Valuing Contraception in GiveWell’s Cost-Effectiveness Analysis

Published: April 2025

Summary

What benefits of contraception do we include?

In our analysis, we focus on the following benefits of contraception:

• Improved health for women and children. Fewer unintended pregnancies, which includes mistimed pregnancies (i.e., pregnancies that occur sooner than the mother desired) and unwanted pregnancies (i.e., pregnancies that the mother did not desire to have at any time), may reduce maternal and child mortality and morbidity by making births safer and lowering the number of births per woman.
• Increased earnings for women. Fewer unintended pregnancies may allow women to spend more time on income-generating activities or education (which leads to increased earnings later).
• Increased resources for existing children. With fewer children in the household, each child may see higher consumption, more educational investments, and/or more parental attention.
• Improved subjective well-being for women. Contraception may reduce stress about potential pregnancies and increase women's sense of control over reproductive decisions.

We chose to focus on these benefits because:

• Our shallow review of surveys found that women in LMICs commonly cite these as key benefits of modern contraception.
• We think this approach is consistent with how we value the benefits of other programs we’ve funded or considered funding. (more)

How large are these benefits?

To quantify these benefit streams, we developed a rough model and sense-checked it against other approaches.

Our best guess is that providing a year of modern contraception in LMICs1 has ~0.7 units of value2 (25th-75th percentile range of ~0.3-1.1). This implies that averting an unintended pregnancy has ~2.3 units of value.3 (more)

This estimate suggests:

• A year of modern contraceptive use is as valuable as a ~60% increase in income for one year, averting a year of severe anemia, or averting a year of moderate anxiety disorder for two people
• Averting an unintended pregnancy is as valuable as doubling income for one person for two and a half years, averting a year of severe anemia for three people, or averting a year of a moderate anxiety disorder for six people. (more)

A breakdown of the modeled benefits is below. Our full estimates are in this spreadsheet.

	Units of value	25th percentile	75th percentile
Improved health for women and children	0.2	0.06	0.4
Increased earnings for women	0.1	0.003	0.2
Increased resources for existing children	0.2	0.04	0.3
Improved subjective well-being benefits for women	0.2	0.1	0.4
Units of value per year of modern contraception	0.7	0.3	1.1

This estimate is based on a shallow evidence review combined with subjective judgments to fill gaps. The evidence is relatively weak, often relying on underpowered studies or non-experimental designs with methodological limitations.

• Improved health for women and children. We estimate that each year of modern contraception leads to ~0.3 fewer unintended pregnancies. This estimate is based on experimental evidence from Bangladesh, Burkina Faso, Zambia, Malawi, and Kenya and a model using survey data from 17 low- and middle-income countries. (more) Among these averted pregnancies, we estimate that ~40% would end in abortion, ~15% would end in miscarriage, ~30% would end in mistimed births (i.e. births that occur sooner than the mother desired), and ~15% would end in unwanted births (i.e. births that were not wanted by the mother). These estimates are based on surveys of women in Africa. (more)

Averting these pregnancy outcomes leads to improved health through:

• Reductions in maternal mortality and morbidity (due to both fewer births overall and better-timed births).
• Reductions in neonatal and child mortality and morbidity (due to better-timed births).4

We estimate the maternal and child health benefits from better-timed births based on observational estimates of the effect of better-spaced pregnancies on these outcomes. We estimate the maternal health benefits of fewer overall births by combining estimates for low-income countries of the health risks of births from the United Nations Interagency Group for Child Mortality Estimation (UN IGME) and the Institute For Health Metrics and Evaluation (IHME). (more)

• Increased earnings for women. We estimate that each year of modern contraceptive use increases income by ~0.2% per year, or ~4% per unwanted birth averted. This is based on three studies of the long-term economic and educational effects of family planning programs in Colombia (a quasi-experimental study), Indonesia (a simulation-based study), and India (an experimental study). For studies on educational effects, we extrapolate from education to income by estimating the returns to education for women using observational data across LMICs and experiments in Colombia and Ghana. We assume the effect on income persists over ~15 years. (more)
• Increased resources for existing children. We estimate that averting an unwanted birth leads to existing children attaining ~0.5 years of additional education and earning ~3% more income once they begin working. This is based on quasi-experimental studies from India and Burkina Faso, combined with return-to-education data across LMICs. We apply a 25% upward adjustment to account for non-educational benefits like improved consumption in childhood and quality of life. We also sense-checked this approach by estimating the direct consumption benefits of fewer household members sharing the same resources, which yielded a similar value. (more)
• Improved subjective well-being for women. We estimate that each year of modern contraception adds the equivalent of ~0.1 disability-adjusted life years (DALYs) from improvements in subjective well-being, separately from the effect of improved health or increased earnings/resources on subjective well-being. This is roughly equivalent to averting a case of mild-to-moderate anxiety for a year. (more)

We compared these estimates with other evidence that directly measures how users in LMICs value different contraceptive options. These suggest a broad range of potential values (~0.02-3.3 units of value, compared to our best guess of ~0.7):

• Willingness to pay. We did an ad hoc review of estimates of willingness to pay for contraceptives in low-income countries and found ranges between 2% and 10% of per capita income. This suggests a weight of ~0.02-0.3 units of value per year of modern contraceptive use among users of contraception, which is lower than our estimate of ~0.7 units of value. (more)
• Subjective well-being studies. We reviewed two randomized trials that measure the effect of programs on uptake of contraception and an index of mental health measures. Comparing these to effects of cash transfers on similar measures suggests a value of ~1.1-3.4 units of value per year of modern contraceptive use. This is higher than our estimate of ~0.7 units of value. (more)

We’ve also compared our health estimates to the health estimates of the Impact 2 model, developed by MSI Reproductive Choices, a family planning charity. They estimate greater health benefits than our model, primarily due to different assumptions about unintended pregnancies averted and child mortality. We think this gap primarily reflects our more skeptical approach to observational data. (more)

We expect the magnitude of family planning benefits to vary across program location, the age of program participants, and program type. Our estimates across seven LMICs ranged from ~0.4 to ~1.0 units of value. We will account for this variance across contexts when evaluating specific programs. (more)

How could we be wrong?

• How to account for autonomy? Our current approach captures the autonomy benefits of contraception through its measurable effects. When contraception gives women more control over their reproduction, this leads to concrete improvements in health, income, and subjective well-being. However, this approach may fail to capture other benefits of autonomy, such as the intrinsic value of bodily autonomy beyond its effects on these outcomes, or long-term shifts in women's agency and gender norms caused by increased contraceptive use. In this initial analysis, we’ve focused on health, income, and subjective well-being because we think the broader benefits are difficult to quantify and compare consistently across programs (for example, we don't currently count broader societal benefits from lower child mortality or autonomy benefits from clubfoot surgery). We plan to investigate this further as we explore grants in family planning and continue to refine our moral weights.
• What would women in LMIC countries say if we asked them to trade off contraception or averting an unintended pregnancy against other outcomes? While we've incorporated some evidence from willingness-to-pay studies and subjective well-being measures, we haven't directly asked women in LMICs to make explicit trade-offs between contraception (or averted pregnancies) and other outcomes we fund. This kind of research could provide valuable information, though responses could be misleading if women are misinformed about the benefits of contraception or do not fully internalize the benefits to other family members (e.g., increased resources for existing children).
• How should we think about lives that don't occur due to contraception? We assign no value to potential lives that don’t occur due to contraception. This is an important choice. Even putting a small positive weight on these potential lives (e.g., ~20% of the value of an existing life) could make family planning programs a net negative. After consulting experts and stakeholders, we chose to assign no value because we think it aligns more with the views of program participants, staff, donors, and local decision-makers. We’re highly uncertain about this choice and don't think we have the objective "right" answer to these complex questions. (more)
• Are there additional benefits and downsides we should include? We exclude several potential effects of modern contraception, including benefits (e.g., cultural shifts toward gender equality, benefits for partners) and downsides (e.g., unintended side effects, disutility from partners opposing contraception). We exclude these primarily because we think they're small relative to included effects or would be inconsistent with our other program evaluations. Some excluded effects could be large but vary significantly across contexts, suggesting they're better evaluated case-by-case. We plan to revisit some of these potential benefits and downsides as we make grants and learn more about their magnitude in specific contexts. (more)
• How can we ensure programs aren't coercive? Family planning programs have a troubling history of coercion, and recent evidence suggests subtle coercion still exists in some programs. We plan to focus on programs that are unlikely to be coercive and assess the potential for coercion for each grant opportunity we investigate. (more)
• What is the true effect of access to contraception on the number of unintended pregnancies? This parameter is highly influential on our estimate since most of the modeled benefits flow through averting unintended pregnancies. The existing estimate is based on a weighted average of five statistically underpowered studies and a model that relies upon observational data. Since we are uncertain about the quality of this evidence, well-powered experiments in relevant contexts for our grantmaking could substantially change our estimate of this parameter. (more)
• How should we quantify subjective well-being benefits? This is a challenging topic, because it is difficult to separate subjective well-being effects from other benefits (e.g. how much of the effect on subjective well-being is downstream of having higher income and better health?) and empirical evidence is limited. (more)
• What about programs that influence whether a woman wants to use modern contraception? Some programs aim to increase participants’ desire to use modern contraception. For these programs, should we consider a person’s “true” preference to be their pre-program lack of desire for modern contraception or their post-program desire for contraception? For example, Development Media International ran a radio campaign that likely increased demand for modern contraception by correcting misperceptions about modern contraception.5 In this case, we think that the program is increasing participants’ ability to make an informed choice, so treat women’s post-program preferences as their “true” preferences. However, we recognize that this is a fuzzy criterion and hope to refine it over time.

Next steps and implications for our grantmaking

We plan to investigate family planning programs that are plausibly above our cost-effectiveness bar. Based on our current estimate, programs providing a year of modern contraception for under ~$20 would meet our bar. Given our uncertainty, we'll also consider higher-cost programs that offer learning value or where specific contexts suggest higher benefits. If we find many promising opportunities, we may conduct further research, including asking potential program participants how they value family planning outcomes compared to other interventions we fund. (more)

1. Why are we looking into this question?

According to the United Nations, 8% of women globally (165 million women) and 16% of women in low-income countries (30 million women) have an unmet need for family planning, meaning that they are fecund, sexually active, and do not want to have a pregnancy, but are not using any contraceptive method.6 7 Unmet need for family planning may result in unintended pregnancies that negatively affect a family’s health, wealth, and subjective well-being.

According to responses to Demographic and Health Surveys (DHSs),8 reasons for not using modern contraception among women in LMICs who wish to avoid pregnancy include concerns about side effects, low perceived risk of conception due to infrequent sexual activity, personal disapproval of contraceptives or the fear of disapproval of others (including partners), and lack of knowledge about or access to contraceptives.

In the next year, GiveWell is considering whether to recommend funding family planning programs that aim to remove barriers to modern contraception in low- and middle-income countries.9 This may include programs that deliver family planning services and contraceptives to communities with limited existing access, or mass media campaigns that try to address misconceptions about contraceptives.

GiveWell relies heavily on cost-effectiveness analysis to decide which programs to recommend funding. To assess the relative cost-effectiveness of family planning programs, we need to estimate the value created by increasing contraceptive uptake, so we can compare the benefits created by funding family planning to the benefits created by funding other programs that we might recommend.10

This report provides a preliminary analysis of how we think about valuing the benefits of family planning programs.

2. What benefits of contraception do we consider? Which effects do we exclude?

2.1 Which benefits do we include?

We think that increased uptake of contraception as a result of family planning programs could lead to several benefits.

The main benefits we model are:

• Improved health for women and children through reduced maternal and child mortality and morbidity from fewer and better-timed births.
• Increased earnings for women from more time for education and income-generating activities.
• Increased resources for existing children from fewer births, leading to better educational outcomes and future earnings.
• Improved subjective well-being for women from reduced pregnancy-related stress and greater reproductive autonomy.

We focus on these benefits because:

• We think these align with reasons given by women in low- and middle-income countries for using contraception. We conducted a shallow review of studies that asked users of contraception in LMICs why they use contraception. In those studies, the primary reasons for using contraception were:
  • Ensures adequate resources to provide for existing children's needs, including food, clothing, and education, leading to better outcomes for existing children. This is captured in benefits from increased resources for existing children.11
  • Enables mothers to allocate more time to economic and educational activities. This is captured in increased earnings for women.12
  • Promotes positive maternal and child health outcomes through increased birth spacing. This is captured by improved health for women and children.13
  • Improves emotional well-being by reducing the burden of pregnancy and uncertainty, and strengthens marital relationships by reducing conflicts related to childcare and sexual abstinence. This is captured in improved subjective well-being for women.14

The first three points were also broadly consistent with high-level findings from surveys conducted by Family Empowerment Media with its listeners in northern Nigeria.15

• We think these benefits are broadly consistent with benefits we model for other programs.
  • We include health benefits (reduced mortality and morbidity) for child-death averting programs (e.g., our Top Charities) and other programs (e.g., iron fortification).
  • We include benefits from increased earnings for health-focused programs,16 as well as for programs focused more directly on increasing income (e.g., eyeglasses promotion or seedlings).
  • We include subjective well-being effects for programs to improve mental health (e.g. the benefits of reduced anxiety and depression). In addition, we use disability weights for some programs (e.g. clubfoot treatment) which includes both physical and non-physical harms of clubfoot (e.g., worry).17

One potential inconsistency is that, when modeling averted child deaths, we do not consider the reduced resources of other household members from having one more child, but when modeling averted births, we consider the increased resources from having one fewer child. We do this because:

• Families report increased resources from averting births as one of the primary benefits of averting births.18 In contrast, a survey conducted by IDinsight asked low-income individuals in Ghana and Kenya about the economic effects of a household member’s death, and they did not cite increased resources as a primary effect. Instead, they described funeral costs or spending time away from income-generating activities due to the funeral/grief as having negative economic effects.19
• We think there is a difference between an existing child dying and averting an unintended birth.20
• Increased resources from averting a birth substantially increase the value of contraception (~30% of total benefits), while we estimate that decreased resources from averting a death would only slightly decrease the value of life-saving interventions (<5% of total benefits).21

We have several key uncertainties about which benefits to include. We discuss these in the sections below.

