Amazing work and well done on trying to wrangle such a complex topic to help inform prioritisation.
I have a few thoughts, especially on the cost side of the equation which may be useful.
On costs, firstly I am not sure that some of your organisational costs are representative.
For example, the costs for Resolve to Save Lives. In 2021 (the year used in your estimation) RSL was split from its parent org Vital Strategies Initiative however this transition was not complete until mid-2022. As such only a very small fraction of activities are covered by this financial report and these activities under the $600,000 grant were for a COVID-19 project (contracted back to Vital Strategies) not for anything to do with salt reduction (this likely was still covered by Vital strategies). There were also no employees on the books. I don’t think this $600K figure is a reliable estimate for costs for salt reduction activities. You may wish to look at financial reports for other years and if possible find ones that directly relate to their salt work.
On the Action on Salt initiative, their costs are also very low but I think it is also important to recognise that this initiative is supported by the Queen Mary University London (e.g. their website is hosted by them, many of the volunteer experts are staff). So I don’t think this expenditure figure accurately reflects the costs it takes to run this project like volunteer time and other costs borne by others like QMUL.
I think this all means you may be underestimating.
Finally, I think there is also a bigger picture question on costs versus expenditures. Above you are using org expenditures but these are not costs. Expenditures are the dollar outlays by a specific group. For example, the cost of going to university is not just the tuition price paid by students. The cost includes all resources to provide university including public subsidies, charitable support etc. In cost analysis we are interested in all resources no matter who pays.
In your CEA you are assuming a magnitude of positive effects related to the successful reduction in hypertension DALYs from activities of salt taxation, food reformulation, school meals, mass public education and package labelling. While the costs for a small policy advocacy charity to nudge these interventions into being may be small that is not the true cost of these interventions. The costs borne by others to actually put them into practice (e.g. govt, industry, schools etc) would be significant and this is not accounted for even though these costs are instrumental to realising the effects of these interventions that you are claiming.
Because you don’t cost in the full cost of these interventions a cross-comparison to a Givewell charity is not necessarily fair. A charity like AMF are responsible for all the costs of running their mosquito net distribution program which support their claimed effects. In your example, you have only costed the relatively small costs of an advocacy charity and left out many other (substantial) costs necessary to realise the effects of these salt reduction activities like paying for mass media campaigns, or changing formulations in factories, or changing food packaging. But you are still claiming all the effects. So it seems to me that the true cost is vastly underestimated and is resulting in a skewed comparison.
Thanks for the feedback on RSL/WASSH—that’s really useful, and something I’ll definitely factor in at the next research stage!
On governmental/implementation costs—I definitely agree that this should be factored in, but just to clarify, the analysis does take this into account, using WHO estimates of the per capita cost of implementing WHO Best Buy policies on sodium (USD 0.03) and on alcohol taxes (USD 0.004, as an imperfect proxy for sodium taxes. Multiplying this with the average country’s population size, as well as the expected years in which implementation will occur (as a function of various discounts like policy reversal rates etc), we get the long-term cost of implementation in the average country.
To this, two discounts are applied: (a) a discoutn for the probability that advocacy succeeds (such that the implementation costs are incurred at all); and (b) a discount for government spending in the average country being far less counterfactually valuable than EA funding which would otherwise have gone to top GiveWell charities or the like. In my experience, discount (b) tends to mean that governmental costs aren’t as significant a factor as they would theoretically be—but it does depend on the country of implementation (e.g. its fantastically cost effective to get rich world governments to do stuff given the counterfactuals; less so if you’re draining sub-Saharan African governments’ budgets).
Hi Joel,
Amazing work and well done on trying to wrangle such a complex topic to help inform prioritisation.
I have a few thoughts, especially on the cost side of the equation which may be useful.
On costs, firstly I am not sure that some of your organisational costs are representative.
For example, the costs for Resolve to Save Lives. In 2021 (the year used in your estimation) RSL was split from its parent org Vital Strategies Initiative however this transition was not complete until mid-2022. As such only a very small fraction of activities are covered by this financial report and these activities under the $600,000 grant were for a COVID-19 project (contracted back to Vital Strategies) not for anything to do with salt reduction (this likely was still covered by Vital strategies). There were also no employees on the books. I don’t think this $600K figure is a reliable estimate for costs for salt reduction activities. You may wish to look at financial reports for other years and if possible find ones that directly relate to their salt work.
On the Action on Salt initiative, their costs are also very low but I think it is also important to recognise that this initiative is supported by the Queen Mary University London (e.g. their website is hosted by them, many of the volunteer experts are staff). So I don’t think this expenditure figure accurately reflects the costs it takes to run this project like volunteer time and other costs borne by others like QMUL.
I think this all means you may be underestimating.
Finally, I think there is also a bigger picture question on costs versus expenditures. Above you are using org expenditures but these are not costs. Expenditures are the dollar outlays by a specific group. For example, the cost of going to university is not just the tuition price paid by students. The cost includes all resources to provide university including public subsidies, charitable support etc. In cost analysis we are interested in all resources no matter who pays.
In your CEA you are assuming a magnitude of positive effects related to the successful reduction in hypertension DALYs from activities of salt taxation, food reformulation, school meals, mass public education and package labelling. While the costs for a small policy advocacy charity to nudge these interventions into being may be small that is not the true cost of these interventions. The costs borne by others to actually put them into practice (e.g. govt, industry, schools etc) would be significant and this is not accounted for even though these costs are instrumental to realising the effects of these interventions that you are claiming.
Because you don’t cost in the full cost of these interventions a cross-comparison to a Givewell charity is not necessarily fair. A charity like AMF are responsible for all the costs of running their mosquito net distribution program which support their claimed effects. In your example, you have only costed the relatively small costs of an advocacy charity and left out many other (substantial) costs necessary to realise the effects of these salt reduction activities like paying for mass media campaigns, or changing formulations in factories, or changing food packaging. But you are still claiming all the effects. So it seems to me that the true cost is vastly underestimated and is resulting in a skewed comparison.
Hi Sophie,
Thanks for the feedback on RSL/WASSH—that’s really useful, and something I’ll definitely factor in at the next research stage!
On governmental/implementation costs—I definitely agree that this should be factored in, but just to clarify, the analysis does take this into account, using WHO estimates of the per capita cost of implementing WHO Best Buy policies on sodium (USD 0.03) and on alcohol taxes (USD 0.004, as an imperfect proxy for sodium taxes. Multiplying this with the average country’s population size, as well as the expected years in which implementation will occur (as a function of various discounts like policy reversal rates etc), we get the long-term cost of implementation in the average country.
To this, two discounts are applied: (a) a discoutn for the probability that advocacy succeeds (such that the implementation costs are incurred at all); and (b) a discount for government spending in the average country being far less counterfactually valuable than EA funding which would otherwise have gone to top GiveWell charities or the like. In my experience, discount (b) tends to mean that governmental costs aren’t as significant a factor as they would theoretically be—but it does depend on the country of implementation (e.g. its fantastically cost effective to get rich world governments to do stuff given the counterfactuals; less so if you’re draining sub-Saharan African governments’ budgets).