Living review of WELLBY cost-effectiveness analyses

Summary

• We curate the living review of how much wellbeing per dollar is created by charities (i.e., cost-effectiveness analyses, CEAs), as measured in wellbeing-adjusted life years (WELLBYs), a new, standardised metric of value. We will add new evaluations to this living review as they are published.
• Our review found 24 charity cost-effectiveness estimates by 4 different evaluators, all produced in the last decade.

1. The dataset includes different types of charities working across the world. While not fully representative, this sample does capture a wide variety of interventions and charities.
2. Beyond using wellbeing, the methods across evaluators are not exactly the same; hence, different evaluators might evaluate different charities differently.

• We find that the wellbeing cost-effectiveness of charities varies dramatically. The best charities in our sample are hundreds of times better at increasing happiness than others. This implies donors can multiply their impact at no extra cost to themselves by funding the best, most cost-effective charities.
• Our sample has two key gaps. It doesn’t include any evaluations of large, well-known charities or a typical act of charitable giving, such as helping the homeless.

Why a living review of WELLBY CEAs?

Can we do good by giving our money to charities? A review of the cost-effectiveness of charities, using the right metric and definition of “good”, is necessary for people to make the right choices with their donations.

We take a wellbeing approach, where what ultimately matters is how people think and feel about their lives (Crisp, 2001; OECD, 2013; Moorhouse et al., 2020; Plant, 2023). This is typically measured with questions like “how happy are you?” on a scale of 0-10 (OECD, 2013). One of the main benefits of using self-reported wellbeing is that we can see, from the data, how much things like wealth and health really matters to people’s lives, rather than assuming we know.

Therefore, the impact of a charity, policy, or intervention is how much it increases wellbeing. We further quantify impact with wellbeing adjusted life-years, aka WELLBYs (see this explainer; Brazier & Tsuchiya, 2015; Layard & Oparina, 2021). One WELLBY is equivalent to a 1-point increase11For context, in high-income countries, being depressed is associated with a 1.3 point decrease in life satisfaction (Happiness Research Institute, 2020), being unemployed is ~a 0.5 decrease (Clark et al., 2018,Chapter 4, p. 63), a doubling of income is a ~0.2 point increase (Clark et al., 2018, Chapter 2; Frijters and Krekel, 2021, Table 2.2), and marriage is associated with a 0.3 point increase a year after getting married (Clark et al., 2008). Close on a 0-10 self-reported wellbeing scale (typically life satisfaction) for 1 person for 1 year. If your wellbeing went from 7/10 to 6.5/10 for 2 years, that would be a loss of 1 WELLBY.

However, resources are limited, so we don’t only care about impact, we care about cost-effectiveness. Not every charity can be funded, so which one does the most good with the money we give? How can we make the biggest difference to world happiness with what we have to spare?

We are not the only people to propose or use WELLBY cost-effectiveness. The WELLBY has been suggested as an alternative or complement to measures of health and wealth in previous editions of the World Happiness Report (Layard & Oparina, 2021), and in academia for evaluating public policy (Frijters et al., 2020, 2024). In 2021, the WELLBY methodology was adopted by the UK’s Treasury as an official way of evaluating the impact of government policies.

We see no good reason not to apply WELLBYs to charities: it gives us a scientifically credible, evidence-based way to work out how much good we do, per dollar, by giving to different charities.

And we are not the only ones to do these evaluations, so we decided to create a living review of charity cost-effectiveness analyses (CEAs) in WELLBYs. This project comes out of our chapter in the 2025 World Happiness Report (Plant et al., 2025). Here we expand on it, and update the review with new or updated evaluations.

