Very briefly, the process was starting with GiveWell’s list of charities they might like to see (this was mostly broad cause areas) and doing further research into each area to determine a short list of promising specific charities.

1) I think my main thoughts around this is that the charity market is a lot less efficient than the for profit market so there are more low hanging opportunities. And even the for profit market isn’t perfect. There’s a specific brand of vegan sausages in the UK that is impossible to get in Canada, despite it being better than anything Canadian-made. Additionally, as we mentioned in our report, there are already charities doing this, it’s just it’s often easy to forget how large the world is. AMF and all the other enormous bed net charities still haven’t filled all the gaps for bednets because there are just so many people. Likewise with SMS. On top of that, mobile health is a relatively new field due to the nature of the technology, so a lot of obvious things are still just starting to permeate the space.

2) This could be true. We do have an explicit category for logistical difficulty, so we certainly have considered that as an important factor. That being said, I think even if we held logistical difficulty consistent across interventions, our team would still favour SMS based on the other criteria.

3) I can see this being tempting, but we do not think of SMS as something that we simply program and wait for the DALYs to roll in. We suspect there will be constant work for further expansion, tweaking of messaging and code, and testing to see what could improve the system.

Thanks for the links.

There is indeed some competition although we found this to be true in basically every area. We are in contact with a few these folks about possible partnerships.

I could imagine most of the value of studies coming from very high quality studies, although I think there is also quite a lot of value in some replication and exploratory studies. I agree high quality studies are both hard to do and generally have their difficulty underestimated. We’re also on the same page when it comes to small problems at the beginning making the research effectively useless later on, which is why we would definitely hire people with a history of running exceptionally rigorous studies if we were to run this intervention. We are generally in favor of hiring capacity when it’s crucial.

The estimate of millions of dollars is a bit higher than the estimates I have seen although I could imagine it being the case. The estimate we used was the higher end of of the $50,000 and $500,000 range that IPA gave for its costs per studies. Do you have any information about the costs of running studies or more specifically the cost of the gold standard type studies?

Thanks for the thoughtful response! Some thoughts on these that I will also email you but thought that other folks on the EA Forum might like being kept in the loop.

With regards to your essay, I really like it and found a lot of parts very interesting. This is in fact an area we looked into a little bit, although did not write much on it. I strongly agree that ‘internal happiness interventions’ (‘IHIs’) are extremely neglected in contrast to the ‘external happiness interventions’ (‘EHIs’) compared to what a rational utilitarian should value.

I did end up with a different conclusion on the issue, although I still think the case for this is strong, I just do not see it as quite as strong as the listed poverty options. In general I think it's very hard to compare the possible expected value of research where it is very hard to get an estimate (e.g. IHI research) to the expected value of a more direct intervention (e.g. AMF). I think we mainly differed in how we weighted the Easterlin paradox and the expected value difference. I see how you came up with the 10-18 number but think that it was generated using a pretty favourable example given the DALY weightings I have seen for different things. For more info, check out Jeff Kauffman’s handy listing of disabilities by weighting https://www.jefftk.com/p/disability-weights.

Although the endline idea of online CBT/mindfulness is very interesting, I also have major concerns about evidence about even the offline versions of these and would expect the effect on happiness to be smaller than is currently stated in many studies if they were stronger studies/less publication bias etc. For example, the following review on mindfulness research found that of 21 pre-registered RCTs on mindfulness meditation, only 38% were published 30 months after registration, showing worrying signs of publication bias, and the median sample size of 124 RCTs was 54, including controls (http://journals.plos.org/plosone/article?id=10.1371/journal.pone.0153220).

I have a lot of sympathy for your argument about death being over-valued (and we are aiming to publishing a blog post on this at some point), but it’s worth taking into account the happiness effects on a parent of losing a child (which can still make reducing death very significantly net positive.) This is even without taking into account the economics arguments mentioned.

I think a combination of all the points made above make me think that while this is a promising area for research, I would give it a smaller chance of revealing something more cost effective than AMF than I would give to doing more direct poverty research (on say SMS or CCT).

I would be very interested in talking more about this as I can imagine my views changing quite a bit. Particularly it would be great to talk about what research you plan on doing going forward on it.

Re: conclusions

CCTs not increasing happiness at all would surprise me. The cash component of the disagreement might come from different general views about the evidence that increased income leads to happiness. But a huge part of the benefit would be from the activity that is being CCTed. For example, even if the cash component of CCT for vaccines only lead to short term happiness, the vaccines might lead to ongoing happiness (details would largely dependent on the condition). There is substantial evidence that parents mourn the passing of a child, and thanks for medical advances, even in the developing world losing a child is a relatively rare event, so it’s unlikely there’s as strong a hedonic adaptation effect.

I do agree that happiness is not nearly as tracked as I would like and I would want to see long term happiness data in conjunction with health/wealth data for any intervention/RCT we end up doing. I think this is particularly true for CCTs, but I would love to see more happiness data almost everywhere.

Research was hard. Some people seemed to think it was much more effective than any direct options and others thought the opposite. I still do not have very confident views despite the time I have put into researching it. Would be super interested if you or anyone else had a cost effectiveness estimate for poverty research based on historical data that might be able to shed some light on the question. The biggest question mark is to what extent research affects where resources are spent counterfactually, rather than simply providing a justification for existing programs to continue.

On tobacco, even if one takes the epicurean view on death there seems to be considerable suffering caused to relatives from premature death (both from sadness and lost income if the person is still working age or due to widows being treated terribly in many cultures). When thinking about how much to value someone’s life we both took the above into account and the average happiness of the population we would be helping.

With regards to iron, this is in fact one of the ones we are more confident is connected directly to happiness. Anemia is quite correlated with negative emotional states relative to other health interventions we considered. It is so cost-effective than even small benefits in happiness really add up.

For SMS would be happy to talk to any behavioural economists who you think might be informed in the area.

I agree personal fit is important. Although we do not have an explicit separate category for this, we definitely took it into account when evaluating options.

I do not know of any research in this area offhand, but I would expect there to be some online. I would be interested in seeing a cost-effectiveness estimate on this if you found one.