While not planning to take this forward ourselves, we anticipated and discussed several future issues.

We assess what went well and what could have been improved in our implementation, and offer a more open-ended discussion on the issues involved.

While many of the issues raised were not critical at our small scale, we had in mind throughout that we should remain as close to a scalable solution as possible.  Consequently the board has discussed issues in detail that were not planned for implementation in the short term.

Donor presentation

Giving donors access to high-level categories gave us a good way to handle difficult issues like setting a funding level for animal charities.

We would have preferred, with donor comprehension in mind, to give detailed descriptions of what each category would and would not include, and the strategic goals of each.  One benefit of not doing this was that we had some flexibility to include later projects that didn’t fit our initial classification system.

Defining categories – as we did for animals and overseas projects – to replace common points of contention with quantifiable choices and a single point of control is a pattern we felt worked well.  For example, having animals as their own category, instead of some donors being “animal people” and others being “people people”, each donor is given their own clear choice about quantifying between human and animal suffering, which we feel is a better outcome.1 gives a list of other common binary distinctions where this approach may work well.

A problem that we have not been able to resolve in this project is how to mitigate the bias introduced by any hierarchical categorisation system.  Focusing on partitioning different concepts and different types of suffering makes it easier for the donor to understand, but may produce categories with vastly different needs, while a natural default response would be to assume that categories presented together are broadly similar in scale and need.  We feel that this could be addressed effectively, but it unavoidably requires some method of quantifying differences between distant categories. This is problematic because, if done poorly or over-simply it will erode donor trust, and we expect that it would require an economics institution to research and maintain a sufficiently compelling model.  With that caveat, we present our initial thoughts here.

Evaluation framework

Putting the primary focus on suffering is how to align well with donor needs.

We focus on two elements for a given cause: the size of the population affected and the suffering experience of the population.  While these are interlinked with mortality, morbidity and longevity metrics, we believe that only focusing on suffering gets close enough to what drives donors.  It is of course harder to describe, measure, and communicate, and is liable to subjective variations.

A provider might determine population sizes and give a descriptive account of their experiences, but it is the donor’s role to determine how they feel about these descriptions and make quantitative distinctions between different experiences.  In essence this is the approach our project encouraged, but the size and experience were implicit or omitted, with the donor left with a lot of guesswork.

A common marketing strategy is to find the single sufferer who best portrays their condition to the audience, and to remove all other information that would distract.  With every condition of suffering, there is going to be a variety of levels of suffering, and it would be achievable and more useful to describe how many people there are at each level.  This will already be happening for some projects behind the scenes, but make these different levels available to donors, giving them the position of judging between accounts of suffering, and they have a level of information that they can use to guarantee their spending aligns with what they want.

Complications arise with measuring the sizes of populations, and accounting for bias and imbalances in different accounts of different experiences, but in the big picture the framework appears viable and would provide a better donor experience.

Donor expression

Overall, we do not expect great results at scale from donors using numbers from 1 to 10 – we expect that donors will feel it is neither clear nor sufficient, but that this high-level perspective is still desirable.  We spent some time working on designs for a better donor experience, focusing on three specific areas.

First, there are several ways to make it clearer and easier for a donor to appreciate the consequences of their choices.  Part of this is visual – showing what part of the whole goes to each category; and part is about giving more context to the quantities – for example showing what an individual’s spend would be over a decade, or showing how we would spend our current budget based on their preferences.

Second, we acknowledged earlier that there are some areas where a considerable proportion of donors might be put off by some aspects of a project: for instance, where medical research involves animal testing, or environmental projects include political lobbying.  Where the distinction is clear, it could be presented to the donor as a yes/no choice, as part of a questionnaire alongside the preference form. Where the distinction is unclear or presenting the choice to donors is unfeasible for some other reason, we would simply exclude the project (and make this clear).

Third, adding feedback mechanisms would enable donors to say where improvements are needed.  We considered how to do this by allowing feedback on individual projects, but felt that giving this level of detail immediately would be unhelpful: directly comparing two different projects is not the same as setting budgets for high-level categories, and our donor profile does not require particular expertise or overall perspective.  We would not rule this out for a more mature service, particularly if effective mechanisms for delegation were in place, but we would be very cautious about tying such feedback directly to individual funding decisions.

