On the 23rd, DFID announced their new strategy for the next 5 years. This included at least £2.5bn in extra funding for research into high priority areas – like new treatments for neglected tropical diseases and malaria. We have been advocating this, because we think it could be much more effective than direct overseas health spending (approximately 15-150x), so we were very pleased to see this outcome. Of course, it’s deeply uncertain to what extent our advocacy shaped this result.
Key points from the new DFID strategy
It’s worth recognising that DFID, over the last few years, has become a true global leader in development. The UK spends more in absolute terms on overseas development than any country except the US. It spends almost twice what any other large country spends in relative terms, 0.7% of GNI. Maintaining that commitment, despite big cuts to other budgets, deserves real recognition and praise.
DFID will refocus on crisis response and resilience. This includes some large pots of funding:
- A new £1bn Ross Fund targeting research and development in “vaccines, drugs, diagnostics, treatments and other technologies” to combat “infectious diseases; diseases of epidemic potential, such as Ebola; neglected tropical diseases which affect over a billion people globally; and drug resistant infections”.
- A new £1.5bn Global Challenges Fund “to strengthen resilience and response to crisis”. This will include “research on challenges like beating antimicrobial resistance and protecting animal and plant health, and emerging viral threats in developing countries”.
- A new £500m crisis response fund to allow the government to respond quickly to natural disasters or initial outbreaks of potential pandemics.
There are opportunity costs associated with the development work DFID is de-prioritising. In net, however, this seems extremely positive. We believe, outlined below, that research into the most dangerous diseases can be far more cost-effective for big funders than direct roll-out of those treatments. Moreover, we believe that the global catastrophic risk posed by pandemics is significant and deserves more attention. This new budget addresses that risk along many axes: increasing disease surveillance, researching new defense strategies, and having money set aside to allow rapid action without dithering.
In addition, DFID will be refocusing its aid spending in ways that should improve international stability. We do not have a strong view on whether this is good or not because we lack the relevant expertise. It seems politically attractive, which might make one suspicious, but equally there is evidence that conflict creates huge opportunity costs. Moreover, we believe that global catastrophic risks are exacerbated by conflict because it prevents international coordination of the sort required to resolve collective action problems, and because it encourages arms races in potentially dangerous technology. As a result, we cautiously welcome this shift.
Why we think funding research rather than just direct deployment is probably right for big funders1
We think it is very difficult to work out robustly whether or not research and development is more cost effective than direct deployment of drugs. What we care about is the marginal ex-ante return on investment (ROI) – the expected value of the next dollar spent on R&D. We want an estimate of this for both the research case and the direct deployment case.
For research into cures and treatments for the diseases that inflict the most suffering, estimates of the ex-ante average ROI (rather than marginal) put number between 10-100 DALYs/$1000.2 (OHE, CGD, BVGH, IAVI) There are a number of factors omitted from these estimates which may skew them upwards, and some downwards. See (Dalton, 2015) for further discussion.
In fact, we probably care more about marginal impact than the average impact estimated above, but our own analysis suggests that these numbers are in the same region, with marginal NTD/Big Three ROI is about 14 DALYs/$1000 [1.4,130], but substantially higher for some diseases (malaria being one). This analysis is from first principles and makes some strong assumptions about the nature of the returns, so we wouldn’t use it on its own, but it is a useful piece of confirming evidence.
The marginal ex-ante cost-effectiveness of direct deployment is also hard to find. Crucially, we do not just want to compare to the top recommended charities by GiveWell, because actual government development spending is not all at this level. I estimate it using the WHO-CHOICE threshold for “highly cost-effective” which is 0.6 DALYs/$1000 in the region with the most demanding threshold (AFRO D).3 So research in neglected diseases is likely to be roughly between 15-150x more cost-effective than the threshold the WHO uses for “highly cost effective” in the part of the world where they have the most stringent standards for cost-effectiveness. If we assume that all of DfID interventions are at the level of highly cost-effective then we get a 15-150x improvement by switching to research.
In fact, the improvement may well be substantially higher:
- Much of DfID spending is probably less cost-effective than the highly cost-effective threshold and because of DFID commitments to value-for-money this is likely to be redirected before the best spending.
- Much DfID spending (and disease burden) is in other regions, with less demanding cost-effectiveness thresholds.
- The research estimates numbers don’t consider long-term scientific progress.
On the other hand, there are some reasons why the improvement may be lower:
- Research funding may not be adequately targeted at neglected diseases;
- The amount of funding may be large enough that it is difficult to hire qualified experts.
As a result, our intuitions are that targeted research is probably between one and three orders of magnitude better than the marginal alternatives.
How good was DfID’s funding shift?
It’s very hard to tell how good DfID’s move was, and it is certainly possible that the shift might have been net negative. However, if we assume that the benefits of research funding are 10-100 DALY/$1000, the health effect of a £2.5bn investment is 38m to 380m DALYs averted.4 Conservatively assuming that this is one order of magnitude better than the alternate spending (we estimated one to three), this means the health effect of DfID moving the funding is around 34m to 340m DALYs averted. The most cost-effective direct charities we know of, such as AMF, have a cost-effectiveness of around 10-20 DALY/$1000. This means that the expected health effect of the DFID shift is equivalent to about $1.5bn – $30bn donated to AMF-tier-charities (at current levels of cost-effectiveness).
NOTE for individual donors: It is hard to fund cost-effective research as an individual donor. Moreover, because you are able to pick the most effective development charities you can get something like 10-15 DALYs/$1000, which is much closer to being competitive with well-targeted research (it is more than one order of magnitude better than my estimate for public development aid).
1 The following section is a slightly tidied up version of the notes from our advice to policy-makers.
2 Throughout I have not adjusted properly for inflation for any of the estimates, which come from times between 2005-2015. The ranges are big enough that inflation adjustments will remain relatively negligible.
3 The threshold is higher in richer regions because it is based on average per capita GDP, so this represents the most demanding threshold in use. This means that the actual marginal cost-effectiveness is likely to be worse, since much of health spending is in richer parts of the world, and some may not be highly cost-effective.
4 Using an exchange rate of roughly $1.5=£1 and rounding
This sounds fantastic.
However, I’m very skeptical about drawing any conclusions based on the ‘WHO-CHOICE’ thresholds. Having a threshold solely based on GDP per capita is kinda weird, because the threshold is supposed to represent the marginal productivity of the healthcare system. For example, the threshold of the English NHS has been estimated to be around £13k, which is roughly half of the UK’s GDP per capita. I’m not even aware of any empirical estimates done in any other country, but very preliminary estimates imply the threshold could be as little as 1% of GDP in the poorest countries:
I want to share this, but want to make sure it’s the most up-to-date information you have and couldn’t find a date on the article. I only could get a ballpark estimate from the comment made by Matt! Suggestion: Put a “Date of publication” on the article so the reader can know whether it’s up-to-date info or not.