life and have landed on climate because it’s a Popular Social Cause, discover the range of other important things to work on. Second, help people like you, who, I assume, care about future generations and want to help solve climate change, work in the most effective way possible. I think we can do both in the future, even if we haven’t in the past.
This is perhaps a bit off-topic, but I have a question about this sentence:
I do actually think there is value on poverty-reduction like work
Would it be correct to say that poverty-reduction work isn’t less valuable in absolute terms in a longtermist worldview than it is in a near-termist worldview?
One reason that poverty-reduction is great is because returns to income seem roughly logarithmic. This applies to both worldviews. The difference in a longtermist worldview is that causes like x-risk reduction gain a lot in value. This makes poverty reduction seem less valuable relative to the best things we can do. But, since there’s no reason to think individual utility functions are different in long- and near-termist worldviews, in absolute terms the utility gain from transferring resources from high-income to low-income people is the same.
The assumption is that a policymaker will use these results to shape how strict climate policy is. Stricter climate policies will reduce present-day consumption in the policymaker’s jurisdiction. The goal is to have a climate policy that is just strict enough to balance the future utility gain from improved climate with current utility loss from reduced consumption.
For most real world applications it is convenient to have marginal damages expressed in monetary terms, rather than in utility units. In a final step, the marginal damage estimate therefore must be converted into monetary units. In principle one would just divide the marginal damage estimate expressed in utility units by marginal utility of income or consumption in the present to convert from utility to monetary units. Income inequality in the present however implies that the marginal utility of income is difference between regions and therefore the conversion depends on the choice of the reference or normalization region. Unless potentially very large transfers equating marginal utilities are allowed this yields different monetary values for different reference regions of the aggregated marginal damages as shown in Anthoff et al. (2009a)
(Anthoff and Emmerling, 2016, p. 5, emphasis added).
I think it’s conventional in the literature to use the US as the reference region: “In order to make the numerical results comparable with previous studies we take the US as reference region x throughout this paper” (ibid).
My current understanding of how we should interpret a social cost of carbon of $y is:
“Given certain assumptions about future climate effects, emissions trajectories, how utility scales with consumption, and how much we discount future utility, a utilitarian policymaker in country x should be willing to reduce current consumption in his country by up to $y in order to abate 1 ton of carbon emissions”
This also means that, due to assumptions about diminishing marginal utility, the choice of reference region majorly affects the SCC. For example, Anthoff and Emmerling show how reference region affects their results. (Remember that this, like all SCC estimates, is subject to lots of assumptions about the effects of climate change and discounting.)
(Sorry for the huge image)
If the SCC estimate is low enough it can actually be negative for some regions, meaning that any reduction in present-day consumption in those countries to mitigate climate change would reduce utility.
I wanted to jump in here while the discussion is active, but I’ll also flag that John, Johannes and I are working on this and should have a post on comparing climate vs. global health/development interventions in the not-too-distant future.
Nice work! This seems like a pretty great overview of our current understanding of a whole range of international development interventions, at least on the micro end of the spectrum. Useful not just for your donors, but the community as a whole.
Two quick points. First, in your appendix you write that it would be interesting to see a rigorous evaluation of Jeff Sachs’ Millenium Village Project. DFID did fund a big evaluation of that project which returned pretty negative results. Marginal Revolution discusses that here, plus there’s a paper by Gelman et al. cited at the end of the MR post.
Second, I do think there’s a bit of tension between the part where you write “the donors are comfortable with impact within the next 1 to 20 years or so, but not keen on options like R&D and policy advocacy which may or may not have an impact at all” and your recommendation of TaRL. As you note, it’s completely possible that J-PAL’s work to scale up TaRL across Sub Saharan African will fail. The line between policy advocacy and policymaking support is pretty blurry.
I don’t want to discourage you from recommending TaRL because I think it’s great! And of course you can’t evalute everything. But I’d be interested in your thinking about what differentiates TaRL advocacy from advocacy for better health or macro policies, which you’ve largely excluded.
One thing to note about the bounds of the FP cost-effectiveness estimate is that they aren’t equivalent to a 95% confidence interval. Instead they’ve been calculated by multiplying through the most extreme plausible values for each variable on our cost-effectiveness calculation. This means they correspond to an absolute, unimaginably bad worst case scenario and an absolute, unfathomably good best case scenario. We understand that this is far from ideal: first, cost-effectiveness estimates that span 6+ orders of magnitude aren’t that helpful for cause prioritization; second, they probably overrepresent our actual uncertainty.