2.2 What about lives that don’t occur due to contraception?

Increased uptake of modern contraception may lead to fewer children born overall.22 How much should we value those lives that don’t occur as a result of contraception?

This is a challenging question, so we solicited feedback from a range of experts.23

Our understanding from these conversations is that there’s no philosophical consensus on how to value potential lives,24 and this gets especially complex when considering unwanted births where valuing potential lives could conflict with respecting autonomy.25

We currently do not assign any value, positive or negative, to these potential lives. We think this stance is consistent with our approach to other normative questions (e.g. our moral weights), where we consider the views of GiveWell’s staff, donors, and program participants.26 That’s because:

• Most GiveWell staff and donors we spoke with were comfortable with this approach, while acknowledging the deep uncertainty involved. Many mentioned respecting reproductive autonomy as a rationale for not placing a value on lives that don’t occur.27
• A relatively high number of women and men in LMICs desire to have no more children.28 While this doesn’t mean they assign neutral value to lives that don’t exist as a result of contraception, it does suggest this consideration is not enough to fully override the perceived benefits of avoiding unwanted pregnancies.
• Local governments29 and organizations like WHO30 support family planning programs, indicating that decision-makers in these contexts (at least implicitly) don't see prevented births as a major downside, even if they don’t take an explicit stance on the issue.

However, we’re uncertain about this, and this decision has significant implications for our assessment of family planning programs. Valuing lives that don’t occur as a result of contraception as 20% as valuable as existing lives would make family planning programs net negative in our model.31

We think there are several ways we could be wrong:

• We spoke with some individuals who argued that potential lives should count equally to existing lives, or that we should at least give them some weight given our uncertainty because:
  • These would be real people with experiences, relationships, and contributions to society.
  • Many religions and ethical frameworks view potential human life as inherently valuable. For example, from a total utilitarian perspective, future individuals' welfare matters equally to current individuals.
  • Choosing not to fund family planning programs does not conflict with respecting autonomy because we are only deciding not to help people fulfill their preferences rather than actively preventing them from fulfilling their preferences.
  • Even if we are uncertain about the value of lives that don’t occur due to contraception, we should apply a discount for moral uncertainty rather than not valuing these lives at all. This approach would likely lead us to value lives that don’t occur due to contraception at some fraction of our value of averting a death.
• We have not deeply investigated the views of individuals in LMICs on this question. It’s possible that significant numbers of individuals in LMICs believe preventing potential lives through contraception has negative value, or we are missing additional considerations that could come from deeper engagement with these individuals.
• We have spoken to a small share of our donors. It’s possible many donors view preventing potential lives as a major downside of contraception and that we should reconsider our approach or offer alternative funding options (e.g., offering “funds” that exclude family planning programs for donors who are opposed to this).

Going forward, we may do additional work to further investigate how program participants and donors view this issue or conduct more systematic surveys on these questions.

For transparency, we’ve set up our model such that individuals who disagree with us can input different values for lives that don’t occur due to contraception to see how it affects the results (see here). We’re also interested in feedback from those who disagree with our stance.

2.3 What other benefits or downsides are we excluding?

In addition to the key benefits we focus on in our analysis, there are several other potential upsides and downsides to providing access to modern contraception that we do not explicitly account for in our cost-effectiveness estimates.

Potential upsides:

• Autonomy. We aim to capture the improved autonomy that results from contraception through instrumental benefits (i.e., contraception leads to fewer unintended pregnancies, which leads to improvements in health, income, and subjective well-being). This approach does not include an intrinsic value of autonomy that is independent of these downstream consequences. Several donors, staff, and experts thought that we should include an intrinsic value of autonomy. Currently, we are not including it, because we did not have a reliable method of quantifying the intrinsic value of autonomy and thought it might be inconsistent with our approach to other programs (e.g., we do not value the increased autonomy from treating clubfoot).32
• Potential "demographic dividend" effects on economic growth due to changes in the age structure of the population. We exclude this because we do not include these considerations in our analyses of other interventions (e.g., child death averting programs). We also are uncertain about how changes to a population’s age structure and/or size affect macroeconomic growth in the settings where GiveWell is likely to fund family planning.
• Possible contributions to environmental sustainability due to reduced population growth. We exclude this because we do not include these considerations in our analyses of other interventions (e.g., child death averting programs). We also think these benefits would be challenging to quantify and may be offset by the contribution of a higher population to scientific progress.
• Potential cultural shifts towards greater gender equality and women's empowerment. We view these macrosocial shifts as a potential benefit of family planning but exclude them because we do not include similar considerations for death-averting programs (e.g., broader improvements in society from lower child mortality). In addition, we would expect these benefits to vary across contexts, so coming up with a uniform value for the effect of a given year of modern contraception in a LMIC seems relatively intractable. If there is a plausible case that a specific grant decision hinged on its ability to cause these cultural shifts, we expect to revisit this issue.
• Improving the quality of one’s long-term partner. Having an unintended pregnancy may cause a woman to end up marrying a “worse” partner than they would otherwise. Marrying a “better” partner could lead to a number of benefits (e.g. higher subjective well-being, higher partner income, etc.) that we do not include in our model.33 We do not include this potential benefit because we were unsure if these benefits would come at the expense of other women who would end up with “worse” partners.
• Possible benefits for men, who may also benefit from a year of modern contraceptive use. We assume benefits are concentrated among women.

Potential downsides:

• Potentially lower per capita economic growth due to a smaller population. Many macroeconomic models imply that a larger population can lead to greater economic growth due to idea generation (more people means more ideas) and scale effects.34 As with sustainability and demographic dividend effects, we do not include this factor in our model due to worries around inconsistency with our other cost-effectiveness models and the intractability of quantifying these effects in our context.
• Benefits of additional children, such as increased care for parents or other family members in old age or increased earnings for the household later in life. Given that women report that averting unintended pregnancies would help their households' economic situation,35 we assume that these longer-run benefits are offset by short-run economic hardships.
• Parents experience increases in subjective well-being from children, even those that were initially unwanted, over time. Empirical work finding increases in subjective well-being for women using modern contraceptives suggests that this consideration is outweighed by other effects on subjective well-being (more), though we have not looked into the magnitude of this effect in detail. In addition, the subjective well-being measures we rely upon may not adequately capture the benefits to the parent of having the child.
• Other children in the family experience increases in subjective well-being from having a sibling. We have not investigated this. It seems plausible that other children enjoy having another sibling, though it also seems plausible that more siblings could lead to lower subjective well-being due to reduced attention from parents or reduced resources generally.
• Disutility from partners or other family members who oppose contraception. We think that this disutility is likely to be small relative to other considerations, but we have not looked into this in-depth.
• Potential reduction in happiness from concealed uptake of contraception. The studies finding increased subjective well-being (more) suggest that this effect is outweighed by other factors, but this may be more of a concern for specific program designs.36
• Unintended side effects of using modern contraceptives. Side effects of some modern contraceptive methods may be unacceptable for certain populations. (e.g. bleeding from implants could be unacceptable for Muslim women due to daily prayers).37 We expect to support programs that enable women to make an informed choice among contraceptive options. These programs should allow women to choose a modern contraceptive method with minimal side effects for them. Thus, we expect that including side effects in our estimate would not substantially change our estimate of the value per year of modern contraception provided by these programs, though we plan to investigate this more for specific grant opportunities.

2.4 What about coercive family planning programs?

There is a troubling history of family planning programs coercing individuals into using contraceptives. We plan to fund only opportunities that enable voluntary uptake of modern contraceptives, because we believe that coercive family planning programs are unethical and counterproductive.38

In addition to focusing on non-coercive programs, we will investigate and monitor for subtle coercion. Recent qualitative research suggests that some family planning program staff in Africa subtly pressure women into using long-acting reversible contraceptives (LARCs).39 To avoid funding subtly coercive programs, we plan to investigate potential grantees’ practices and, when possible, support monitoring systems to identify and minimize subtly coercive practices. We will be especially careful when funding LARCs given the evidence of subtle coercion in past LARC-focused projects.

3. How large are the benefits of contraception we model?

To try to quantify the different benefit streams (improved health for women and children, increased earnings for women, increased resources for existing children, and improved subjective well-being), we developed a rough model that estimates the value of providing a year of modern contraception to a woman in an LMIC. (more)

We’ve also tried to sense-check this model against other benchmarks: willingness-to-pay for modern contraceptives, studies of the effect of modern contraceptives on subjective well being, the value of benefits caused by other programs we fund, and other models of the value of contraception. (more)

3.1 Rough quantitative model

To model the benefits of modern contraception:

• We estimate the number of unintended pregnancies averted by a year of modern contraception. We estimate that each year of modern contraception averts ~0.3 unintended pregnancies. (more) This includes ~0.1 abortions, ~0.05 miscarriages, ~0.1 mistimed births, and ~0.05 unwanted births. (more)
• We then project the benefits of averted pregnancies for women and their household. This leads to the following benefits:

Unintended pregnancies averted by a year of modern contraception (more)	0.30
Benefit streams:	Units of value
Improved health for women and children
Reduction in maternal mortality due to fewer unwanted births, miscarriages, and abortions (more)	0.04
Reduction in maternal mortality due to fewer mistimed births (more)	0.01
Reduction in maternal morbidity due to fewer unintended pregnancies (more)	0.04
Reduction in child mortality due to fewer mistimed births (more)	0.09
Reduction in child morbidity due to fewer mistimed births (more)	0.006
Reduction in healthcare spending due to improved woman and child health (more)	0.04
Increased earnings for women
Increases in household income through increased economic activity for women (more)	0.05
Increased resources for existing children
More resources for existing children due to fewer unwanted births (more)	0.19
Improved subjective well-being for women
Improvements in subjective well-being not captured by improvements in health or economic outcomes (more)	0.23
Units of value per year of modern contraception	0.70
Units of value per unintended pregnancy averted	2.3

Our full model is in this spreadsheet. The sections below provide an overview of the model and evidence we use to estimate the benefits of contraception.

We have relatively high uncertainty around our estimate of the units of value per year of modern contraception (25th percentile: 0.3 units of value; 75th percentile: 1.1). Our largest source of uncertainty is our estimate of unintended pregnancies averted per year of modern contraception, because many of the benefits flow through averted unintended pregnancies and we have imperfect evidence for this estimate. Otherwise, the uncertainty is relatively evenly distributed across the benefits, though we have relatively more uncertainty on the economic benefits for women and the benefits for existing children, due to the lower-quality evidence for these benefits.40

Unintended pregnancies averted by a year of modern contraception

We think the primary benefit of modern contraception is to avert unintended pregnancies. We define unintended pregnancies as births that occur two or more years sooner than desired (“mistimed”) and births that were not wanted at all by the mother (“unwanted”).41 We think unintended pregnancies bring health risks, disrupt income-earning activities and human capital investments, and lower subjective well-being (which we model in the sections below).

As a result, a key parameter in our model is the effect of modern contraception on the number of unintended pregnancies per year among women who take up contraception due to a Givewell-funded program (i.e., women who would not have used modern contraception in the absence of our program, but would use modern contraception if the program existed).

Our best guess is that a year of modern contraception leads to ~0.3 unintended pregnancies averted per year of modern contraception.

We used two approaches to reach this estimate:

• Model using survey data from LMICs. We estimate the difference in unintended pregnancies between the average user of modern contraception and the average non-user of modern contraception with a need for family planning in a sample of 17 LMICS, using data from the Guttmacher Institute’s Adding It Up project.42 We adjust this difference downward by ~15%, because women who take up contraception due to a Givewell-funded program are likely to differ from the average user of modern contraception and the average non-user of modern contraception with need for family planning. (e.g., they are likely more motivated to avoid pregnancies than the average non-user with unmet need). (more)
• Estimates from studies of family planning in LMICs. We take a weighted average of empirical estimates of the effect of modern contraception on pregnancies among women who choose to use modern contraception as a result of family planning programs in LMICs. We adjust these effects downward by ~20% to account for internal validity (e.g., one of the studies we rely on, the “Matlab study,” bundles several interventions so likely overstates the effect of contraception alone) and external validity (e.g., some of the studies are in settings with higher fertility rates than LMICs today, which we guess leads to higher effects). (more)

Both approaches yield a similar effect. You can see detailed calculations in this spreadsheet.

Approach	Value
Model using survey data from low- and middle-income countries (more)	0.34
Model using survey data from low- and middle-income countries (more)	0.27
Average	0.30

As a sense-check, we calculate the number of lifetime averted pregnancies and births implied by our estimate. Our best guess estimate implies that modern contraception would reduce lifetime unintended pregnancies by 4.6.43 Using the breakdown of unintended pregnancy outcomes described in Types of unintended pregnancies averted, this implies that using modern contraception through one’s reproductive years would lower lifetime fertility by ~0.8 births.44 This estimate aligns with our impression that low usage of modern contraception is one of the drivers of high fertility rates, but not the only one (e.g. desired number of children, levels of economic development, child mortality, and women’s empowerment are also likely to affect fertility rates).