Two barriers help explain why donors often do not give effectively: information (they may not know where to give) and motivation (they may prefer causes to which they feel personally attached; Caviola et al., 2020a, 2021). Notably, donors substantially underestimate the differences between charities. In one study, people predicted that the best charities helping the global poor were only about 1.5 times more effective than average ones (Caviola et al., 2020b). In a survey conducted by the Happier Lives Institute, in collaboration with Prolific Academic, we found that people predicted that the best charities at improving happiness were 3 times more effective than average ones (Happier Lives Institute, 2025). The primary difference between the studies is that we removed the reference to charities serving the world’s poorest – changing the implied reference to any charity operating anywhere – and framed the cost-effectiveness as the “amount of happiness created” instead of “the number of lives saved”. If one believes charities differ little in cost-effectiveness, then searching for a better option is not worth the effort. We find that this belief is mistaken.

Method and inclusion criteria

We aim to collate and present CEAs in WELLBYs of charities.

The data is publicly available here, and the code is publicly available here.

We will update this regularly. We will keep a log of different versions.

We plan to have a formal system for people to send their CEAs for consideration.

Inclusion criteria

Here are our current inclusion criteria for a CEA to be included. We reserve the right to edit these over time.

• The CEA must be published in a report, white paper, or academic publication with written explanation. Not a simple spreadsheet, or number without explanation. Not an unfinished or back-of-the-envelope calculation (BOTEC). We need to be able to refer to it / point others to it.
• The CEA must be based on empirical evidence.
• The CEA must be of an actual charity that could potentially receive funds, not of a general intervention or policy. For example, StrongMinds, a charity delivering psychotherapy, rather than an evaluation of psychotherapy itself or a policy to promote access to psychotherapy.
• The CEA must be conducted by an independent organisation, not the charity itself. Unless there is a collaboration with an external reviewer that would significantly reduce the risk of bias.
• The outcome should be improving wellbeing22We focus on interventions that are primarily about improving quality of life, and currently do not include interventions that are primarily about increasing the quantity of life (life-saving). This is because – whether one uses wellbeing or not – increasing the quantity of life has some philosophical complications. We would consider adding them in the future in a manner that shows sufficient variation across philosophical questions. You can read more about this issue in Plant et al. (2022).Close, using subjective wellbeing measures, such as life satisfaction, happiness, or affective mental health33For more discussion about the relation between each of these, see Dupret et al. (2024).Close.
• The CEA must contain a cost-effectiveness figure, or at least a cost figure and a wellbeing impact figure that could be used to build the cost-effectiveness figure.

Variables

Simply put, a CEA of a charity would look at the costs of a charity for providing a service, how many people are reached, and the average benefit per person affected. So, if a charity spent $1 million, reached 50,000 people, and they each got a 1 WELLBY benefit, that is 50,000 WELLBYs for $1 million, a cost-effectiveness of $1 million / 50,000 = $20 per WELLBY (CpWB). This can also be presented as 50 WELLBYs created per $1,000 donated to the charity (WBp1k).

We do, however, include our subjective assessments of:

• the relevance of the evidence; namely, how directly about the charity itself is the evidence? Often there is a tradeoff, where evidence directly about a charity tends to be of lower causal quality (e.g., pre-post data) when more general evidence about the intervention they deploy (e.g., RCTs of psychotherapy) can have higher causal quality.
• the depth of analysis; namely, how much depth has gone into this evaluation. Some of these include simply the impact of an RCT combined with cost, whereas others involve large meta-analyses with analyses for different adjustments, etc. 

These can be understood as indicating uncertainty44We do not include confidence intervals around the central estimate – which is a typical way of representing uncertainty – for two reasons. One is that the evaluations themselves often don’t include them. The other, more conceptual issue, is that confidence intervals only capture statistical uncertainty There are other, harder-to-quantify uncertainties to pay attention to. For example, is the quality of the data sufficient to determine a causal effect? Is the data relevant to the context being evaluated? Has the analysis considered all the relevant parameters? Hence, providing confidence intervals would create false precision.Close, and therefore informing the reader of how confident they should be in the numbers presented.

