Different Levels of Vetting Based on Grant Size
Vetting grants is a complex procedure. Vetting is required to determine which are the best grants, and what projects are worth supporting. There’s no controversy in the concept that larger grants should get more scrutiny compared to smaller grants. A larger grant means bigger expectations; however, it can be hard for these expectations to scale proportionally, particularly as different grantmakers typically give different size grants. If you are giving 5 million across a whole year it can seem worth putting a huge amount of time into vetting. Conversely, if you are giving away 500 million, a 5 million dollar project might seem like a small investment. Objectively, across the philanthropic space, it would make more sense to have a more equal application of vetting based on grant size.
The implications of a more equal vetting system are quite radical relative to the normal amount of differentiation that happens between grant sizes. For example, two recent grants we received were for 150,000 and 1.5m; a 10x difference in grant size. However, the difference in vetting was only about 1.5x. This can be even more dramatic with even larger grants. Grants can differ significantly in size; ranging from $5k small scale starting grants, all the way to $50 million megaproject grants. That is a 10,000 x difference in total funding. If linearly proportioned, a 5k grant worth one hour of consideration would mean that a 50 million dollar grant would be worth almost 5 years of investigation. This might seem like a lot, but it's not way off the numbers that GiveWell uses when vetting their applications. This year GiveWell moved ~$500 million to effective charities, but GiveWell also has a staff team of ~40. There are likely some factors of time that are non-linear with scale (e.g. a set amount of time per grant regardless of its size for due diligence, general understanding of the organization etc). Taking this into account we can make a table applying a consistent 5x increase in vetting for every 10x increase in funding.
We can use this information to make a more practical comparison. We can create a table for some organizations and see what the vetting ratio is relative to grants given. The green vetting ratios are organizations that apply more than the suggested fixed ratio level of vetting into an application.
Hours of vetting per organization
72,000 hours (~40 FTE)
7,200 hours (~4 FTE)
2,700 hours (~1.5 FTE)
7,200 hours (~4 FTE)
4,500 hours (~2.5 FTE)
900 hours (~0.5 FTE)
7,200 hours (~4 FTE)
1,800 hours (~1 FTE)
We can compare EA grantmakers with the ratios that occur in the typically large foundation (bottom row) and see that most, but not all, are putting in more scrutiny than the suggested 5x increase per 10x increase in spending.
Why half scaling of ratios instead of full scaling?
A case could be made for a standard x10 increase in vetting per every x10 increase in funding, but it does lead to unintuitive outcomes such as GiveWell not putting in enough vetting (something most would consider unlikely) and small actors like D-Prize putting in far too much (again, few would find likely). Some reasons for the half scale, as opposed to the full scale, include:
Fixed non-evaluative costs such as the time taken to read a one pager, etc
Diminishing returns to the value of further vetting after a certain point such that even if one is spending 10x as much funding on a grant, it doesn't make sense to spend significantly more time vetting the grant (because you're not deriving any extra value from that vetting time)
Full-time grantmakers may be more skilled or practiced at the process, to the point that they can spend less time and get the same outcome
Full-time grantmakers or those with larger funding pools have higher counterfactuals to their time
It does seem like there should be some semi-scaling level of vetting, likely at a higher ratio than is traditional in the philanthropic grantmaking space. This level of incremental vetting could likely improve the ecosystem and allow more smaller projects to start.