Model using survey data from LMICs

Our first approach is to compare unintended pregnancy rate among modern contraceptive users and non-users45 with unmet need in LMICs. The main downside to this approach is that the women who take up contraception due to a Givewell-funded program are likely to differ from the typical user of modern contraception and the typical non-user with unmet need. We try to adjust our estimates to account for these differences, but we are highly uncertain about the magnitude of those adjustments.

This table shows our estimates before and after adjusting for potential differences between women who take up contraception due to a Givewell-funded program and the average user or nonuser with unmet need:

Parameter	Estimate
Unintended pregnancy rates without modern contraception (unadjusted)	0.43
Unintended pregnancy rates with modern contraception (unadjusted)	0.04
Unintended pregnancies averted per year of modern contraception (unadjusted)	0.40
Unintended pregnancy rates without modern contraception (adjusted)	0.38
Unintended pregnancy rates with modern contraception (adjusted)	0.04
Unintended pregnancies averted per year of modern contraception (adjusted)	0.34

The unadjusted estimate is the average46 unintended pregnancy rate among users of modern contraception in 17 LMICs minus the average unintended pregnancy rate among non-users of modern contraception in the same 17 LMICs.47 These estimates come from the Guttmacher Institute’s Adding It Up project,48 which produces estimates of unintended pregnancy rates from 2015-19 using survey data49 from LMICs. While we have some reservations with their approach, we think these are the best available estimates of unintended pregnancy rates by contraceptive use for LMICs.50 51

We apply adjustments to these estimates to account for the fact that women who take up modern contraception due to GiveWell-funded family planning programs are likely to differ from the average user of modern contraception or non-user with unmet need.52

These differences could affect the impact of modern contraception on unintended pregnancies in three ways:

• Contraceptive methods used. In the absence of the program, we assume that women who take up modern contraception due to GiveWell-funded programs would use traditional methods at 20% higher rates than the typical non-user with unmet need. This captures that those most motivated to avoid unintended pregnancies are likely to be disproportionately represented among women who take up modern contraception due to GiveWell-funded programs.53 54
• Failure rates of contraception methods. We assume that women who take up modern contraception due to GiveWell-funded programs have 10% higher failure rates when using modern contraceptives relative to the typical user of modern contraception.This adjustment is meant to capture that the women who take up due to our programs are likely to be more inexperienced using modern contraception than the average user.
• Risk of unintended pregnancy. Women who are technically not in need of family planning may still adopt modern contraceptives due to GiveWell’s programs. The Guttmacher Institute includes women who are “experiencing postpartum amenorrhea after an intended pregnancy” as women not in need of family planning,55 but these women experiencing postpartum amenorrhea adopt modern contraceptive methods fairly frequently.56 We roughly estimate that ~10% of women who take up modern contraception due to GiveWell-funded programs will be women not in need of family planning.57

Estimates from studies of family planning programs in LMICs

Our second approach is to combine estimates from empirical studies of family planning programs.

An advantage of this approach is that these studies estimate effects on the women who take up modern contraception due to family planning programs. This subset of women should be a closer match for the women who take up modern contraception due to GiveWell-funded programs than the average user of modern contraception or the average non-user with unmet need.

Downsides of this approach include that the studies are often underpowered (leading to noisy estimates), have internal validity issues (e.g. bundling family planning with maternal and child health interventions), or have external validity issues (e.g. conducted many years ago when fertility rates were higher).

We identified five family planning programs in LMICs with studies evaluating the effects on both modern contraceptive usage and pregnancies/births in a moderate depth literature review.58

We take a weighted average of these five programs to estimate the effect of modern contraception on unintended pregnancies:

Studies	Estimate of unintended pregnancies averted per year of modern contraception	Weight
Matlab experiment (Bangladesh)	0.59	0.41
Glennerster et al. (2023) experiment (Burkina Faso)	0.45	0.25
Karra et al. (2022) experiment (Malawi)	0.31	0.06
Kosgei et al. (2011) retrospective cohort study (Kenya)	-0.09	0.24
Ashraf et al. (2014) experiment (Zambia)	-0.57	0.04
Weighted average (unadjusted)	0.33
Adjustment for internal validity of studies	-10%
Adjustment for external validity of studies	-10%
Weighted average (adjusted)	0.27

For each included program, we divide the effect on unintended pregnancies by the effect on years of modern contraception usage to get an estimate of unintended pregnancies averted per year of modern contraception. This estimation procedure assumes that all of the effect of the program on unintended pregnancies comes through modern contraception usage rather than e.g. changes in norms and attitudes.

The weight on each study is proportional to the number of women who chose to adopt modern contraception as a result of the program. This captures the fact that studies where more women are caused to use modern contraception have more statistical power to estimate the effect of modern contraception on unintended pregnancies. A potential weakness of our approach to weighting the programs is that we do not account for clustering of treatment assignment or study quality.

We downwardly adjust our estimate by 10% for internal validity concerns and 10% for external validity concerns. This is mainly due to concerns with the evaluation of the “Matlab program” a “maternal and child health and family-planning program that was implemented in Matlab, Bangladesh in 1977.”59 The Matlab program had the highest estimate of unintended pregnancies averted per year of modern contraception and has the most weight of any study in our analysis.60

• Details of the Matlab experiment: The program ran from 1977-1996 in Matlab, Bangladesh and involved community health workers visiting married women of childbearing age every two weeks. The community health workers advised women on modern contraceptive use, provided modern contraceptives, and referred women to local clinics/hospitals when necessary.61 Outcome measurements covered 282 villages62 and ~5000 women.63 The program increased modern contraceptive use by ~28 pps among women.64
• Internal validity concerns: It’s difficult to isolate the impact of modern contraception, because the program bundled family planning services with other maternal and child health services. This means that the reduction in births could partially be a result of these other services rather than modern contraceptive usage.65 In addition, the potentially nonrandom selection of intervention and comparison areas could influence the validity of the results.66
• External validity concerns: The Matlab experiment was conducted in Bangladesh in 1977 when fertility rates were ~50% higher than the rates in low-income countries today.67 While some of this difference is driven by users of modern contraception (i.e. not our target population), it is likely that even non-users have lower fertility and unintended pregnancy rates in low-income countries today than in Bangladesh in 1977 due to changes in other determinants of unintended pregnancies.68 This would mean that unintended pregnancies averted from a year of contraception would be lower in low-income countries today than in Bangladesh in 1977.

Refer to this sheet for more details on the five programs and potential internal and external validity concerns with the evaluations of these programs.

Uncertainties about unintended pregnancies averted per year of modern contraception

We are relatively uncertain about this estimate because of the lack of high-quality evidence. It is plausible that this parameter is substantially higher (75th percentile is 0.4) or lower (25th percentile is 0.2).

• We think it is unlikely that the effect is higher than 0.4 because this would imply very large effects on lifetime unintended pregnancy rates that seem unlikely to us.69
• We think it is unlikely that estimates are below 0.2, because this would imply implausibly widespread usage of traditional contraceptive methods (e.g. withdrawal or periodic abstinence).70

Since most of the benefits flow through averting unintended pregnancies, being wrong about this estimate would substantially influence our overall value of modern contraception. Our 25th percentile estimate (0.2) results in an estimate of 0.6 units of value per year of modern contraceptive use. Our 75th percentile (0.4) results in an estimate of 0.8 units of value per year of modern contraceptive use.

Key drivers of this uncertainty include:

• What populations are likely to be served by family planning programs? Whether women who adopt modern contraception have relatively low or high unintended pregnancy rates is the main factor determining the number of unintended pregnancies averted. Unfortunately, we did not come across much evidence on this question. In the future, we may explore this question by looking into empirical studies or the characteristics of women who begin using modern contraception in panel data sets.71
• Do women misreport their pregnancy outcomes? The survey-based estimates are based on women’s retrospective reports about their births. There is some evidence that “women who say they do not want to become pregnant but then do become pregnant and have a child will later report that the pregnancy was intended” (Sedgh, Singh, and Hussain, 2014, p. 8). Staveteig (2017) found that women who took a DHS survey in Ghana often reported different fertility preferences in a follow-up interview. Valente et al. (2024) find that women in Nigeria give different responses to direct and indirect survey questions. We have not reviewed this evidence in-depth. In the future, we may adjust our estimate if evidence suggests that women’s self-reports of pregnancy outcomes appear to be systematically biased.

We have several other (less critical) uncertainties too.72

Types of unintended pregnancies averted

The effects of an unintended pregnancy are likely to differ by the outcome of the unintended pregnancy. We estimate the number of abortions, miscarriages, unwanted births (meaning the mother did not want to have a child), and mistimed births (meaning the mother wanted a child eventually but not in the next two years).

Unintended pregnancies averted
Unintended pregnancies averted per year of modern contraception	0.30
Breakdown of unintended pregnancies:
% of unintended pregnancies that result in abortion	37%
% of unintended pregnancies that result in miscarriage	14%
% of unintended pregnancies that result in unwanted or mistimed birth	49%
of those, % that are unwanted	33%
Number of abortions averted per year of modern contraception	0.112
Number of miscarriages averted per year of modern contraception	0.042
Number of mistimed births averted per year of modern contraception	0.099
Number of unwanted births averted per year of modern contraception	0.049

Our breakdown of unintended pregnancies is based on Sedgh, Singh, and Hussain (2014) and Bearak et al. (2020). In a shallow review of the literature, Sedgh, Singh, and Hussain (2014) was the only paper with estimates for each pregnancy outcome of interest to us.73 Sedgh, Singh, and Hussain (2014) only provide estimates at the regional-level. We use their estimate for Africa, because we thought this would be the best proxy for the likely locations of GiveWell-funded programs. For % of unintended pregnancies that result in abortion, we use Bearak et al. (2020)’s estimate that 37% of unintended pregnancies in sub-Saharan Africa end in abortion. We have not thoroughly vetted Sedgh, Singh, and Hussain (2014) or Bearak et al. (2020)’s estimates.

Main uncertainties:

• Have there been substantial changes in pregnancy outcomes from unintended pregnancies over the last decade? Since the estimates from Sedgh, Singh, and Hussain (2014) are from 2012, we are not accounting for any changes over the past decade when we use these estimates.74
• What percentage of pregnancies end in miscarriages? In Sedgh, Singh, and Hussain (2014), “miscarriages are estimated to equal approximately the sum of 20 percent of live births and 10 percent of induced abortions.” (p. 5) Alternatively, Dugas and Slane (2022) mention that “it is estimated that as many as 26% of all pregnancies end in miscarriage and up to 10% of clinically recognized pregnancies.” We have not looked into this issue in-depth so are unsure why these estimates differ.

Improved health for women and children

By allowing women to control their fertility, modern contraception should improve maternal and child health. Maternal mortality (more) and morbidity (more) should be lower as women are able to avoid unwanted pregnancies and make the pregnancies they do have safer by having well-timed rather than mistimed pregnancies. For child health improvements, we only consider the benefits of replacing mistimed births with well-timed births (more), because if an unwanted birth is averted, we do not expect the child to exist and do not place a positive or negative value on lives that don’t occur due to contraception.75 We also account for the reductions in health spending due to improved maternal and child health (more).

Maternal mortality averted due to fewer unintended pregnancies

By averting unintended pregnancies, modern contraceptives decrease a woman’s risk of dying due to a pregnancy. This decreased risk comes from reducing the number of pregnancies across a woman’s lifetime (more) and improving the timing of a woman’s pregnancies (more).

Reducing the number of pregnancies and maternal mortality

We start with estimating the effects of reducing the number of pregnancies over a woman’s lifetime. We assume that averting unwanted births, abortions, and miscarriages reduces a woman’s number of pregnancies in their lifetime by one. That’s because we assume that averting these pregnancy outcomes does not affect the number of children a woman wants to have later in life.76 In contrast, we assume that averting a mistimed birth does affect the number of children a woman wants to have later in life. That’s because women who avert a mistimed birth will still want to have the child later, meaning that this later birth will offset the effect of averting a mistimed birth on lifetime births.77

Since maternal mortality risk varies by pregnancy outcome, we separately calculate the maternal mortality rate for abortions, miscarriages and live births.

Reduction in maternal mortality due to fewer unwanted births, miscarriages, and abortions
Number of unwanted births averted per year of modern contraception	0.049
Number of abortions averted per year of modern contraception	0.11
Number of miscarriages averted per year of modern contraception	0.042
Maternal mortality ratio (Ratio of maternal deaths to live births)	0.0035
Maternal deaths per birth	0.0031
Maternal deaths per abortion	0.0011
Maternal deaths per miscarriage	0.0007
Maternal deaths averted per year of modern contraception	0.0003
Units of value per maternal death averted78	125
Units of value from less maternal mortality due to fewer unwanted births and abortions per year of modern contraception	0.04

We calculate maternal deaths per pregnancy outcome by using the maternal mortality ratio in low-income countries (~350 per 100,000 live births79 ) and the percentage of maternal deaths attributable to each pregnancy outcome.80

The main uncertainty with this estimate is the validity of our assumption that averting unwanted births, abortions, and miscarriages do not affect a woman’s subsequent fertility decisions. For example, if they increased the likelihood of subsequent pregnancies, then the health gains from averting an unintended pregnancy might be offset by a later, intended pregnancy. It is plausible that these experiences affect a woman’s attitude toward child-bearing, but we are unsure of the direction of the effect. We have not looked for or come across any empirical evidence on this, so do not have a strong enough opinion to justify an adjustment to our estimate.81

Improving the timing of pregnancies and maternal mortality

When a woman controls the timing of their births through modern contraception, she can avoid giving birth when it would be unsafe for her (e.g. the birth is short-spaced, she has few resources, or she is very young).

We calculate the maternal deaths averted due to births at a woman’s preferred time being safer than mistimed births below.