Although all donors completed their preference data, we anticipate that many people would prefer to delegate this, and there are several options here.  Apart from completely unrestricted donations, we looked at different ways of aggregating donor preferences to create an “average” profile (i.e. all donors’ preferences, averaged), and thought about how we might set up a “champion” model – where public figures and experts explain how they’ve turned their knowledge and feelings into figures.  Alternatively, we considered ways of setting up a social networking element – though looking briefly into that option raised several practical concerns, and we think it would be challenging to get it right.

A further thought for scaling is that we would expect a large proportion of donors to interact only once with a preferences system, so there is a challenge around whether and how to improve the system over time while honouring donors’ original expressions.

70% of funds being allocated within the UK was reasonable given the preference data.2Global causes, Emergency & Poverty, and Development & Welfare altogether accounted for approximately 50% of the initial budget.  The process of putting together our portfolio of projects made it clear where our earlier versions of our preference model could be improved, and if we were to support new donors now then we could give more explicit definitions of categories that would give clearer answers to questions like this.  There may be more we can do; ultimately, if we can find simpler ways to visualise the whole picture and the practical impact of each decision then that is helping our donors.

Another way to characterise charitable activity is based on how a project addresses a problem.  How to split funding between long-term strategic research into the root of a problem and immediate action to mitigate its effects will generally not be a decision that belongs primarily to experts on the specific problem, but having the donor decide or delegate will be a better outcome.

Some data clusters were visible in our database search, and we can see this becoming a more pressing question at scale.  Homelessness and refugees might be more prevalent in cities, with unemployment affecting rural areas more; different health problems are more pervasive for different socio-economic groups.  Overall, we don’t think that this dilutes the usefulness of the general causes concept, but it highlights that there may be more to do to ensure donors are represented effectively.

There are two kinds of delegation at play here.  One is for decisions which a donor can’t make effectively because they don’t have access to the appropriate information, which we call expert delegation, and which has no decent alternative.  The other is for decisions where a donor trusts someone else to represent their feelings, which we call sympathetic delegation. Some donors will want to avoid sympathetic delegation as much as possible, while others will want to use it heavily in order to maintain an emotional distance from the causes they support.  We want the donor to be able to separate the two types. The basic role of an expert is to provide advice as a service; going beyond this towards a “we know best” mentality without donor consent produces a conflict of interest that erodes confidence in the service relationship.

Overall, partitioning of the charitable sector is a highly constrained problem.  We were pleased with the results from our solution, and expect that it could be used at scale with only minor tweaks, but there are several places to look for a smarter approach.

Work discovery

We were very happy with our results from using the database to find projects without relying on advertising, and it took some of our grant recipients by surprise to receive personalised grant offers seemingly out of nowhere.  This regulator-maintained registry of which organisations are working in which areas gave us the ability to find eligible charities at low cost to both sides and with low bias, and we strongly hope it will continue to be available in a usable form.

As mentioned, we needed to work around a large proportion of charities that were inappropriate for us, primarily schools and churches.  These naturally dominate the register, as they are usually eligible for charitable status. We checked our search procedures and results for other biases, with no major findings, and so overall we can recommend using this database for future projects.

While we have stated that we would have liked to fund more overseas projects, there was less variation in the content of projects we found abroad.  We do not know whether this is because the charities running them were UK-based. We had no objection in principle to running very similar projects in many places, but it raises a question about how we might get better assurance that the right needs are being met and there aren’t opportunities that we’re missing.  A key requirement for running a similar service at scale would be to promote the identification of gaps in existing provision and ensure such gaps can be filled. This is a more noticeable need for overseas projects but will also apply domestically.

Project selection

At scale, it would become more important to define strategic objectives in each category – although the overall aim is to leave nothing forgotten, this gives no guidance about how to attribute different amounts to different projects within a category.  We considered setting up panels of experts to advise the board in each category. We also considered whether, going beyond the “human story” case studies in common use in marketing material, we might find new ways to represent beneficiaries more effectively.  While we have no strong ideas to report here, in principle this could provide us with a more nuanced view of suffering across different categories, and be used more authoritatively for making quantitative decisions.

Beyond this, project selection is already carried out effectively by other granting bodies.  It is in the other service components where we see more immediate opportunity or need for improvement.

Continue: recommendations

References   [ + ]

1. gives a list of other common binary distinctions where this approach may work well.
2. Global causes, Emergency & Poverty, and Development & Welfare altogether accounted for approximately 50% of the initial budget.