Ground-breaking analysis by Giving Evidence disproves the popular idea that charities should spend less on administration.
This is the first analysis which shows (doesn’t just argue) that high-
So it’s unhelpful of Parliament’s Public Accounts Committee to be considering limiting charities’ admin costs. It’s unhelpful of donors such as Gina Miller to suggest that admin costs be capped. The data indicate that such caps would nudge donors towards choosing weaker charities, at untold cost to their beneficiaries. It’s time for this to change.
Judging a charity’s quality is hard. Some of the most rigorous analysis is by GiveWell, a US non-profit run by former Wall Street analysts, whose analysis is often dozen of pages. GiveWell looks for various sensible indicators of quality, including: a strong documented track record of impact; highly cost-effective activities; and a clear need for more funds.
Charities which GiveWell reviewed in 2011 and recommended, spend 11.5% of their costs on administration, on average. Charities which GiveWell reviewed and didn’t recommend spent less on administration, only 10.8% on average.
This is no freak result. The same pattern was true in 2009:
In 2009, GiveWell had four levels of ranking, and the pattern is even more pronounced if we use those:
Admin costs are discussed in detail in It Ain’t What You Give, and when we understand what’s included, we can understand the pattern these data show. ‘Admin’ includes systems for capturing learning, for improving, for reducing costs. It’s spending on those things which enables good performance. Scrimping on them is often a false economy.
Assessing a charity by its admin spend is like assessing a teacher on how much chalk they use, or assessing a doctor on how many drugs they prescribe: they’re easy measures but don’t relate to performance. As Einstein said: ‘Not everything that counts can be counted. Not everything that can be counted counts.’
This isn’t to say that there isn’t waste in charities. There is: masses, much of it avoidable, and good charities try to avoid it. But don’t expect to find it clearly labelled in the financial statements.
To be clear, we used all the GiveWell data which were available at the point that the analysis was done. We have stated the sample size on all graphs.
Why is this the case?
Imagine a water charity which operates in several less developed countries to improve irrigation. If it’s run well, it will have a system for recording what works and what doesn’t in particular circumstances, and for sharing that learning between its various country offices. Now, should the costs of that system count as ‘administration’? On one hand, the system isn’t directly helping people: it probably involves databases and conference calls, rather than pipes and water. As a result, it may well be classified as ‘administration’ in a charity’s accounts. However, the system will reduce the charity’s costs and increase its effectiveness, and therefore certainly isn’t waste. Aha – in this case, money spent on administration increases performance.
Let’s consider finance costs. Perhaps the Finance Director purchases a better invoice-handling system. Same thing. That system should reduce work for finance staff by reducing processing times and/or mistakes, which frees up their time (and/or frees up money) to improve the quality and quantity of service to beneficiaries.
Let’s take a real example. Chance UK provides mentors for primary school children who are at risk of developing anti-social behaviour and possibly being permanently excluded from school (formerly called ‘being expelled’). The charity spent some money evaluating its work. It found that male mentors were best suited to children with behavioural difficulties, whereas children with emotional problems responded best to female mentors. Again, the money spent on that evaluation would normally count as ‘admin’, but for the children receiving support which has improved because of that insight, it was money well spent.
We tried to update this analysis in 2019, hoping that there might be new data that we could add. There weren’t really, and the method kind of broke. See here.