Reduction in maternal mortality due to fewer mistimed births
Number of mistimed births averted per year of modern contraception	0.099
Maternal deaths per birth	0.0031
Reduction in maternal mortality due to improved birth timing	27%
Maternal deaths averted per year of modern contraception	0.00008
Units of value per maternal death averted82	125
Units of value from less maternal mortality due to fewer mistimed births per year of modern contraception	0.01

We assume that the effect of better-timed births on maternal mortality is roughly equivalent to the effect of modern contraception on maternal mortality. We base our estimate on the effect of modern contraception on maternal mortality in Cleland et al. (2012).83 Cleland et al. (2012) estimated the change in maternal mortality associated with changes in the contraceptive prevalence rate using nationally-representative data from 40 countries.84 Their estimate implies that contraceptive use decreased the maternal mortality rate by 67%.85 We adjust this estimate downwards by 60% to account for the fact that country-level increases in contraceptive prevalence are likely to be positively correlated with other changes that would reduce maternal mortality (e.g. women's economic empowerment or improvement of public healthcare provision).86

We are uncertain about the magnitude of these benefits, but these benefits are a relatively small percentage of the overall benefits of modern contraception, meaning that they are unlikely to influence our bottom line.

Main uncertainties:

• Is the effect of modern contraception a good proxy for the effect of improved timing of the birth? The effect of modern contraception on maternal mortality compares all births when using modern contraception to all births when not using modern contraception. Since modern contraception only improves the timing of a fraction of births, this comparison might understate the effect of improving birth timing on maternal mortality.
• Is our adjustment for confounding factors too large or too small? We only have cross-country observational data for this estimate and we are unsure about the degree of confounding.

Maternal morbidity averted due to fewer unintended pregnancies

In addition to maternal mortality, unintended pregnancies are also associated with increased prevalence of various health conditions for women.87 To quantify the benefits from averting these conditions, we make subjective judgements on the DALYs lost due to a given pregnancy outcome based on the evidence on the health risks associated with each outcome.

Reduction in maternal morbidity due to fewer unintended pregnancies
Units of value per DALY	2.3
Number of unwanted births averted per year of modern contraception	0.049
DALY loss from unwanted births	0.11
Units of value from lower maternal morbidity due to fewer unwanted pregnancies	0.01
Number of mistimed births averted per year of modern contraception	0.099
DALY loss from mistimed births	0.02
Units of value from lower maternal morbidity due to fewer mistimed pregnancies	0.004
Number of abortions averted per year of modern contraception	0.11
DALY loss per abortion	0.05
Units of value from lower maternal morbidity due to fewer abortions	0.01
Number of miscarriages averted per year of modern contraception	0.04
DALY loss per miscarriage	0.04
Units of value from lower maternal morbidity due to fewer miscarriages	0.003
Additional morbidity benefits	30%
Units of value from lower morbidity due to fewer unintended pregnancies per year of modern contraception	0.04

We estimate DALYs lost per pregnancy outcome:

• We estimate 0.11 DALYs lost per unwanted live birth. Following the approach taken in the Maternal Health Initiatives (MHI)’s CEA, we add up the DALYs lost from common health conditions caused by pregnancy.88 This leads to an estimate of ~0.08, which we adjust upward by 25% to account for the fact that we are likely missing some health conditions caused by pregnancy.
• We estimate 0.02 DALYs lost per mistimed births. This is based on Bauserman et al. (2020) Table 4, which shows that short-spaced pregnancies are associated with ~30% higher rates of adverse delivery outcomes. We adjust this estimate downward by 40% to account for confounding factors since Bauserman et al. (2020) is observational, and not all mistimed pregnancies are short-spaced.
• We estimate 0.05 DALYs lost per abortion. This is based on Higashi et al. (2014) which uses estimates from IHME’s Global Burden of Disease project. They report 0.12 DALY loss per abortion with 55% coming from maternal deaths.89 To avoid double-counting DALYs from maternal deaths, we adjust the .12 DALY estimate downward by 55%.
• We estimate .04 DALYs lost per miscarriage. We could not find an attempt to quantify the DALY loss due to miscarriages.90 In lieu of rigorous evidence, we assume that DALY loss from miscarriages is two thirds of the DALY loss from abortion. This assumption is based on the suggestive evidence that miscarriages pose less health risk to the mother than abortions.91

We adjust our overall estimate upward by 30% to account for near-misses of severe maternal health outcomes and the disability associated with being pregnant.

Overall, we are highly uncertain about this estimate since we do not have reliable, rigorous estimates of DALYs lost per pregnancy outcome. However, these morbidity benefits are small relative to other benefits, so errors in this estimate are unlikely to change our bottom-line.

Main uncertainties:

• Is our analysis missing important conditions?
• How has the DALY loss from abortion changed since 2015? For example, has the proportion of unsafe abortions gone up or down? We also did not have time to review Londoño-Vélez and Saravia (2025), a recent paper that seems to imply relatively large effects of unsafe abortions on women’s health.
• What are the health conditions associated with miscarriages? Do they imply a higher number of DALYs lost per miscarriage than we estimated?

Child health improvements due to fewer mistimed births

When a mother controls the timing of a birth through modern contraception, the child is likely to be healthier. This is because the mother can avoid having a child in unhealthy circumstances for the child (e.g. when the family has few resources, the birth is short-spaced, or the mother is very young).

We quantify these benefits, primarily, in terms of reduction in child mortality and morbidity from averting mistimed births. Essentially, we are comparing the health of the child if born in a mistimed birth to the health of the child if born in a well-timed birth.

Reduction in child mortality due to fewer mistimed births
Number of mistimed births averted per year of modern contraception	0.099
Infant mortality rate	4.7%
Reduction in infant mortality due to improved birth timing	11%
Infant deaths averted due to fewer mistimed births	0.00053
Units of value per infant death	84
Under-5 mortality rate	6.9%
Reduction in under-5 mortality due to improved birth timing	11%
Under-5 deaths averted due to fewer mistimed births (excluding infant deaths)	0.00024
Units of value per under-5 death	116
Additional benefits from long-term income increases ("development effects")	25%
Units of value from fewer neonatal deaths and stillbirths due to improved timing per year of modern contraception	0.09

Based on a shallow review of the evidence, we roughly estimate that averting mistimed births lowers infant mortality by 11%.92

• We average together estimates from observational studies (Shukla et al. (2020), Tsui et al. (2009), and Finlay (2013)) on the effects of modern contraception usage on the infant mortality rate.
• We adjust our averaged estimate downward by 30% to account for confounding factors. This is because the relationship between contraceptive use and infant mortality estimated in observational data is likely confounded because of correlations between contraceptive use and, for example, other health-related practices, such as sanitation usage, adoption of preventive health technologies, and investment in child's health. This adjustment would be larger (~40%) but using modern contraceptive usage to proxy for the effect of mistimed births may lead to underestimation (see here for more discussion).

Using a similar methodology, we roughly estimate that averting mistimed births lowers under-5 mortality by 11%.

• We use an observational study estimate of the effect of modern contraception on under-5 mortality (Shukla et al., 2020).
• We adjust this estimate downward by 65% to account for confounding factors. This adjustment is larger than the adjustment for infant mortality because the outcome is less proximate to modern contraceptive usage, so there is more scope for unobservables (e.g. parenting practices) to influence the outcome. In addition, we are only relying on one observational study from India, rather than the three observational studies used to calculate the infant mortality rate.
• Importantly, when we calculate the benefits from averting under-5 deaths, we subtract out the infant deaths averted to avoid double-counting.

To account for the fact that healthier children are likely to have better economic outcomes in adulthood, we apply a 25% upward adjustment to our estimate. The magnitude of this adjustment is based on our analysis of other interventions that avert child deaths.93

Our main uncertainties:

• Are our adjustments for confounding factors accurate? The size of these adjustments are subjective judgments and we do not have a reliable way to benchmark them, so we are highly uncertain of their accuracy. These adjustments are consistent with our approach to other potential cause areas.94
• Are our estimates of the effect of improved birth timing on child mortality accurate? They are based on observational studies that likely suffer from confounding and have wide confidence intervals.
• Is the effect of modern contraception on child mortality a good proxy for the difference between mistimed and well-timed births? We are equating the difference in child mortality between users and non-users of modern contraceptives with the effect of averting mistimed births which could be wrong because non-users of modern contraception will still have some well-timed births and users will have some mistimed births. To assess the importance of this choice, we also estimate the effect of averting mistimed births using evidence on the effect of short-spaced births on child mortality. This approach focuses on births that are likely to be mistimed and risky for the child’s health. Using this approach, we estimate a 23% effect on neonatal mortality and a 11% effect on under-5 mortality.95 96 Using these estimates instead of our current ones would not substantially affect our bottom line.
• Are our adjustments for morbidity and long-run income correct? We use adjustments based on global health programs but family planning programs may have different effects. In the future, we could investigate the literature on the association between family planning programs (e.g. Barham, 2012) or maternal age (e.g. Finlay et al., 2011) and child health and economic outcomes to more precisely estimate these adjustments.

Health spending averted due to improved woman and child health

In general, improving the health of women and their children should lower healthcare spending. To account for lower health spending, we adjust upwards the benefits from improving the user’s health (maternal mortality and morbidity averted) and the child’s health (child mortality and morbidity averted).

Reduction in healthcare spending due to improved woman and child health
Units of value from improved woman and child health	0.19
Additional benefit from medical costs averted (as % of health effects)	20%
Units of value from medical costs averted per year of modern contraception	0.04

We did not research what the magnitude of the adjustment as a % of health effects should be for family planning programs in particular. Instead, we applied the same adjustment that we use for our Top Charities.97

Our main uncertainty for this adjustment is whether the reduction in healthcare spending per health benefit from family planning programs is similar to our estimate of the reduction in healthcare spending per health benefit for our Top Charities. We do not have a strong prior on this question. A Guttmacher Institute report attempts to estimate healthcare spending caused by unmet need by adding up the costs associated with each pregnancy outcome. We have not reviewed their methodology in-depth.

Increased earnings for women

Modern contraceptives allow women to control the number, timing, and spacing of their births, which may allow them to pursue educational and labor market opportunities that they would otherwise forgo. These opportunities should, in the long-run, increase the income of women using modern contraceptives relative to the counterfactual of not using modern contraceptives.

Effect of modern contraceptives on income

To estimate the economic benefits of modern contraception for women, we identified three studies that estimated the long-run economic effects of family planning programs. Our understanding is that there are relatively few studies looking at the long-run effects of family planning, so we do not expect that we are missing many important studies on this topic.98

Percentage increase in income per year of modern contraception
Matlab study (Barham et al., 2020)	0.00%
Miller (2010) assuming family planning program increased modern contraception by 17%	0.23%
Miller (2010) assuming same ratio of pregnancy outcomes as in our model	0.06%
Angeles et al. (2005) assuming same ratio of pregnancy outcomes as in our model	0.38%
Best guess	0.17%

More detail on these studies:

• Barham et al. (2020)99 looked at the long-run effects of the Matlab program. They estimate statistically insignificant and slightly negative effects on consumption and income.100 Since we think it is unlikely that family planning reduces income, we round this estimate to 0.
• Miller (2010) is the only other relevant quasi-experimental study we identified. It estimates the effect of family planning program expansion in Columbia. He estimates that expansion of the program led to an increase in female educational attainment (Table 5) and a decrease in lifetime fertility (Table 3).
  • For Miller (2010), we have to extrapolate from effects on education to effects on income. To do this, we use evidence on the returns to education for females in LMICs from observational data (Psacharopoulos and Patrino, 2018) and experiments (Bettinger et al., 2017; Dupas, Duflo, and Kremer, 2024).101
  • We also need to make adjustments to account for the fact that Miller (2010) does not estimate the effect of the program on modern contraceptive usage, which we need to calculate the effect of modern contraceptives on income. We use two approaches to overcome this issue:
    • Make an assumption about the program’s effect on modern contraceptives: We know that modern contraceptive usage increased by 34% in Colombia between 1964 and 1986.102 We assume that half of that increase came from the program.103
    • Assume the ratio of family planning outcomes is the same as estimated in Types of unintended pregnancies averted: This approach allows us to use Miller (2010)’s estimated effects on lifetime fertility to identify the magnitude of the income effect per unwanted birth averted.104
• Angeles et al. (2005) uses a simulation-based approach to estimate the effect of family planning on educational attainment in Indonesia. Since Angeles et al. (2005) also does not provide information on the magnitude of increase in modern contraceptives, we assume the ratio of family planning outcomes is the same as estimated in Types of unintended pregnancies averted. In contrast to Miller (2010), this paper did not provide enough information for us to make an informed assumption about the program’s effect on modern contraceptive use.105

Units of value from increases in income

To quantify the value of increases in income, we calculate the net present value of the income stream and apply our moral weight for consumption increases.

