Moral Weights

Published: March 2025

In our effort to identify and direct funding to outstanding, highly cost-effective programs that save and improve lives, our researchers spend thousands of hours reviewing the empirical literature, talking to experts, meeting with government officials, and visiting programs in person, with the goal of understanding the impact of the programs as fully and accurately as possible.

Drawing on all of those sources, our research team builds cost-effectiveness analyses (CEA) to estimate the programs’ impact. These CEAs incorporate a range of judgment calls, such as judgments regarding the strength of the evidence or regarding how similar we believe an organization’s proposed program is to the randomized control trials of a related intervention.1

Judgment calls about evidence can be more straightforward than the judgment calls required to compare programs with different types of outcomes to each other. Such comparisons fall more heavily in the realm of ethics.

For example, which of the following might be preferable, assuming they all cost the same?

• Give six very poor households a cash transfer that would double their income for a year
• Correct clubfoot for five children
• Prevent the death of one child from malaria

Whenever someone chooses to support one of these programs over another, they’re making a judgment about the value of those outcomes and, at least implicitly, ranking one over another. Because we’re also making these kinds of judgments as we evaluate funding opportunities, we make our calculations explicit to make sure that people reading our recommendations can understand and evaluate the judgments we’re making.

In our cost-effectiveness models, we use moral weights—that is, the units of value that we assign to different positive outcomes—to compare outcomes and help us understand how much a particular program is likely to help people per dollar donated. Though we believe moral weights are ultimately subjective, we’ve worked to ground ours empirically by incorporating surveys and widely used global health metrics. We believe that drawing from these sources helps us reach a reasonable and well-balanced conclusion, though we acknowledge there is no one correct approach. We expect to engage in ongoing reflection on and revision of our moral weights over time.

Our current moral weights focus primarily on three outcomes: increased consumption, lives saved, and years of poor health or disability averted. While these weights cover the primary outcomes generated by our current recommendations, we are exploring ways to adapt our approach as we consider a broader range of programs.
For example, we plan to evaluate programs that aim to remove barriers to contraception and may also evaluate programs that aim to reduce interpersonal violence. These programs often have effects on income and health that are similar to what we currently model. However, they may also have significant impacts on factors we haven’t modeled in depth yet, such as subjective well-being, autonomy, and mental health. In these cases, we’d look for ways to incorporate these additional factors that aren’t captured by outcomes we currently include in our moral weights.

What are GiveWell’s current moral weights?

Consumption. Our current benchmark moral weight is the value of doubling consumption for one person for one year. We have assigned this a value of 1.

Mortality. We have assigned a value of 134 to saving the life of a five-year-old child. That means that we believe that saving that life is 134 times as valuable as doubling the consumption of one person for one year.

Like many others, we believe that the value of averting a death changes depending on the age of the person whose life is saved. To account for this, we adjust our mortality-related moral weights by age. The values we assign follow a rough curve, from 84 units of value for a newborn, peaking at 134 units for children five to nine years old, and falling gradually to 12 units of value for people over 80 (see here for a full age breakdown and here for information about how we arrived at these values).

Morbidity. We have assigned a moral weight of 2.3 to averting one year of healthy life lost to disability (YLD). This widely used global health metric represents the equivalent of one full year of healthy life lost due to disability or ill health. Particular conditions are then weighted based on their severity.2

Outcome examples	Moral weight
Doubling consumption for one person for one year	1
Averting a lifetime of disability associated with clubfoot	30
Averting the death of a newborn	84
Averting the death of a forty-year-old adult	86
Averting the death of a five-year-old child	134

How does GiveWell use moral weights to estimate the value of specific interventions?

Let’s say that GiveWell is evaluating a program in Nigeria that provides medicine to prevent children under five from dying of malaria. To calculate the value of that program, the researcher would first find data on the age distribution of malaria deaths in Nigeria (we typically use data from the Institute for Health Metrics and Evaluation (IHME)). Next, they would multiply the relevant share of deaths at various ages with the moral weight of averting death at that age (e.g., 84 for early neonatal, 127 for children aged one to four, and so on). This weighted average would then be multiplied by the number of deaths averted by the program to generate the intervention’s total units of value from averting mortality.3

Let’s say that in this very simplified hypothetical example, we estimate that the funding we provide averts 100 deaths of children at an average age of one year old. That mortality reduction would produce 12,700 units of value (100 lives multiplied by our moral weight of 127 for the life of a one-year-old). Other benefits of the program would produce additional value, and those benefit streams would be added together to generate the total value for the funding opportunity. To see how this plays out for an actual malaria program, see this section of our cost-effectiveness analysis for Malaria Consortium.