Comparing the distributions

We average cost-effectiveness ratios into certain groups for presentation: top 5, charities operating in low- and middle-income countries (LMICs), charities operating in HICs, and BOTECs, or within different evaluators. These are ratios, so for the information to be consistent both in “cost per WELLBYs” and “WELLBYs per $1,000”, we use geometric means rather than arithmetic means.

Different evaluators

We collate and present information from different evaluators. We present the results as they are, at face value, with minimal adjustments (sometimes we need to combine the cost and impact reported; and we apply a ‘depth of analysis’ evaluation, see above). We do not take responsibility for how different evaluators conduct their evaluations.

Briefly, all the evaluations produce a similar output: WELLBY cost-effectiveness. However, they differ in terms of their inputs: the evaluations are not all done in the same way. The main differences are:

1. The depth of the analyses, and the quality and quantity of the evidence used55It is widely accepted that there is a ‘hierarchy of evidence’, where research that uses evidence that is higher up is considered more reliable and better at demonstrating causality than those which use data from further down. At the top of the hierarchy are studies of groups of causal studies (meta-analyses), then singular causal studies, then groups of studies with non-causal methods, etc. One difference between the charity evaluations in our dataset is that they do not all rely on evidence at the same level of the hierarchy.

   Beyond the quantity of studies and whether they demonstrate causality or not, academics often use frameworks like ‘GRADE’ which assess the quality and relevance of study to whatever question is at hand. See Guyatt et al. (2011) for a discussion of the role of indirectness in GRADE assessments and Guyatt (2008) or Siemieniuk and Guyatt (n.d.) for more about GRADE generally. Close.

2. Modelling choices, such adjusting for internal and external validity, and whether researchers tried to include longer-term and spillover effects to households or communities (as opposed to just the short-term impacts on direct beneficiaries).

Details about each charity CEA

The charity CEAs we collate on usually consist of technical reports. For brevity and readability, we do not describe every charity CEA here. In the appendix we have a summary of each evaluation.

Results

Wellbeing CEAs of charities only started appearing in the last few years. To wit, the earliest ones were in Plant (2019), although not explicitly in WELLBYs, with the first ones in WELLBYs all published in 2021: by Frijters and Krekel (2021), State of Life (2021a), and the Happier Lives Institute (McGuire & Plant, 2021d).

At the time of writing, we found 24 charity evaluations from 4 evaluators:

• Happier Lives Institute (8 charities): This is us, a UK wellbeing research institute doing charity evaluations focusing on finding the most cost-effective ways of improving global wellbeing.
• State of Life (3 charities): a UK organisation doing social value impact evaluation.
• Pro Bono Economics (10 charities): a UK organisation focused on wellbeing impact in the UK.
• Krekel and colleagues (3 charities): an umbrella term we use to describe the academic publication of Professor Christian Krekel from LSE and his colleagues (Frijters and Krekel, 2021; Krekel et al., 2021; Krekel et 2024).

Table 1. Details about the charities and BOTECs.

Note. Country income: LMICs = low- and middle-income countries; HIC = high-income country. WBp1k = WELLBYs created per $1,000 donated. CpWB = cost per WELLBY.

Explaining the Distribution

Across the curated set (see Figure 1), cost-effectiveness ranges from, at the top, $9.49 per WELLBY or 105.4 WELLBYs per $1,000 donated (Pure Earth) to, at the bottom, $8,690.85 per WELLBY or 0.12 WELLBYs per $1,000 donated (Football Beyond Borders). The most cost-effective charities in the sample are therefore hundreds to thousands of times more cost-effective than the least (see Figure 2).

The most natural explanation for the range is what the charities do and where they operate. The most cost-effective charities deliver inexpensive, high-impact interventions in LMICs, while the least cost-effective work in HICs, where comparable help is far more expensive. The nine charities in the sample operating in LMICs average $30 per WELLBY, more cost-effective than the 15 operating in the UK, which average $1,204 per WELLBY. The UK charities nonetheless remain cost-effective by the standard of a country whose government values one WELLBY at £13,000 (HM Treasury, 2022).