Increases in household income through increased economic activity for women
Percentage annual increase in income per year of modern contraception	0.17%
Share of household income from user of contraception	50%
Household size with contraception	3.8
Duration of benefits	15
Units of value per increase in ln(consumption)	1.44
Discount rate	4%
Units of value from increases in income per year of modern contraception	0.05

We make the following assumptions:

• We roughly assume that the partner at-risk of pregnancy contributes 50% of household income. This is a guess based on the idea that this partner will be one of two income-earning adults in the household. It could be an underestimate because some users are single-parent households or an overestimate because women are less likely to participate in income-earning activities and/or earn less than men in many contexts.
• We estimate the household size with contraception as the median of the estimates for 17 LMICs based on a UN report, subtracting out the projected unwanted births averted over a lifetime due to contraception.106
• We guess that the duration of benefits lasts 15 years. This estimate likely depends on the age profile of workers. Younger users who receive more education due to modern contraception might receive lifelong economic benefits. In contrast, older users may only receive temporary benefits because they have fewer years left in their career and are less likely to be making long-term investments in their human capital.
• Consistent with our other programs, we apply a 4% discount rate.107

Uncertainties on economic benefits for women

We are relatively uncertain about the magnitude of the economic benefits for women. This comes from the following uncertainties:

• What is the effect of modern contraception on earnings among women in LMICs? We have limited evidence on this effect, so are uncertain about our estimate (25th percentile estimate=0.001%; 75th percentile estimate=0.7%)
• Should we separately incorporate economic benefits of avoiding pregnancies that are independent of education? The Miller (2010) and Angeles et al. (2005) estimates only account for benefits from increasing educational attainment. Our approach applies these benefits to all women (not just young women) which assumes that older women improve their earning capacity in similar ways when they avert pregnancies and childcare (e.g. by starting a business rather than going to school). We could try to separately estimate the economic benefits for women who are not in school and apply this to older women.
• How long do the benefits last? We could not find convincing evidence on this, so we made a rough assumption. It is plausible to us that our estimate could be substantially off (25th percentile estimate=10 years; 75th percentile estimate=20 years).
• How to incorporate recent evidence? Preliminary evidence from Karra, Maggio, and Canning (2023) find an increase in female employment due to a family planning program, though this analysis is not finalized. Kleven, Landai, and Leite-Mariante (2024) estimate the child penalty across countries and find smaller effects in poorer countries. We have not reviewed this evidence in depth.

Increased resources for existing children

Women in low- and middle-income countries cite a desire to care for their existing children as a primary motivation for using modern contraceptives.108 By having fewer children, households increase the resources (e.g., money and time) available for other children.

We capture these benefits by modeling the increase in years of education for existing children. This approach captures that women in low-income countries report that family planning enables them to send their existing children to school for more years and to invest in providing children with the resources to thrive in school (e.g. food and clothing).109 We model the primary benefits from increased resources as more years of education for existing children which leads to a stream of higher income over the working career of these existing children:

More resources for existing children due to fewer unwanted births
Number of unwanted births averted per year of modern contraception	0.049
Effect of averting an unwanted birth on years of schooling for existing children	0.53
Effect of years of schooling on income	4.8%
Effect of averting an unwanted birth on income for existing children	2.6%
Duration of benefit	40.00
Average number of years between intervention and the beginning of long-term benefits	13.00
Discount rate	0.04
Benefit on one year's income (discounted back because of delay between intervention and working for income)	0.02
Present value of income per affected child in terms of ln(consumption)	0.32
Multiplier for resource sharing within households	2.00
Value assigned to increasing ln(consumption) by one unit for one person for one year	1.44
Total fertility rate in low-income countries	4.55
Annual under-20 mortality rate in low-income countries	0.90
Averted unwanted birth (lifetime)	0.80
Number of children surviving to adulthood (with contraception)	3.42
Additional benefits for existing children	25%
Units of value from more resources for existing children due to fewer unwanted births	0.19

We estimate the effect of averting an unwanted birth on years of schooling for existing children based on two studies of how quasi-random variation in number of children in a family influences existing children’s educational attainment:

• Bougma et al. (2015) use subfecundity110 as an instrument for the number of other children in one’s family with data from Burkina Faso.
• Kugler and Kumar (2017) instrument for number of children in a family using gender of the first child in India (where parents have a strong son preference so tend to have more children if the firstborn is a girl).
• We discount these estimates by 40% due primarily to concerns around publication bias and the validity of the instruments. These studies were not pre-registered and if they had found null results then they may not have been published. We are unsure of whether these instruments are uncorrelated with children’s years of education. This concern is mostly based on it being difficult to find good instruments rather than specific concerns with the instruments used.111 112

As we did for women’s income effects, we extrapolate from years of education to income by estimating the returns to schooling. We still use the evidence on the returns to education for females in LMICs from observational data (Psacharopoulos and Patrino, 2018) and experiments (Bettinger et al., 2017; Dupas, Duflo, and Kremer, 2024), but we estimate returns to schooling for men and women rather than just for women (since the existing children could be male or female).113

We assume that the duration of benefits last 40 years (our best guess of career length) and that there are on average 13 years between the birth being averted and the child working for income.114 These are rough guesses but varying them by +/- 5 years shifts our estimate of the benefits from more resources for existing children by less than 0.05 units of value.

Following our deworming CEA (which also estimates the benefits of long-term income increases), we use a multiplier for resource sharing within households of 2 to account for the fact that these children are likely to eventually have families who are dependent upon their income.

We multiply the benefits from this stream of income by the number of children of women using contraception who we expect to survive until adulthood. To estimate this parameter, we take the total fertility rate in low-income countries,115 subtract our projection of the number of unwanted births averted,116 and subtract the children who would not make it to 20 years old.117

We add a 25% upward adjustment to this estimate to account for additional benefits for existing children. This adjustment captures that these children will have more resources during their childhood and suggestive evidence that family planning enhances child physical and cognitive development (Maggio, Karra, and Canning, 2023). We are highly uncertain about this adjustment, but we thought we would be underestimating the benefits if we did not attempt to account for these additional benefits.

Since we are uncertain about this estimate, we also estimate the value if the increased resources due to the averted unwanted birth went toward increasing the consumption of all existing household members.118 This method yields an estimate of 0.24 units of value.119 It gives us more confidence in the rough magnitude of our main estimate that this other method yields a roughly similar number. We prefer our main estimate to this alternative for two reasons. The first is that women’s own reports suggest that investing in their existing children is a greater motivation for using contraception than their own consumption, which implies that they are likely to spend additional resources by investing in existing children. The second is that this estimate is highly sensitive to assumptions about when children start contributing to the household’s resources and how to model children’s contribution to household resources once they become adults.120

Our estimate of this benefit could be off due to the following uncertainties:

• How accurate is our estimate of the effect of averting unwanted births on years of schooling for existing children? We rely on relatively low-quality evidence from studies using instrumental variables, so we would not be surprised if this estimate was off.
• Is it inconsistent to account for additional resources due to not having a child but not due to child mortality? We do not think it is inconsistent. We discuss this further here.
• How should we incorporate evidence of improvements in child physical and cognitive development from family planning programs? Maggio, Karra, and Canning (2023) and Barham et al. (2012) find evidence of cognitive gains in RCTs of family planning programs. However, Barham et al. (2012) find evidence that it was the child health intervention aspect of Matlab that caused the effects on physical and cognitive development rather than the family planning aspects.121 Maggio, Karra, and Canning (2023)’s mediation analysis is under-powered but it does not find convincing evidence that contraceptive usage mediates the effect on cognitive development.122 We apply a 25% upward adjustment to account for this and consumption gains but we are highly uncertain about the magnitude of this adjustment.
• Should we incorporate increased resources for adult household members? Our guess is that this is not one of the major benefits of averting an unwanted birth but we have not tried to estimate it explicitly. As mentioned above, parental reports suggest investing in their existing children is a greater motivation for using contraception than their own consumption.123

Improved subjective well-being benefits for women

We think contraception benefits women’s subjective well-being in ways not captured by improvements in health, economic outcomes, or outcomes for existing children. This could include reduced anxiety about becoming pregnant or improved subjective well-being from greater perceived control over reproductive decisions.

We roughly estimate that each year of modern contraceptive use adds the equivalent of ~0.1 DALYs due to improvements in the woman’s subjective well-being (~0.2 additional units of value). This is based on an intuitive guess on what magnitude of effect feels plausible. This is roughly equivalent to averting mild-to-moderate anxiety for a year.124 If the effects only came through unintended births, it would be equivalent to assuming that each unwanted or mistimed birth reduces well-being by the equivalent of ~0.2 DALYs (slightly more than averting a year of moderate anxiety) for 5 years.

Improvements in subjective well-being not captured by improvements in health or economic outcomes
DALY improvement per year of modern contraception	0.10
Units of value per DALY	2.3
Units of value from improvements in subjective well-being not captured by improvements in health or economic outcomes per year of modern contraception	0.23

We’re highly uncertain about these benefits because there’s less empirical evidence to inform us, and we haven’t seen others try to quantify these benefits.

Specific uncertainties we have are:

• What is the appropriate quantitative range for subjective well-being improvements? We’ve relied on rough guesses, and it’s possible that women in these settings would value this reduction in anxiety or improved autonomy to a greater or lesser extent than we’ve assumed.
• Should we include similar subjective well-being benefits for other interventions we fund? For example, does uptake of vaccination or use of insecticide-treated bednets lead to less stress or anxiety about children becoming sick with malaria or vaccine-preventable diseases? We guess these benefits are more salient for contraception, but we haven’t dug deeply into this question.
• How do subjective well-being effects change over time, particularly regarding initially unwanted children? It’s possible that initially unwanted children lead to positive subjective well-being from parents over the long-run.
• To what extent do these subjective well-being benefits overlap with other benefits we've modeled? We’re assuming that these benefits accrue separately from increases in household resources and improvements in health. However, these benefits will be overestimated if reduced anxiety is driven by improved health or economic outcomes.
• How should we incorporate studies that measure the effect of contraception on subjective well-being directly? We discuss these studies below. These studies estimate the effect on subjective well-being overall rather than the effect independent of improvements in health and economic outcomes. However, in the future, we may investigate if we can use these studies to have a better-informed estimate here, by, for example, comparing the treatment group’s subjective well-being to a matched subset of the control group who did not experience a pregnancy during the study period.

3.2 Outside-the-model benchmarks

We compare our modeled estimate to several benchmarks:

• Studies of willingness to pay (more) and studies of the effect of contraception on subjective well-being (more). The willingness-to-pay estimates imply a lower value of modern contraception, while the subjective well-being estimates imply a higher value. Since these effects go in different directions and we don’t have a strong reason to favor one over the other, we don’t view these checks as a major update on the bottom line from our model.
• Comparing the benefits of modern contraception to the benefits of other programs we fund (e.g., increasing income, averting a year of anemia, or averting or child death) (more). While it is challenging to think through these trade-offs, these comparisons make us think our estimates are within the right ballpark.
• Comparison to other models of the benefits of contraception (more). MSI's Impact 2 model estimates greater health benefits than our model due to different assumptions about unintended pregnancies averted and child mortality. We think this reflects our stronger adjustments to estimates from observational data.

Studies of willingness to pay for modern contraception

These studies ask women in LMICs how much they would be willing to pay for modern contraception.

Advantages to these studies:

• Provide input directly from women in LMICs
• Provide a “revealed preference” for contraception
• Provide an explicit trade-off against another outcome we consider in our moral weights (increases in income and consumption)

We conducted a very shallow review of the literature of willingness to pay for modern contraception. Specifically, we reviewed:

• A June 2020 report from USAID125 reported on willingness-to-pay for three family planning products socially marketed by USAID (Protector Condoms, Injectaplan, and Pilplan Plus) in eight districts in Uganda. Highest willingness to pay for these products were:
  • Protector condoms: 1800 UGX, ~$0.50 USD, 120 units per year of contraception, ~$60 per year of contraception
  • Injectaplan: 5,000 UGX, ~$1.40 USD, 4 doses per year of contraception, ~$6 per year of contraception
  • Pilplan Plus: 4,500 UGX. ~$1.30 USD,126 12 doses per year of contraception, ~$16 per year of contraception
• Athey et al. (2021) measured the effect of counseling on willingness to pay for long-acting contraceptives at clients of a women’s hospital in Yaounde, Cameroon. They report 32% of clients were willing to adopt LARCs at a price of $7.25.127 LARCs provide 3-10 years of modern contraceptive use each,128 so this is ~$1-$3 per year of contraception.
• Prata et al. (2013) elicit willingness to pay for injectable contraceptives in Tigray, Ethiopia. Average willingness to pay was 11 birr ($0.65 USD) per injection, and 3% were willing to pay between 50 and 200 birr129 (or $3-$12). Assume 4 doses per year of contraception, then average WTP is $3 and max is $12-$48.
• Guttmacher Institute estimates an annual cost of providing contraceptive services at ~$13 per user annually in sub-Saharan Africa.130

Based on this review, we estimate a range of willingness to pay of roughly $5-$60 per year of contraception. If we assume baseline annual consumption of ~$300,131 this is equivalent to ~2%-20% of annual consumption. This would imply ~0.02-0.3 units of value.132 This is lower than our estimate of ~0.7 units of value (which implies a ~60% increase in consumption for one person for one year).

These push us toward a lower value per year of contraception. However, we think there are reasons willingness-to-pay studies would underestimate benefits:

• Individuals may not fully internalize some of these benefits, which may lead to undercounting of benefits.
• “Liquidity constraints” may lead to undervaluing (e.g., maybe an individual is willing to pay up to $20 but knows she only has $5 to spend currently so reported willingness to pay undervalues true value).
• WTP for contraception may be influenced by respondent’s level of knowledge about the available options and their understanding of the potential benefits. If individuals underestimate the benefits of contraception or overestimate the side effects, this would cause willingness-to-pay estimates to be underestimates of the “true” value of modern contraception.
• Individuals who use contraception as a result of programs we support might be those who value it the most and are therefore at the higher end of willingness-to-pay.

We have several uncertainties about this evidence:

• Are there other studies we are missing? We only conducted a shallow review so are likely missing some relevant studies.
• How much stock to put in these estimates? We guess these undercount benefits, but we’re not sure by how much.
• We have not thoroughly reviewed these studies. We’re taking the estimates at face value, and it’s possible we’re misinterpreting them.

Studies on the effect of contraception programs on subjective well-being

Another approach is to consider the effect of access to contraception on measures of mental health and well-being directly. These studies are paired with randomized controlled trials that aim to increase contraception (e.g., via radio campaigns or vouchers) and measure, as one of the trial outcomes, effects on subjective well-being or other mental health measures.