On TaRL specifically, the effects seem really good—whether or not we can get governments to implement TaRL effectively seems to be where most of the uncertainty lies.
Great post, thanks for this. I’ll stop chucking in “antibiotic resistance” as a reason to reduce factory farming. I’ll focus on stronger reasons. I think a longer post on this topic would be useful.
On horizontal gene transfer, you write “This last mechanism could potentially be the most important one, but we do not know how common such transfer is or what share of the resistance burden for humans it causes.” Without more information this is not particularly reassuring for me. Do we truly know nothing about how common or potentially important this is? I’d love to see you give a sense of your intuitions here, even if they’re based on theorizing, speculating, or very weak evidence.
Good forecasts seem kind of like a public good to me: valuable to the world, but costly to produce and the forecaster doesn’t benefit much personally. What motivates you to spend time forecasting?
Most of the forecasting work covered in Expert Political Judgement and Superforecasting related to questions with time horizons of 1-6 months. It doesn’t seem like we know much about the feasibility or usefulness of forecasting on longer timescales. Do you think longer-range forecasting, e.g. on timescales relevant to existential risk, is feasible? Do you think it’s useful now, or do you think we need to do more research on how to make these forecasts first?
Lots of EAs seem pretty excited about forecasting, and especially how it might be applied to help assess the value of existential risk projects. Do you think forecasting is underrated or overrated in the EA community?
I’m slightly confused by the part where you say you’re struggling to understand effectiveness on an “emotional” level. Are your doubts about the state of our knowledge about charity effectiveness, or are you struggling to feel an emotional connection to the work of the charities we’ve identified as highly effective?
Wow, this is really fantastic work! Thank you for the effort you put into this. Overall I think this paints a more optimistic picture of lobbying than I would have expected, which I find encouraging.
To follow up on a couple specific points:
(1) Just in terms of my own project planning, do you have an estimate of how long you spent on this? If you had another 40 hours, what uncertainties would you seek to reduce?
(2) Your discussion of Bumgartner et al. (2009) is super interesting. You write “Policy change happens over a long time frame.” I wonder if you could expand on this briefly. Do you mean that it takes a lot of lobbying over years before a policy change happens, or do you mean that meaningful policy change happens through incremental policy changes over time?
(3) Your finding that lobbying which protects the status quo is much more likely to be effective seems particularly actionable. I mean, once put into words it seems obvious, but it’s a point I hadn’t thought about before. I notice, though, that your list of ideas seems to consist of positive changes rather than status quo protection. I wonder if it would be worth brainstorming a list of good status quo issues that might be under threat. Protecting these would be less exciting than big changes, but for exactly the reasons you outline here more likely to work!
(4) I’m interested in thinking a bit more about uncertainty about policy implementation. This is something that we’re currently grappling with in our models of policy change where I work (Founders Pledge). On the one hand, the Tullock Paradox suggests that we should expect lobbying to be extremely difficult (otherwise everyone would do a lot more of it). On the other hand, we’ve noticed that very good policy advocates seem to quite regularly affect meaningful policy changes (for example, it seems like the Clean Air Task Force regularly succeeds in their work).
In your model you write that “the change in probability of policy implementation lies with 95% confidence between 0 and 5%, and is distributed normally.” I’m not sure about this, but I imagine the distribution of “chance of affecting policy success” over all the possible policies we could work on is much flatter than this. Or perhaps it’s bimodal: there are some issues on which it is near impossible to make progress and some issues where we could definitely get policies implemented if we spent a certain amount of money in the right way.
Perhaps we want to start with a low prior chance of policy success, and then update way up or down based on which policy we’re working on. Do you think we’d be able to identify highly-likely policies in practice?
(5) I found this post super helpful, but overall I think I’m still quite puzzled by the Tullock Paradox. If anything I’m more confused now, given that this post made me update in favour of policy advocacy. I think perhaps something that’s missing here is a discussion of incentives within the civil service or bureaucracy. A policy proposal like taking ICBMs off hair-trigger alert just seems so obvious, so good, and so easy that I think there must be some illegible institutional factors within the decision-making structure stopping it from happening. I don’t blame you for excluding this issue considering the size of this post and the amount of research you’ve already done, but it seems worth flagging!
Thanks again for a great post! I’m really excited about more work in this vein.
This sounds like a promising update! Well done, and I’m looking forward to seeing how things progress in the coming months.