Now let’s say we want to compare that life-saving vaccine program to a program that addresses disability or the health effects of a disease. Here we typically begin with the disability weights provided by the IHME, which range from 0 for full health to 1 for death. For example, in our cost-effectiveness analysis for MiracleFeet’s clubfoot treatment program, we start with a disability weight of 0.237, which we base on a condition that we think is comparable to bilateral clubfoot, multiply that by the years of disability averted to arrive at our total YLD figure, then multiply that result by our morbidity moral weight of 2.3 per YLD to generate the intervention’s total units of value from health effects.4

Let’s say that in this very simplified hypothetical example, we estimate that our funding resolves 100 cases of bilateral clubfoot and that those benefits persist for 50 years after treatment. That reduced disability would generate 2,725 units of value (100 cases multiplied by the disability weight of 0.237 multiplied by 50 years multiplied by our morbidity moral weight of 2.3). Other benefits of the program would produce additional value, and those benefit streams would be added together to generate the total value for the funding opportunity. To see how this plays out for an actual grant, see our cost-effectiveness analysis of MiracleFeet.

Now that we have estimated the units of value created by these two hypothetical programs, we can more easily compare them.

How did GiveWell set these moral weights and how have they changed over time?

Early history of GiveWell’s moral weights

We’ve been thinking about moral weights since GiveWell was founded, and our approach has changed a number of times since then. From the beginning, we’ve recognized the subjective nature and difficulty of comparing different outcomes.

This blog post from 2008 is one of our first attempts to think through an approach. In the post, we laid out a series of possible outcomes, such as “life-years saved” and “cases of extreme suffering prevented/rectified,” without any attempt to judge among them. We explained, “We prefer to leave major judgment calls to our donors when practical. This means that rather [than] adopt a single definition of value (such as Disability-Adjusted Life-Years), we hope to find the most cost-effective interventions for several different definitions of value.”

Later, we compared many programs in terms of a “lives saved equivalent” measure that incorporated both mortality and non-mortality effects, but we did not yet have a fully consistent approach, particularly for programs like cash transfers that focused on increasing consumption. For more on our approach during those years, see this blog post and this page.

Using staff moral weights

By 2017, as we explain in this blog post, we had developed a highly subjective conversion factor that enabled us to compare years of healthy life with increases in income. Our moral weights at the time were primarily determined by the personal values of staff. (See an example spreadsheet here.) We invited all staff to come up with a set of moral weights, then combined their resulting impact estimates in our calculations.5

This was a useful method because our staff are very familiar with and engaged with the questions we’re trying to answer and how the moral weights fit into our analyses. Staff had a variety of approaches in developing their own moral weights.6 Some considered factors such as life expectancy, the development of personhood over time, the grief associated with death at different ages, economic contributions made at different ages, and so on. Staff also used literature on the value of a statistical life, sought input from philosophers, and researched how other actors, such as governments and the World Health Organization, make these judgments.

This approach led us to set the following moral weights (as of early 2019):

• 1 unit of value for doubling the consumption of one person for one year
• 48 units of value for averting the death of a child
• 85 units of value for averting the death of an adult

While this approach was useful, it had a number of shortcomings. First, staff have limited insight into the lives of the people directly affected by the programs we recommend. In addition, staff composition changes over time, which sometimes resulted in substantially different moral weights for the same outcomes from year to year, making comparisons difficult.

2019 Beneficiary survey

In order to gain more insight into the moral preferences of those affected by our funding, we reviewed empirical literature from low- and middle-income countries relevant to moral weights, but we found that there was little research that we could use to inform our decision-making. Because of that, we decided to commission our own research.

In 2019, we funded IDinsight to survey approximately 2,000 people in Ghana and Kenya living in poverty about their preferences around averting deaths at different ages and around trading off averting a death with increasing income.