Advantages of these studies are that they:

• Measure impact on well-being for women in low- and middle-income countries.
• Permit comparisons to studies of the effect of cash transfers on subjective well-being, which provides an explicit trade-off against another outcome we consider in our moral weights (increases in income and consumption).

However, this approach has limitations, too:

• Subjective well-being measures are not “objective” measures and so may be less reliable for various reasons.
• This may not capture all of the effects we’re intending to capture (e.g., longer term effects on birth spacing or number of pregnancies may not be captured by short term impacts on subjective well-being).

We identified two randomized controlled trials that measure the effect of a program that increases years of modern contraceptive use and measures the effect on subjective well-being or a similar measure.

• The two studies we identified suggest a 1.5 SD and 4.6 SD increase, respectively, in an index of well-being for each woman who takes up contraception.
  • Ashraf et al. (2014) conduct a field experiment in Lusaka, Zambia, on the effect of vouchers for free and immediate access to long-term contraceptives (injectables and implants). They find that the treatment increased uptake of modern contraceptives by ~4 pp and led to a ~0.06 SD increase on a mental health index.133 We roughly estimate this implies an effect of ~1.5 SD per woman taking up modern contraceptives.
  • Glennerster et al. (2023) estimate the effect of radio campaigns on uptake of modern contraceptives in Burkina Faso. They find a ~6 pp increase in modern contraceptive use and a 0.27 SD increase on an index of health and well-being. We roughly estimate this implies an effect of ~4.6 SD per woman taking up modern contraceptives.

We can translate this into our current moral weights using studies on the effect of cash transfers, which increase household consumption, on SDs in mental health and well-being.

• A meta-analysis from the Happier Lives Institute finds that $1,000 from GiveDirectly leads to ~1 SD-year improvements in mental health and subjective well-being scores.
• In our GiveDirectly CEA, we estimate GiveDirectly increases recipients’ income by ~75%.
• Using our moral weights, this suggests each SD improvement in mental health and subjective well-being measure is worth ~0.7 units of value.
• The effects from Ashraf et al. and Glennerster et al. imply an effect of ~1.1-3.4 units of value per person taking up modern contraception.134

Taken at face value, these estimates suggest an effect equivalent to ~1.1-3.4 units of value, which is higher than our current estimates (~0.7 unit of value). These push us toward a higher value for a year of modern contraception.

However, we think that these suggested impacts could be over-estimates because we worry about publication bias in one of the studies and whether modern contraceptive usage (vs. some other aspect of the program) drives the subjective well-being effects in both studies:

• Ashraf et al. (2014). As far as we can tell, Ashraf et al. (2014) was not pre-registered, which raises concerns about publication bias (would this result have been reported if it was insignificant? Does the statistical test adequately account for multiple hypothesis testing?). In addition, this result has not been accepted by an academic journal yet so has not gone through peer review. Results from (we think) the same experiment were published but with a focus on the randomization of interviewing voucher recipients alone or with their husband (Ashraf, Field, and Lee, 2014). In that paper, the authors report negative mental health effects on the voucher recipients most likely to use the voucher (Table 3). If this result is from the same experiment, it somewhat weakens our confidence that increases in modern contraceptive usage drove the positive mental health effects in Ashraf et al. (2014).
• Glennerster et al. (2023): Glennerster et al. (2023) pre-registered estimating effects on subjective well-being, but their results are relatively imprecise. After adjusting for multiple hypothesis testing, their result is only marginally significant (p-value=.063; p. 26). It seems likely that this effect is not entirely driven by increases in modern contraceptive usage. The women in the top 25% of predicted effects on modern contraception actually exhibit ~30% lower effects on subjective well-being relative to the overall effects (Table 7), which suggests that other aspects of the program besides the effect on modern contraceptive usage may have driven the effect on subjective well-being.

Key uncertainties:

• Is it appropriate to compare these indices across studies? We think these studies are using different measures than the GiveDirectly studies, and we’re assuming that a 1 SD increase in measures used in Ashraf et al. and Glennerster et al. is equivalent to a 1 SD increase in measures used in the studies in the Happier Lives Institute meta-analysis.
• Have we appropriately benchmarked to GiveDirectly?
• Are there studies that we’re missing?
• Have we misinterpreted these studies?
• Is it plausible that these effects would be so large? Could we sense-check these estimates against differences in subjective well-being across countries (e.g., if we found that our estimates implied access to contraception explained >50% of differences in well-being across countries, we might be more skeptical).

Comparison to other outcomes of programs we fund

As an additional sense check, we can compare the value of a year of modern contraception to other outcomes of programs we fund.

Our estimate of ~0.7 units of value per year of modern contraception suggests that:

• The value of a year of modern contraceptive use is comparable to:
• a ~60% increase in income for one year (~0.7 units of value)135
• averting a year of severe anemia (~0.7 units of value)
• averting a year of clubfoot (~0.6 units of value)
• averting an 18 month-long mild episode of major depressive disorder136 (~0.7 units of value)
• averting two years of a moderate anxiety disorder137 (~0.8 units of value)
• Providing modern contraception for ~20 years to ~8 women is comparable in value to averting the death of a child under five.

Our estimate of ~2.3 units of value per unintended pregnancy averted suggests that

• The value of averting an unintended pregnancy is comparable to:
• Doubling income for two and a half years for one person (~2.4 units of value)138
• Averting a year of severe anemia for three people (~2.1 units of value)
• Averting a year of clubfoot for four people (~2.4 units of value)
• Averting a year of a mild episode of major depressive disorder for five people (~2.2 units of value)
• Averting a year of a moderate anxiety disorder for six people (~2.5 units of value).
• Averting ~50 unintended pregnancies is comparable in value to averting the death of a child under five.

We’re uncertain how to incorporate these intuitive comparisons into our best guess of the value of modern contraception. At a high-level, these comparisons seem broadly reasonable, so increase our confidence in our estimate. In the future, we plan to revisit our moral weights, which could involve asking potential program participants to compare the benefits of modern contraception (e.g. averting an unintended pregnancy) to the other benefits listed above. This would allow us to test whether their preferences are in-line with our estimate of the value of a year of modern contraceptive use.

Comparisons to other models

We compared our model to existing models of the health effects of modern contraception to surface any potential errors and understand the source of our differences.

The most commonly cited model in conversations we had was the Impact 2 model from MSI Reproductive Choices. We asked ReThink Priorities to summarize and replicate the health benefits from a simple scenario in the Impact 2 model, Version 6.139 Then, we compared their replication’s health benefits to an earlier draft of our quantitative mode. Based on our initial comparison, we added under-5 deaths to our model and made some minor adjustments to our parameters.

The remaining major differences between GiveWell’s model and Impact 2’s model are the estimates of unintended pregnancies averted per year of modern contraception and child deaths averted per mistimed birth averted:

• Unintended pregnancies averted per year of modern contraception: Impact 2 uses a range of values for this estimate depending on the contraceptive method being modeled (values range from 0.05-0.50). MSI staff shared that the most commonly used values in Impact 2 are between 0.26-0.50, meaning that the Impact 2 model will typically use a higher value than our estimate (0.3). The Impact 2 model is based on Guttmacher Institute’s Adding It Up project estimates, which also inform our estimates (see here). This estimate is peer-reviewed and aligns with the Foreign, Commonwealth & Development Office (FCDO)-supported Family Planning model harmonization process (see Askew, et al. 2017). We think the differences in our estimates are because:
  • We adjust the Guttmacher Institute’s estimates for the fact that women caused to use modern contraception by our programs are likely to be different from the average user of modern contraception or non-user with unmet need in observational data (see above). These adjustments drop our estimate based on Guttmacher’s model from 0.40 to 0.34. Since Impact 2 is modeling the average user of modern contraception (rather than the impact of GiveWell’s grants), it might not make sense for them to make these types of adjustments.
  • We incorporate results from five family planning programs in LMICs (see here). The weighted average of these results is 0.33 and we adjust this estimate to 0.27 primarily based on internal and external validity concerns with evaluations of the Matlab program (see here).
• Child mortality averted per mistimed birth averted: Impact 2 estimates 0.06 child deaths averted per mistimed birth averted.140 We estimate 0.008 child deaths averted per mistimed birth averted. Impact 2 notes that “estimates of child deaths averted may be unreliable because there is currently very limited data about the linkages between CPR [contraceptive prevalence rate], birth spacing, and child mortality. This part of Impact 2 will be updated as improved research becomes available.” (p. 38) We think the differences in our estimates are primarily because:
  • We adjust estimates for potential confounding of the relationship between child mortality and birth spacing (see above for details on how we account for this). Impact 2 does not make these adjustments.141
  • Impact 2 uses data from 2010 when child mortality rates were higher.142 We use estimates from 2021 and 2022.143

We are planning to continue discussing these estimates with MSI as we refine our model and MSI updates the Impact 2 model.

3.3 How will we account for heterogeneity across programs and contexts?

We expect the value of family planning programs to vary across settings due to differences in program structure and target population.

When we investigate specific programs, we plan to focus on two program characteristics that we expect to be the main drivers of heterogeneity:

• Location: Locations are likely to vary in terms of the burden of unintended pregnancies due to differences in parameters such as the number of unintended pregnancies and the total fertility rate. We think it will be relatively quick to update these parameters so expect to make these adjustments for most grant opportunities. Across seven LMICs, our adjustments result in four estimates within 0.2 units of value of 0.7 (our best guess), a high estimate of 1.03, and a low estimate of 0.34. (more)
• Program participant age: We expect the pregnancy outcomes averted, the economic benefits, and the health benefits of modern contraception to differ by user age. We are uncertain on how to make these adjustments, so we would need to spend more investigation time to do so for a specific grant opportunity. (more)

We also considered adjusting for:

• Rates of traditional contraception usage in a population: 12-month failure rates for traditional contraceptives are ~14% according to Polis et al. (2016). Thus, moving women from traditional to modern contraceptives should have a much smaller effect than moving women from not using any contraception to modern contraception. Our country-level estimates from Bearak et al. (2023) of averted unintended pregnancies already account for traditional contraception usage, so we usually do not need to make an adjustment for this. However, if we expect traditional contraceptive usage rates among our target population will not match the country-level traditional contraception usage rates, then an adjustment could be warranted.
• Rates of postpartum insusceptibility in a population: Similarly, a woman who is postpartum insusceptible adopting modern contraception will have a relatively small effect on unintended pregnancy rates. Based on DHS surveys in LMICs, we think ~10-20% of women 15-49 are likely to be postpartum insusceptible at a given time.144 In cases where we think postpartum insusceptible is likely to be substantially different from 10-20%, it may make sense to adjust unintended pregnancy rate accordingly (i.e. adjust upwards if postpartum insusceptibility is low and downwards if postpartum insusceptibility is high).

Heterogeneity across locations

We expect to investigate grant opportunities across a variety of locations. These locations are likely to differ in terms of the burden of unintended pregnancies on the population. We plan to adjust our model to account for these differences.

To adjust parameters to fit a specific grant location, we plan to incorporate publicly-available country-level averages into our estimates:

• Unintended pregnancies averted per year of modern contraception: We plan to use a weighted average of our best guess for low-income countries and estimates that adjust for the country-level estimates of the Guttmacher Institute.145
• % of unintended pregnancies that result in abortion: We plan to average the low-income country estimate and the country-level estimate from the Guttmacher Institute.146
• Maternal mortality ratio, neonatal mortality rate, under-5 mortality rate: We plan to average the country-level estimates from the IHME and the UN.147
• Under-20 mortality rate: We plan to use the country-level IHME estimates.148
• Total fertility rate and household size: We plan to use estimates from the UN.149

To provide examples of adjusting parameters to new locations and to understand the magnitude of variation across locations, we adjust the relevant parameters for seven LMICs.150 151

	Low-income countries	Nigeria	Malawi	Indonesia	Burkina Faso	Togo	DRC	Chad
Units of value per year of modern contraception	0.70	0.86	0.62	0.35	0.66	0.71	1.05	0.94
Breakdown of benefits (units of value coming from each benefit):
Health effects to women and newborns	0.23	0.36	0.19	0.08	0.23	0.22	0.33	0.39
Increased resources for existing children	0.19	0.20	0.16	0.05	0.17	0.19	0.36	0.26
Economic benefits for women	0.05	0.06	0.05	0.04	0.06	0.05	0.07	0.07
Subjective well-being benefits for women	0.23	0.24	0.22	0.18	0.21	0.25	0.29	0.23

Most of the countries are within ~0.2 units of value of the low-income country estimate. This suggests that ~0.7 units of value provides a relatively accurate guide to the size of benefits we expect from grants that increase usage of modern contraceptives.

The units of value per year of modern contraception for the Democratic Republic of the Congo (DRC) and Indonesia diverge substantially from the low-income country estimate, which highlights the importance of adjusting the estimates based on program location. The largest driver of the high estimate for the DRC is the higher rate of unintended pregnancies averted per year of modern contraception (0.37 in the DRC vs. 0.30 in low-income countries; accounts for 47% of difference between DRC and low-income country estimates). The second largest driver of the DRC’s higher estimate is their higher fertility rate (accounts for 21% of the difference between DRC and low-income country estimates).152

Our key uncertainties for updating the model across locations include:

• How should we estimate unintended pregnancies averted across countries? We currently put weight on our best guess estimate for low-income countries as well as the Guttmacher Institute’s estimates of unintended pregnancies for each country. We are uncertain how much weight we should put on each of these sources.
• How should a woman’s economic outcomes or subjective well-being vary by unintended pregnancies averted? Since we have not investigated this in-depth, we make the rough assumption that the benefits from a woman’s economic and subjective well-being vary in proportion to unintended pregnancies averted. With more investigation, we might learn that this assumption is importantly wrong in some cases.
• How should we adjust for targeting specific locations within a country? The example adjustments use country-level estimates because these are the only publicly available estimates for unintended pregnancies we could find. If we have more detailed data on these parameters based on surveys in a region or experiments in the target population, then we would use those estimates instead.