A natural concern is that the spread reflects differences in evaluator ‘generosity’ in evaluation rather than in the charities themselves, since average cost-effectiveness also varies across evaluators and the most cost-effective charities have been evaluated by the Happier Lives Institute (see Figure 3). Two features temper this concern. First, most evaluators have produced only a few evaluations, so a single charity can move an evaluator's average substantially. Second, the evaluators sample differently: State of Life, Pro Bono Economics, and Krekel and colleagues largely analyse UK charities they were asked or found convenient to analyse, whereas the Happier Lives Institute deliberately searches for the most cost-effective charities and concentrates on LMICs. Relatedly, the one charity in a LMIC-context that State of Life evaluates (i.e., Tearfund) is the most cost-effective charity they have evaluated. Cost-effectiveness varies strongly between charities even within each evaluator.

Figure 1. The cost-effectiveness of charities by depth and evaluator.

Figure 2. Comparing cost-effectiveness across the categories.

Figure 3. Comparing cost-effectiveness across evaluators.

Note. The points are individual charities. The crossed circle is the geometric mean. The text above the crossed circle is the geometric mean, and the text below is the ratio between the smallest and the largest.

Methodological Differences Between the Evaluators

There are several notable differences in how the evaluators assess their charities beyond the fact that the Happier Lives Institute focuses on the most cost-effective charities in LMICs. Our comments here focus in large part on the differences between the Happier Lives Institute and the other evaluators because, as the authors, we know the Happier Lives Institute’s analyses best.

Causality and Relevance

A difference between evaluators is the relevance of the data they use to the charities they evaluate and how much of a causal interpretation these data afford. These differences do not necessarily mean one evaluator will have larger or smaller estimates than another. It is an outstanding problem in science to understand how to trade off relevance (generalisability, or external-validity concerns) against the quality of the causal evidence (internal-validity concerns). Our confidence should be highest when an evaluation is supported by high-quality causal evidence (for instance multiple large RCTs) that is also highly relevant, reflecting the programme’s particularities in practice. This is hard to come by, a notable exception being the evaluation of GiveDirectly, which used several large, high-quality RCTs of the programme as it is deployed in practice.

The Happier Lives Institute’s evaluations typically draw on natural experiments or RCTs, which afford a stronger causal interpretation of the relationship between the intervention and the effect on wellbeing. Among evaluations other than the Happier Lives Institute's, only the Krekel and colleagues analyses of Action for Happiness and of the Royal Voluntary Service used causal evidence (an RCT) as the basis of the analysis. Pro Bono Economics’s evaluations use matching estimators combined with a difference-in-differences analysis, which does not afford as much causal interpretation as an RCT but is better than purely associative data such as pre-post results. State of Life has also used methods which, while affording less causal interpretation than the methods above, are better than purely associative estimation strategies. For example, for Parkrun they apply adjustments for the counterfactual, and for Tearfund there is no randomisation but they compare results between participants and non-participants. The remaining evaluations used other non-causal strategies, such as looking at the association between programme participation and wellbeing, or at the change in wellbeing before and after participation without a randomised control group for comparison.

Analyses from Krekel and colleagues, Pro Bono Economics, and State of Life tend to rely almost exclusively on directly relevant (but mostly non-causal) evidence. By relevance we mean the degree to which the evidence relates to the programme a charity deploys as it is delivered in practice. For example, Pro Bono Economics bases its analysis of Football Beyond Borders on a dataset that includes participants directly affected by the programme as delivered. The Happier Lives Institute’s evidence comes from a mix of more and less relevant sources, incorporating direct evidence when available. For example, in the analyses of StrongMinds and Friendship Bench, we used a meta-analysis of RCTs of psychotherapy in LMICs in general, as well as direct evidence (RCTs of the charities' programmes and pre-post data from the charities' own monitoring and evaluation).

