Cause overview: cause prioritisation



I recently conducted a ‘shallow investigation’ (see GiveWell) into cause prioritization, with the help of Nick Beckstead. It covers the importance of cause prioritization; who is doing it, funding it, or using it; and opportunities to contribute. We had conversations with eight relevant people (see the three most useful here, here and here). The full document is here and the collection of related interview notes and such is here. This blog post is a summary of my impressions, given the findings of the investigation.

Cause prioritization research seems likely enough to be high value to warrant further investigation. It appears that roughly billions of dollars per year might be influenced by it in just the near future, that current efforts cost a few million dollars per year and are often influential, and that there are many plausible ways to contribute. It also seems like things are likely to get better in the future, as more work is done.

Funding which might be substantially influenced by cause prioritization research is probably worth at least several billion dollars per year.

Prioritization research is likely more relevant to new funders than established ones, but we can get an idea of the scale of new cost-effectiveness sensitive philanthropists by looking at existing ones. This study found nine private funders (or groups of them) who appear to care about cost-effectiveness, and did not look that hard. The Gates Foundation spends around $3.4bn annually, The Hewlett Foundation spent $304M in 2012, and Good Ventures around $10m in 2014. The others spent less or I did not find data on them. Given these figures, it seems reasonable to expect more foundations worth hundreds of millions of dollars per year in the future, who care about cost-effectiveness.

I think funders not explicitly focused on cost-effectiveness are probably also influenced by prevailing beliefs about cause effectiveness, which are likely influenced (gradually) by research. $300bn is spent annually on philanthropy in the US, and probably somewhat less than twice that much is spent globally. Development assistance is at least $125bn per year. Government domestic spending on some kinds of programs is also a worthy target for prioritization research, not measured here.

Few organizations work on cause prioritization.

Those identified in this study to be working directly on public cause prioritization research have a total annual budget of around $3M. Philanthropists also sometimes invest in private cause prioritization, and other kinds of organizations do related work.

Current efforts are plausibly successful, though I haven’t investigated a lot

There are some examples of cause prioritization redirecting large volumes of funding, however I do not know of enough such examples, or know enough about how the money moved, to be confident that the cost has been worth it. We are told of cases of prioritization influencing $4bn of government spending, and substantially moving $208M of government spending and $750M of private funding. These are largely from the Copenhagen Consensus Center (CCC), which has probably spent very roughly $15M ever. If we suppose their only output was moving $820M (i.e. if we ignore their apparent influence on billions of dollars, and all other less clear cases of influence), then in order to break even, they would need to improve the quality of that spending by 1.8%. This seems very plausible, though I have not investigated the strength of their evidence. To be a highly effective use of money they would have to do better. However on the other hand, they appear to have many other good effects, and everyone seems to agree that most of the value from cause prioritization should come in the future anyway, since we are only just learning how to do it.

Many opportunities appear to exist for contributing to cause prioritization.

Existing organizations seek funding, and cause prioritization seems amenable to small-scale research projects, such as individual researchers working. A large variety of research approaches and questions are plausibly valuable, and experimentation is probably unusually good, given the early stage of the research. There are also a variety of non-research routes to contributing to cause prioritization, such as encouraging people to use it, arranging for other researchers to use comparable metrics, organizing relevant academic research, outreach such as workshops at foundation conferences, and encouraging sharing of private research. This project has not investigated the value of any of these specific ideas.

My own views

If I had some money to spend, cause prioritization is one of the top ways I would consider spending it. For an example of the kind of thing I’d consider: I think a pilot project investigating the indirect and long-term effects of relevant actions would be valuable. For instance, when you give cash to a person in the developing world, does it help the country develop faster? Does it change the population? Does it make the world better or worse than you would expect if you just looked at that person’s wellbeing?

Many findings in this area seem applicable to evaluating a range of causes, and would probably remain applicable for a long time. There appears to be relevant academic research (e.g. into the effects of economic growth on violence, or on the extent to which sub-optimal standards persist), and many people suspect long run effects are important, yet when choosing interventions it is common to ignore them. There is disagreement over whether this kind of research is tractable, and I think it is worth checking more thoroughly.

This project would not involve prioritizing causes directly, however it would provide an important building block in the prioritization of many causes, and I expect would reveal some valuable interventions that we were not thinking about. I’m not sure if this is among the best suggestions in the longer document. But it is an example of the kind of project that appears to be tractable, cheap, and promising.

Author: Katja Grace

Katja Grace is a PhD candidate studying probability at Carnegie Mellon University, and a former visiting fellow at the Singularity Institute for Artificial Intelligence.