The results suggested that, when compared to the moral weights developed by GiveWell staff, the respondents valued averting death more highly than increasing consumption. In addition, those participating in the survey valued the lives of young children much more highly than the lives of adults. This also differed significantly from our existing moral weights, which had valued adult lives more highly. More details about the survey, its findings, and its shortcomings are available here.

Based on our analysis of the results, we revised our moral weights as follows:

• 1 unit of value for doubling the consumption of one person for one year
• 100 units of value for averting the death of a child
• 100 units of value for averting the death of an adult

These moral weights provided us with the ability to compare children and adults, but did not provide information about more specific age ranges. Unfortunately, the survey data did not provide useful information about this question, so we tried another approach.

2020 Donor survey

In 2020, we surveyed about 70 of our largest donors about how they valued deaths averted at different ages in order to help us create more granular weights. We did not ask them to compare the value of deaths versus consumption, only to assess deaths at various ages. This information was easy to gather and was from a key stakeholder group external to GiveWell. Unlike other methods, it also allowed us to distinguish among weights for stillbirths, early neonatal deaths, and deaths at age one. We’re particularly interested in distinguishing moral weights in early life, as many of the programs we recommend focus on reducing infant mortality.

At the same time, donors have limited information about the lives of people affected by the programs GiveWell funds, and the donors we surveyed were not very diverse across characteristics such as race, gender, income, and country of origin. Because of this, we were careful to ensure that the other sources also continued to inform our moral weights.

GiveWell’s current approach to moral weights

We currently use the three sources described above. Each source has its own limitations; we think combining them helped us come up with sensible numbers that are sufficiently granular to compare programs effectively. We encourage those who are interested in looking further into our moral weights to use this moral weights tool to explore the impact that valuing different outcomes has on cost-effectiveness.7

As indicated here, we assigned a value of 1 to the value of doubling consumption for one person for one year and a value of 2.3 to averting one year of healthy life lost to disability (YLD). Using the moral weights of 100 to avert the death of a child and 100 to avert the death of an adult as a starting point, we then made granular adjustments using the donor survey to generate age-specific weights, as indicated in the chart below.8

How might GiveWell’s moral weights change?

Our moral weights are a work in progress. We have a number of open questions that we have not fully explored. We expect that as we dig into those questions, we will update and refine our moral weights. For example, we might:

• Do research to update our trade-off between consumption and mortality. This trade-off has a substantial impact on our grantmaking. We fund few livelihoods programs—that is, programs aimed primarily at increasing income rather than improving health. If we doubled the value of consumption, the cost-effectiveness of livelihoods programs would increase, such that some programs that are currently below our cost-effectiveness threshold might be above it.9 We expect this would lead us to fund more livelihoods programs and fewer life-saving programs.

  We think there are a few ways we could collect more information on this tradeoff:

  • Incorporate data from value of a statistical life (VSL) studies. There’s a large literature that tries to measure the “value of a statistical life,” generally based on either “revealed preferences” or “stated preferences.” Revealed preferences are based on how much people must be paid to accept a particular risk of death. These estimates use real-world choices, such as how much additional money someone needs to be paid to take a job with a 1% higher mortality risk than another similar job. Stated preferences are based on directly asking people questions about these tradeoffs, such as how much they would be willing to pay to reduce their risk of death by 1 in 10,000. Our impression is there is a growing VSL literature looking at low- and middle-income countries specifically. To date, we have not reviewed this literature or incorporated it into our moral weights.
  • Fund additional surveys. This could include surveys of more individuals in low- and middle-income countries (beyond Kenya and Ghana, where we did our initial beneficiary preferences survey); these surveys could use a broader sample or alternative approaches.10 We could also survey others, such as experts in global health and development or decision makers in low- and middle-income countries.
• Update our moral weight for morbidity. As GiveWell has begun to recommend more grants beyond our Top Charities, we’ve begun to evaluate more programs focused on reducing morbidity rather than mortality. This includes programs to reduce anemia, correct clubfoot, and provide surgery to repair fistulas. As we learn more about these conditions, we may also want to conduct more related research. For example, we could survey beneficiaries about their preferences related to decreasing disease symptoms (such as avoiding a year with anemia) compared to increasing income. It’s possible that the results of such a survey would lead us to update our moral weight related to morbidity.
• Incorporate new factors such as subjective well-being. Subjective well-being covers how someone feels about their life. It’s distinct from income or health but can be tricky to measure. In 2023, we conducted a cost-effectiveness analysis of Strong Minds, a psychotherapy intervention, using WELLBYs, a subjective well-being approach based on life satisfaction scores,11 then analyzed the cost-effectiveness of one of our live-saving Top Charity programs using the WELLBY metric and compared it to StrongMinds. While we concluded that StrongMinds was less cost-effective than our Top Charities,12 we believe that subjective well-being deserves more study. If we chose to analyze all possible outcomes such as mortality and consumption in terms of their effect on subjective well-being, the result could be a different set of moral weights.
• Adapt our approach to account for other outcomes. We think our existing moral weights for increased consumption, lives saved, and years of poor health or disability averted capture the main benefits for many programs we’re considering funding. However, there may be some cases where these exclude key benefits. For example, one of the primary outcomes of removing barriers to contraception may be recipients’ increased autonomy or reduced feelings of anxiety, in addition to impacts on income and health. Fistula surgery may lead to improvements in well-being through reduced stigma or improved mental health, in addition to reduced disability. We’d want to make sure we’re capturing these other outcomes. We expect to continue to look for ways to adjust our approach so that we’re accounting for the kinds of good done by different programs.
• Update the comparison of our moral weights to other actors. We’ve previously compared our moral weights on consumption and mortality at different ages to the moral weights of other actors, such as governments and global health organizations. At the time, we found that their values for increases in income or consumption relative to deaths averted at different ages were similar to our own, and we think this is still true, given our updated moral weights.13 However, we haven’t updated our analysis to include more recent data that we would use to form our best guess of other actors’ moral weights. This includes data from VSL studies (including from low- and middle-income countries), how other organizations compare deaths at different ages, and how they deal with morbidity and other outcomes.

***

Ultimately, we hope to continue updating and developing our moral weights so we can evaluate and support opportunities that do good.

• 1

  This is one of the reasons that we share our research and cost-effectiveness models: so others can evaluate our conclusions. Transparency is one of GiveWell’s core values.

• 2

  The weights vary from 0 (perfect health or equivalent) to 1 (death). We calculated the moral weight of averting one YLD by dividing the value of averting a death at a particular age by the average remaining years of life at that age, then averaging the results. We only included ages 10 and up. If we were to calculate moral weights only on the basis of remaining life for all ages, then averting the death of a one-year-old would have a greater value than averting the death of a five-year-old. However, our current moral weights place greater value on the life of a five-year-old. In other words, the moral weights we’ve assigned to averting the deaths of younger children are lower in comparison to their average number of years of remaining life. Thus, to avoid skewing the results, we only included people aged 10 and over in this calculation.

• 3

  This is a simplified example that leaves out most of the parameters in our model. For example, to determine how many deaths are averted by the program, we need to know the mortality rate associated with malaria among children under five in the parts of Nigeria where the program is being implemented, how much we estimate the medication will reduce mortality caused by malaria, the percentage of children we expect will receive the medication who wouldn’t otherwise have, and much more. In addition, programs that avert mortality may also have other benefits, such as increasing children’s long-term income. See here for links to some of our full models.

• 4

  Again, this is a simplified example. In our full cost-effectiveness analysis for MiracleFeet’s program, we incorporate the estimated effectiveness of the Ponseti treatment method in averting clubfoot, the number of people whose clubfoot recurs after treatment, the length of time between identifying and fully treating clubfoot (as that affects the average life expectancy after treatment), and more. See our cost-effectiveness analysis here. Note also that in this process, we sense-check disability estimates rather than taking them at face value. For example, in our research on iron fortification, we reviewed the DALY weights associated with anemia and compared them to other analyses of physical capacity/fatigue and cognitive ability. See our write-up here.

• 5

  For more about our approach at the time, see our 2017 write-up here.

• 6

  For more on the sources and approaches used by staff, see this section of our 2019 write-up.