Other Factors

Factors leading to the Happier Lives Institute’s analyses being relatively more conservative are as follows. We applied external-validity adjustments to compensate for the limited relevance of some of the data we used and we applied internal-validity adjustments when the causal interpretation afforded was limited. We also applied adjustments for publication bias and/or small evidence bases. Other analyses do not tend to apply adjustments (a few Pro Bono Economics analyses do), and most adjustments reduce the cost-effectiveness. We assume any cost not going to the evaluated programme has no effect, whereas some analyses look only at the costs of a particular programme when a charity delivers several.

The factor leading to the Happier Lives Institute’s analyses being relatively more generous is that we include spillover effects, such as the effect on the recipient's household.

There are two factors that could push in either direction. First, we attempt to estimate the duration of effects from data rather than assuming a fixed duration (typically one year), which sometimes yields a longer and sometimes a shorter effect. Second, we typically use meta-analyses that standardise effects into standard-deviation (SD) changes, where we convert SD changes into points on a 0 to 10 life-satisfaction scale by multiplying by 2 based on the average SD of the Cantril Ladder in the World Happiness Report data, treating changes in typical subjective wellbeing scales and affective mental health scales (e.g., distress, depression, and anxiety) as equivalent in value (Dupret et al., 2024). Other evaluators sometimes convert from scales with different ranges (e.g., Pro Bono Economics often converts results on one scale to match a 0 to 10 scale).

Discussion

We provide the first global review of charity cost-effectiveness in a wellbeing framework. We found 24 charity wellbeing cost-effectiveness evaluations from four different evaluators. The cost-effectiveness between the most and least cost-effective charities evaluated varies hundreds to thousands of times; therefore, this is the first direct evidence in support of the typical claim by experts in effective giving that the best charities can be hundreds of times more cost-effective than average charities.

For donors, the central message is that charities vary enormously in cost-effectiveness, so careful choice can multiply the wellbeing produced per dollar at no additional cost. For example, this could mean that someone in the US with a median yearly income of $65,416 (U.S. Bureau of Labor Statistics, 2026, Q2) who donates 1% of their income (i.e., $654.16) could have approximately the same impact as someone donating 654.16×1000 = $654,160 simply by choosing to donate to the most cost-effective charities.

Analogous WELLBY analyses of public policy also show wide variation. Some policies, such as publicly funded psychotherapy in the UK, can register a negative cost per WELLBY because they save the state money over time, while others appear highly ineffective (Frijters & Krekel, 2021; Frijters et al., 2024; Frayman et al., 2024). Therefore, we think that similar logic could be used by policymakers.

Limitations

For CEA to be credible, good data and analysis is needed. But we will be waiting forever if we want perfect data and analysis. Thereby, in this living review, we present a range of evidence-based cost-effectiveness estimates of charities informed by data, not ‘facts’, and we hope further work will refine them. Given the alternative to estimates is to rely on our intuitive judgments, we see real value in making and using estimates for decision-making purposes.

Our sample has two key gaps. First, it doesn’t include any evaluations of large, well-known charities such as Oxfam. We call charities like this Multi-Armed NGOs (MANGOs). It is difficult to evaluate them because they have so many different programmes (‘arms’), which likely have varying levels of cost-effectiveness. We discuss this more in our chapter in the 2025 World Happiness Report (Plant et al., 2025) and in our blog.

Second, we do not have WELLBY CEAs or a typical act of charitable giving, such as helping the homeless or donating to a foodbank. We explored some of this with BOTECs in our 2025 World Happiness Report chapter (Plant et al., 2025).

Future updates

We plan to keep updating this living review, with new CEAs and with new features:

• Outline a clear, formal process for having one’s CEA added to the living review.
• Trying to do harmonisation exercises to make the comparison more useful.

Contact us

If you would like us to consider your WELLBY CEA of a charity for our living review, please contact samuel[at]happierlivesinstitute.org. We have yet to formalise a process to review applications, but please consider our inclusion criteria and we will do our best in the meantime.

For media enquiries about the living review, please contact hello[at]happierlivesinstitute.org.