Heterogeneity across age

Maternal age influences the health and economic effects of a pregnancy. Thus, programs that target different age groups (e.g. adolescents vs. older mothers) will have different benefits. However, we are very uncertain about how to make these adjustments, so it would take substantial investigator time to be confident that these adjustments are making the model more accurate.

When the target population of a grant is particularly young or old, it may make sense to adjust the following parameters:

• Unintended pregnancies averted per year of modern contraception: If sexually active and not using contraception, younger women are more likely to become pregnant due to greater fecundity. Thus, we would expect higher unintended pregnancies averted per year of modern contraception among this population. We could adjust this parameter using the Guttmacher Institute’s estimates at the age group and country-level.153
• Percentage of unintended pregnancies that result in abortion: Darroch et al. (2016) suggests that 53% of unintended pregnancies for 15-19 yos end in abortion compared to the 33% we estimated for all women of reproductive age.154 We could adjust this parameter using the Guttmacher Institute’s estimates at the age group and country-level as well.
• Percentage of unintended births that are unwanted: We expect that unintended births among younger women are more likely to be mistimed (because they are more likely to want another child eventually but not right now), while births among older women are more likely to be unwanted (because they are more likely to have reached their desired fertility). We potentially could adjust this parameter based on country and age group-level data from the DHS.155

We think that the economic benefits for younger women will be larger than for older women. To account for this, we will consider adjusting the following parameters:

• Percentage annual increase in income per year of modern contraceptive use: Younger women who avert unintended pregnancies are more likely to increase their years of education relative to older women. This educational investment should lead to larger income effects for younger women. We could adjust this estimate using the estimates in Miller (2010) by age group.156
• Duration of benefits: The returns to investments in education are also likely to last longer, which suggests that the income benefits for younger women will last longer. In addition, younger women have more of their career left to reap the benefits of increased education.

We think that maternal and child mortality effects would vary by program participant age, so would consider adjusting these parameters:

• Maternal mortality: Nove et al. (2014) find “a J-shaped curve for the age distribution of maternal mortality, with a slightly increased risk of mortality in adolescents compared with women aged 20–24 years… and the highest risk in women older than 30 years.” This evidence suggests we should adjust maternal mortality ratios upward for program participants who are 15-19 year olds or over 30 year olds. We could adjust this parameter using Nove et al. (2014)’s maternal age by country-level estimates.157
• Child mortality: Finlay et al. (2011) finds that “the first-born children of adolescent mothers are the most vulnerable to infant mortality and poor child health outcomes. Additionally, first time mothers up to the age of 27 have a higher risk of having a child who has stunting, diarrhoea and moderate or severe anaemia.” These findings suggest that the benefits to delaying birth for young mothers could be especially large. We could adjust our estimates based on the evidence in Finlay et al. (2011).158

There are multiple ways that adjusting to account for age-related differences could go wrong:

• How much do we trust the evidence on differences in benefits across program participant age? We have not done an in-depth review of this evidence. If this evidence is flawed, adjusting based off of it could cause us to prioritize the wrong age groups.
• Should we account for altering the life path of younger women? For example, younger women who become pregnant unintentionally may end up marrying a worse partner which could reduce their subjective well-being. While speculative, this effect could be large and could substantially enhance the case for focusing on younger women.
• Should we adjust for older women having more existing children? We do not do this because we expect younger women to have a similar number of children to older women in the long-run. These future children will still benefit from their mother’s use of modern contraceptives before they were born. We did not spend much time thinking through this issue, so our judgment on it could be incorrect.
• Are we missing other important parameters that vary by age? This is a fairly shallow take on what important parameters vary across age. If a grant opportunity arose that focused on a specific age group, we expect to re-investigate whether other important parameters vary by age.

4. What are our takeaways for our grantmaking and next steps?

GiveWell’s New Areas team is exploring potential grant opportunities related to family planning.

They plan to start from the model described above to decide whether these programs are likely to meet our cost-effectiveness bar, but will also consider:

• Potential for coercion: As mentioned above, we would not fund programs that coerce women into using modern contraception, including subtly coercing women into choosing specific methods rather than ensuring that they can make an informed choice of the method that works best for them. Grantmakers will evaluate the potential for coercion and rule out any grants where this risk is substantial.
• Value of information: GiveWell often explicitly incorporates the value of information into our grant-making decisions.159 We expect that grants to family planning programs will have substantial learning value for us. For example, grantmaking should allow us to closely observe and learn from how grantees implement family planning programs, which could help us refine parameters in our cost-effectiveness analyses.
• Heterogeneity by location and age:As the prior section described, we plan to account for potential differences in the value of modern contraception across program location and program participant age when this is likely to be decision-relevant.160

While our estimate of ~0.7 units of value per year of modern contraception implies that a program would need to cause a year of voluntary modern contraception for under $20 to be above our bar, we will be open to considering programs above this cost if value of information or heterogeneity adjustments substantially increase the benefits of the program.

With this approach, we expect to identify some family planning programs above our 10x bar. We expect to learn more about cost-effectiveness as we explore specific programs so we do not have well-developed estimates for programs we may fund. However, as one benchmark, Glennerster et al. (2023) cites a range of $30 to $60 per year of modern contraceptive use.161 With a moral weight of ~0.7, this implies a cost-effectiveness of 3.5x-7x.162 As a result, we guess that, while the typical program is below our cost-effectiveness bar, there are some exceptional programs that are above our bar (though we view this as a very rough guess, given our uncertainty both on the value of modern contraception and on cost per year of modern contraception from family planning programs).

If we find a large number of promising programs in this space, we may conduct further research on how to value a year of modern contraception. This could include conducting additional preferences research to better understand how individuals in areas we would consider funding family planning programs might trade off a year of modern contraceptive use or averting an unintended pregnancy against other outcomes.

Forecasts

Confidence	Prediction	By time
65%	We will conduct additional in-depth research on this topic. This would not include minor updates based on new studies or evidence that we come across.	December 2027
50%	We will fund a “beneficiary preferences” study that asks individuals in areas where GiveWell may fund programs how they trade off modern contraception and/or averting unintended pregnancies against other outcomes	December 2027
15%	Conditional on conducting additional research, we will update our moral weight on a year of modern contraceptive use to >1.5 units of value, based on follow-up work or new research	December 2029
33%	Conditional on conducting additional research, we will update our moral weight on a year of modern contraceptive use to <0.5 units of value, based on follow-up work or new research	December 2029

Sources

Document	Source
Alano and Hanson (2018)	Source
Andre, Delesalle, and Dumas (2020)	Source
Angeles et al., 2005	Source
Anik, Islam, and Ramik, (2022)	Source
Ashraf et al. (2014)	Source (archive)
Ashraf, Field, and Lee, (2014)	Source
Askew et al. (2017)	Source
Athey et al. (2021)	Source (archive)
Barham (2012)	Source
Barham et al. (2021)	Source
Bauserman et al (2020)	Source
Bearak et al. (2020) (a)	Source
Bearak et al. (2020) (b)	Source
Bearak et al. (2023)	Source
Bettinger et al., 2017	Source (archive)
Bongaarts, (2017)	Source
Bougma et al. (2015)	Source
Boydell and Smith, 2023	Source
Brown and Eisgenberg, The Best Intentions: Unintended Pregnancy and the Well-Being of Children and Families.	Source (archive)
Cleland et al, (1994)	Source (archive)
Cleland et al. (2012)	Source
Darroch et al. (2016)	Source (archive)
Demographic and Health Surveys Program, Homepage	Source (archive)
DHS Statcompiler	Source (archive)
Duflo et al. (2021)	Source
Dugas and Slane (2022)	Source (archive)
Dupas et al., 2024	Source
Dupas, Duflo, and Kremer, 2024	Source (archive)
Effective Altruism Forum, Maternal Health Initiative is Shutting Down, Williamson and Eustis-Gutherie (2024)	Source (archive)
Family Empowerment Media, Blog	Source (archive)
Family Empowerment Media, FEM’s Research: What we learned about our audience and how best to reach them	Source (archive)
Family Empowerment Media, Homepage	Source (archive)
Family Empowerment Media, Survey data	Source (archive)
Family Empowerment Media, Survey data (Drivers of modern contraception)	Source
Farmer et al. (2015)	Source
Finlay (2013)	Source
Finlay et al., (2011)	Source
FP 2030, Commitment makers	Source (archive)
GiveWell, 2020 update on GiveWell's moral weights	Source
GiveWell, 2023 GiveWell cost-effectiveness analysis — version 3	Source
GiveWell, AMF and Population Ethics	Source
GiveWell, Approaches to Moral Weights: How GiveWell Compares to Other Actors	Source
GiveWell, Assessment of Happier Lives Institute’s Cost-Effectiveness Analysis of StrongMinds	Source
GiveWell, Child mortality rates IHME	Source
GiveWell, Context for "Average number of years between deworming and the beginning of long-term benefits" Parameter	Source
GiveWell, Facility-Based Maternal and Neonatal Health Interventions	Source
GiveWell, Facility-based MNH BOTEC	Source
GiveWell, Family Empowerment Media	Source
GiveWell, Family planning radio campaigns BOTEC with FEM Kano data	Source
GiveWell, GiveDirectly CEA	Source
GiveWell, GiveWell and I2 health effects of modern contraceptives comparison	Source
GiveWell, GiveWell's CEA of vitamin A supplementation, 2024	Source (archive)
GiveWell, GiveWell's Cost-Effectiveness Analyses	Source
GiveWell, GiveWell's cost-effectiveness analysis for deworming	Source
GiveWell, GW's moral weights and discount rate	Source
GiveWell, How Do You Value Information?	Source
GiveWell, IDinsight — Beneficiary Preferences Survey (2019)	Source
GiveWell, IMHE U20 Death Rate	Source
GiveWell, Iron Fortification	Source
GiveWell, London School of Hygiene & Tropical Medicine — RCT of Eyeglass Provision (May 2024)	Source
GiveWell, Mass Distribution of Insecticide-Treated Nets (ITNs)	Source
GiveWell, Maternal Health Initiatives (MHI) CEA	Source
GiveWell, MiracleFeet — Clubfoot Treatment (January 2023)	Source
GiveWell, MMR across countries	Source
GiveWell, Moral Weights	Source
GiveWell, Neonatal mortality rate IGME	Source
GiveWell, New Incentives (Conditional Cash Transfers to Increase Infant Vaccination)	Source
GiveWell, One Acre Fund — Tree Program RCT and Data Collection (January 2023)	Source
GiveWell, Our Top Charities	Source
GiveWell, Re-evaluating the Impact of Unconditional Cash Transfers	Source
GiveWell, Research on Moral Weights - 2019	Source
GiveWell, Sayana Press	Source
GiveWell, Seasonal Malaria Chemoprevention	Source
GiveWell, Total fertility rate WBG	Source
GiveWell, Under 5 mortality rates IGME	Source
GiveWell, Valuing contraception BOTEC	Source
GiveWell, Vitamin A Supplementation	Source
Glennerster et al. (2023)	Source (archive)
Greaves (2017)	Source
Gribble and Voss (2009)	Source (archive)
Guttmacher Institute Data Center	Source (archive)
Guttmacher Institute, Adding It Up: Costs and Benefits of Meeting the Contraceptive Needs of Adolescents in Sub-Saharan Africa	Source (archive)
Guttmacher Institute, ADDING IT UP: Investing in Sexual and Reproductive Health 2019 Methodology Report	Source (archive)
Guttmacher Institute, Adding It Up: Investing in Sexual and Reproductive Health 2019—Methodology Report (2020)	Source (archive)
Guttmacher Institute, Adding It Up: Investing in Sexual and Reproductive Health 2019, Data	Source (archive)
Guttmacher Institute, Pregnancy and Abortion Worldwide (AWW)	Source (archive)
Higashi et al. (2014)	Source
Hoyt et al. (2021)	Source
IDinsight, Homepage	Source (archive)
IDinsight, Measuring people’s preferences and values to inform funding decisions	Source (archive)
IHME’s Global Burden of Disease	Source (archive)
Institute for Health Metrics and Evaluation, Global Burden of Disease Study 2019 Disability Weights	Source (archive)
Institute for Health Metrics and Evaluation, Global Burden of Disease Study 2021	Source (archive)
Institute for Health Metrics and Evaluation, Global Burden of Disease Study 2021 Disability Weights	Source (archive)
Institute For Health Metrics and Evaluation, Homepage	Source (archive)
Joe Carlsmith, Against neutrality about creating happy lives	Source (archive)
Jones (2022)	Source (archive)
Joshi and Shultz, (2012)	Source
Karra et al., (2022)	Source
Karra, Maggio, and Canning (2023)	Source (archive)
Kleven, Landai, and Leite-Mariante (2024)	Source
Kosgei et al. (2011)	Source
Kremer (1993)	Source
Kugler and Kumar (2017)	Source
LaLonde (1986)	Source
Londoño-Vélez and Saravia (2025)	Source
Maggio, Karra, and Canning (2023)	Source
Miller, (2010)	Source
Mosha et al. (2013)	Source
MSI Reproductive Choices, Homepage	Source (archive)
MSI Reproductive Choices, Impact 2	Source (archive)
MSI Reproductive Choices, Impact 2 Methodology Report (2023)	Source (archive)
Nove et al., (2014)	Source
Ontiri et al., (2023)	Source
Pelras and Renk, (2023)	Source
Performance Monitoring for Action dataset	Source (archive)
Polis et al. (2016)	Source (archive)
Prata et al. (2013)	Source
Psacharopoulos and Patrino, 2018	Source
Richard Y Chappell, The Profoundest Error in Population Ethics	Source (archive)
Riley T et al. (2020)	Source
Rutgers Institute for Health, Health Care Policy and Aging Research, IFH Team Provides Ethical Advice to GiveWell (2024)	Source (archive)
Sedgh, Singh, and Hussain (2014)	Source
Senderowicz (2019)	Source
Shukla et al. (2020)	Source
Stanford Encyclopedia of Philosophy, Robert Nozick’s Political Philosophy (2022)	Source (archive)
Staveteig (2017)	Source
Su (2012)	Source
The Challenge Initiative, The Measurement, Learning & Evaluation Project	Source (archive)
Tsui et al. (2009)	Source (archive)
UN IGME, Homepage	Source (archive)
UNICEF Multiple Indicator Cluster Surveys	Source (archive)
United Nations, Family Planning Indicators (2024)	Source (archive)
United Nations, Household Size and Composition Around the World (2017)	Source (archive)
United Nations' World Family Planning 2022 report (2022)	Source (archive)
USAID, Couple-Years of Protection (CYP)	Source (archive)
USAID, SURVEY REPORT ON THE WILLINGNESS TO PAY FOR THE USAID SOCIALLY MARKETED PILPLAN PLUS, INJECTAPLAN AND PROTECTOR CONDOM (2020)	Source
Valente et al. (2024)	Source
WHO, Birth spacing — report from a WHO technical consultation (2006)	Source (archive)
WHO, Family planning/contraception methods (2023)	Source (archive)
Wikipedia, Demographic Dividend	Source (archive)
World Bank Group, Fertility rate, total (births per woman)	Source (archive)
World Bank Group, Primary completion rate, female (% of relevant age group)	Source (archive)
World Health Organization, The global burden of unsafe abortion in the year 2000	Source (archive)