• 7

  For fuller analyses of moral weights for our Top Charities, see here.

• 8

  To generate this breakdown of moral weights by age, we combined donor preferences, years of life lost (YLLs), and one staff person’s weights that were near the median as a proxy for all GiveWell staff. We placed 60% weight on donor responses, 10% on the GiveWell staffperson, and 30% on YLLs (both as a commonly-used metric itself and as a proxy for the IDinsight survey). Donor responses received the majority of the weight primarily because in our view, that curve represents the most plausible set of weights. In particular, the other two do not distinguish between early neonatal deaths and deaths at age one, which is an age range we're especially interested in. We use donor weights as our result for stillbirths because we don't have data from the other sources on stillbirths.

  The absolute weights were determined by ensuring that, if we average the value of averting the death from malaria of a child under five with the value of averting the death from malaria of a person over five, the result is 100. This allowed us to keep the same relative value between increased consumption and deaths averted that emerged from our analysis of the beneficiary survey data.

  In 2021, we updated our estimate of the value of averting a death under five and over five from malaria, based on IHME data about the age distribution of malaria fatalities. In doing so, we kept our 2020 moral weights for deaths across each age. As a result, our current values do not average to 100.

• 9

  Changing the ratio of moral weights for consumption compared to deaths averted from 1:100 to 2:100 by doubling our estimate of the value of consumption would double the value of livelihoods programs and increase the value of death-averting programs by about 20% (because programs that address child mortality also generally improve childhood development, leading to higher later-life income and consumption). As a result, after changing the moral weights in this way, a death-averting program that we currently estimate to be 10x cash would be 12x cash (10 * 120%), and a livelihoods program that we currently estimate to be 6x cash would be 12x cash (6 * 200%). This is a simplification—in particular, if we reevaluated the moral weight of consumption increases, our estimate of the value of cash transfers would not stay static, as assumed in the above calculation.

• 10

  In January 2025, we recommended a ~$300,000 grant to IDinsight to review the literature on revealed preferences between consumption, health, and contraception, and to pilot new approaches to eliciting these trade-offs via discrete choice experiements in a low- or middle-income country. These experiments present participants with hypothetical questions (“how much would you pay for a vaccine that reduced the probability of your child dying by X%?”) and use the responses to assess their preferences.

• 11

  Life satisfaction scores attempt to quantify how individuals feel about their life overall, for example by measuring how people respond to the question “All things considered, how satisfied are you with your life as a whole these days?” on a scale of 1 to 10.

• 12

  Our analysis suggests that StrongMinds creates 17 life satisfaction point-years (or WELLBYs) per $1,000 spent whereas insecticide-treated nets create 70 life satisfaction point-years per $1,000 spent.

• 13

  Our takeaways in 2017:

  • "Governments and other prominent actors often use “value of a statistical life” estimates to compare the value of improving health relative to raising incomes. These estimates often imply that a year of healthy life is roughly two to three times as valuable as doubling someone’s income for a year. However, there is little relevant research to inform such estimates in low- and middle-income country contexts; we understand that how income is valued relative to health may shift when a population is much poorer.
  • There does not seem to be a standard approach for comparing the value of life at different ages; the most commonly used framework that we have seen (the disability-adjusted life year framework) explicitly does not provide judgments on this topic. Nevertheless, most other analyses that we have seen assume that averting death during childhood is about one to two times more valuable than averting death during adulthood.
  • Our initial analysis suggests that using relatively “standard” moral weight assumptions (i.e., the assumptions in the previous two bullet points) instead of our staff’s moral weights would not change our overall view of the relative cost-effectiveness of our current Top Charities."

  We’ve updated our moral weights since we published our 2017 write-up. However, the conclusions are similar.

  • At the time of the write-up, our best guess for other actors’ moral weights was 0.4 for doubling consumption, 37 for averting the death of a child under 5, and 30 for averting the death of a person over five. See here. If we assign a value of 1 to doubling consumption, that would imply a value of 93 for averting the death of a child under five, and 75 for averting the death of someone over five.
  • We currently assign a value of 1 to doubling consumption, 116 for averting the death of a child under five from malaria, and 73 for averting the death of someone over five from malaria. See here.