• 1

  In the rest of this write-up, when we refer to the value per year of modern contraception, we are referring to the value of a program causing a year of modern contraception usage that otherwise wouldn't occur for a woman in an LMIC who wants it. We will typically use the shorter term – value per year of contraception – for simplicity and to enhance the readability of the write-up. Our definition differs from the commonly-used couple-years of protection, because we are interested in the increase in contraception usage relative to the counterfactual (the world in which we did not fund the program) rather than the amount of contraception a program distributes.

• 2

  In our cost-effectiveness analyses, we use “units of value” as a way to convert different types of benefits (e.g., increases in income or improvements in health) into a single unit. This lets us combine different benefit streams from a given program and compare across programs. For more on units of value and our moral weights, see this page.

• 3

  See our calculation here. This number comes from dividing the units of value per year of modern contraception by the number of unintended pregnancies averted per year of modern contraception. This method will slightly overstate our estimate of the value of averting an unintended pregnancy since some of the value from a year of modern contraception would occur even if there was no effect on unintended pregnancies (e.g. women likely feel less anxiety due to contraception use regardless of whether they would have an unintended pregnancy counterfactually).

• 4

  We do not include child health effects for births that do not occur due to contraceptive use (i.e. we do not count averting a neonatal death of a birth that would not have occurred with contraception). Refer to What about lives that don’t occur due to contraception? for more discussion of births that do not occur due to contraceptive use.

• 5

  Glennerster et al. (2023) describes a study of Development Media International’s programming, “Eight of 16 geographically and linguistically distinct community FM radio stations were selected to receive a media campaign designed by Development Media International (DMI).” p. 3
  From their abstract, they report the result that, “Women receiving radios in status quo areas reduce contraception use by 5.2 percentage points. This negative effect is concentrated among those who wanted fewer children, consistent with mass media increasing social pressure to conform to the modal behavior in the media market. In contrast, receiving a radio in campaign areas increases contraception use by 5.8 percentage points. Comparing women in campaign vs noncampaign areas we find contraception use is 5.9 percentage points higher, births 10% lower, misperceptions about contraception lower, and reported welfare 0.27 standard deviations higher in campaign areas.” p. 1

• 6

  The full definition of women with unmet need for family planning is “women who are fecund and sexually active, who wish to stop or delay childbearing, but who are not using any form of contraception. A woman is also considered to have an unmet need if she was pregnant at the time of data collection, but reported that the pregnancy was unwanted or mistimed, or if a woman was postpartum amenorrhoeic, not using family planning and her most recent birth was unwanted or mistimed.” (United Nations' World Family Planning 2022 report, p. 4)

• 7

  United Nations, Department of Economic and Social Affairs, Population Division (2024). Model-based Estimates and Projections of Family Planning Indicators 2024, custom data acquired via website

• 8

  DHS surveys of women from 16 LMICs

• 9

  “For analytical purposes, contraceptive methods are often classified as either modern or traditional. In this report, modern methods of contraception include female and male sterilization, intra-uterine devices (IUD), implants, injectables, oral contraceptive pills, male and female condoms, vaginal barrier methods (including the diaphragm, cervical cap and spermicidal foam, jelly, cream and sponge), the lactational amenorrhea method (LAM), emergency contraception and other modern methods. Traditional methods of contraception include rhythm (e.g., fertility awareness-based methods, periodic abstinence), withdrawal and other traditional methods.”
  (United Nations' World Family Planning 2022 report, p. 4)

• 10

  We have previously done a shallow investigation of Sayana Press, an injectable contraceptive intended to be easy to administer, and recently made a grant to Family Empowerment Media to help support a randomized controlled trial on the effect of a radio program aimed at addressing barriers to contraception. For that grant, we used an initial estimate of the benefits of a year of modern contraceptive use to inform this grant.
  This report provides a slightly more in-depth but still preliminary analysis of how to value contraception and aims to give more transparency on how we came up with this estimate and areas of uncertainty.

• 11

  Family Empowerment Media, a nonprofit that promotes health education via radio in Nigeria, has run surveys amongst listeners to understand their views of contraception.The Family Empowerment Media surveys found that the most common drivers of interest in contraception among women in Nigeria were “Better ability to pay for necessities (e.g. food and clothing) for children” and “Better ability to educate/train their children”. A summary of the results of their surveys on the drivers of modern contraception in Nigeria’s South-South region is available here.
  Farmer et al. (2015) a qualitative study in Rwanda found that “over two-thirds of the women reported their discussions about family planning were related to having the financial means to properly manage family needs”.
  Hoyt et al. (2021), a qualitative study in Benin, Ethiopia, Kenya, Malawi, and Uganda found that improving “a child’s prospect for growth and future opportunity” was seen as a key benefit of family planning.
  Mosha et al. (2013), a qualitative study in Tanzania, found that women interviewed alone cited “better ability to provide for their children” and couples interviewed together cited “economic reasons for utilizing family planning, wanting manageable families they can afford to care for” as the main way that family planning provided value.

• 12

  Alano and Hanson (2018), a phenomenological qualitative study in Southern Ethiopia found that “When women are able to postpone unwanted pregnancies and childbirths, they have more time to plan and engage in non-reproductive issues such as income generation.”
  Miller (2010), a quasi-experimental study in Colombia, found that “women gaining access to family planning as teenagers obtained 0.05 more years of schooling, [and] were 7% more likely to work in the formal sector” (p. 711).

• 13

  Hoyt et al. (2021), a qualitative study in Benin, Ethiopia, Kenya, Malawi, and Uganda stated that, “At all sites, women expressed a desire to space their children, primarily because of the perceived health benefits for themselves and their children, including the ability to feed and care for their children.”
  Farmer et al. (2015), a qualitative study in Rwanda’s southern Kayonza district, found that “Participants often viewed family planning as a means of improving the socioeconomic condition of one’s family and community, explaining that contraception facilitates birth spacing and timing, which can, in turn, improve marital relations, sanitation, and hygiene; give a mother time to work and care for her children; and lead to better health outcomes for pregnant women”

• 14

  Hoyt et al. (2021), a qualitative study in Benin, Ethiopia, Kenya, Malawi, and Uganda found that, “Women’s decisions to act and take up [a modern contraception method] was strongly influenced by their desire to space births and capitalise on the perceived health and wellbeing benefits for themselves and their children.”

  Ashraf et al. (2014), a field experiment in Lusaka, Zambia, found that “Improved access to long-term contraceptive methods with lower failure rates contributes to reduced anxiety and greater feeling of control, particularly among women who are subject to greater anxiety over becoming pregnant. Social and economic constraints have been found to be significantly correlated with depression, and in this setting easing an important constraint on health production benefited women via greater psycho-social wellbeing.”

• 15

  Family Empowerment Media relayed to us that their research in northern Nigeria revealed that contraceptive adoption was primarily motivated by parents' desire to better provide resources and time for their existing children, rather than perceived health benefits. Users also valued having more control over pregnancy timing, even when they didn't have strong preferences about total family size.
  Among community stakeholders across regions, government officials and religious leaders emphasized the maternal health benefits of family planning and noted the strain on health services from high fertility rates. These religious leaders also framed family planning in terms of responsible parenthood, ensuring families can provide adequate care, moral upbringing, and enable children to become contributing members of society. While some stakeholders in northern Nigeria noted concerns about potential links between large families and youth vulnerability, this was not a widespread view among government officials.
  For more on the results of Family Empowerment Media’s research on family planning, see this 2021 post on ‘What we learned about our audience and how best to reach them’ and similar posts on their blog.

• 16

  For example, we believe that roughly 20% of the benefits of New Incentives’ conditional cash transfers for vaccination come from long-term income increases and for mass distribution of insecticide-treated nets this figure is 25% -40%.

• 17

  Because the IHME GBD 2019 disability weights does not include a weight for clubfoot directly, we use the weight for “disfigurement level 2 with pain and moderate motor impairment due to congenital limb deficiency" as an approximation. This disability weight includes both physical and non-physical harms described as, “has a visible physical deformity that causes others to stare and comment. As a result, the person is worried and has trouble sleeping and concentrating” and “has some difficulty in moving around, and difficulty in lifting and holding objects, dressing and sitting upright, but is able to walk without help.” Global Burden of Disease Study 2019 (GBD 2019) Disability Weights

• 18

  See discussion here.

• 19

  See pp. 99-101 of IDinsight Beneficiary Preferences Final Report. Social desirability bias may have caused some respondents to avoid mentioning the upsides of a household member’s death, so we take this evidence with a grain of salt.

• 20

  This seems like an important distinction, though we do not have a fully-fleshed out argument for why this distinction is important. One partial argument is that the former goes against the family’s preferences while the latter fulfills their preferences. Another is that intended children are likely to have less of an effect on family resources than unintended children who may have been unintended because of the family’s precarious economic situation.

• 21

  For example, the higher consumption of other household members due to an averted birth (~4 units of value) would only account for ~3% of the value of averting an under-5 death (116 units of value), meaning that incorporating this cost into our model of programs that avert deaths would not substantially affect our cost-effectiveness estimates of programs like seasonal malaria chemoprevention.
  To test this assumption, we performed a sensitivity analysis for our simple CEA of seasonal malaria chemoprevention here that estimates a 2% reduction in units of value created if we account for reduced resources for existing children due to averted deaths.

• 22

  We think it is likely that modern contraception reduces the number of births overall. This is based on empirical evidence from family planning programs (e.g. Joshi and Shultz, 2012 find that a family planning program in Bangladesh led to a ~17% decline in fertility), the negative correlation between fertility and modern contraceptive use across countries (e.g. Bongaarts, 2017; Table 1), and the fact that unwanted births (meaning births a mother did not want at any time) are a substantial portion of the unintended births that could be averted by increasing modern contraception (33% of unintended births in Africa are unwanted according to Sedgh, Singh, and Hussain (2014)’s Table 4). If contraception only improved the timing of births without affecting the overall number of births (i.e. only reduced mistimed births but not unwanted births), then there would not be an issue with lives that don’t occur due to contraception. However, since we think the benefits from averting an unwanted birth are greater than the benefits from averting a mistimed birth (see our model for more on this), our estimate of the benefits of modern contraception would be substantially lower if it only improved the timing of births.

• 23

  For example, we solicited feedback from a group of ethicists from the Rutgers Institute for Health, Health Care Policy and Aging Research and their colleagues (see description of engagement here). We also solicited feedback from implementers of family planning programs, other global health practitioners, GiveWell staff, and GiveWell donors.
  We reviewed literature on these questions and considered other approaches (such as revealed preference) for inferring views of program participants.

• 24

  This question relates to population ethics/axiology, “the study of the conditions under which one state of affairs is better than another, when the states of affairs in question may differ over the numbers and the identities of the persons who ever live.” (Greaves (2017), p.17) Population ethics is a notoriously difficult field of philosophy. As Greaves (2017) writes, “every extant population axiology is open to serious objection: if it does not entail the Repugnant Conclusion then it entails the Sadistic Conclusion, or is anti-egalitarian, or has obviously unacceptable implications concerning future-people cases, or otherwise leads to some similarly serious objection.” (p. 17)

• 25

  If family planning programs might be decreasing the population by avoiding unwanted births and one adopts a population axiology where an additional life is a positive good, there could still be a conflict between respecting people’s autonomy and one’s preference for the world with the additional life.

• 26

  See this write-up for details on why we consider the views of GiveWell’s staff, donors, and people in LMICs who could be affected by our funding.

• 27

  One version of this rationale is the following:

  • Bodily autonomy (including about reproductive choices) is often considered a right.
  • Rights can be considered "side constraints" to utilitarian-style thinking. See discussion here.
  • A naive total utilitarian might think that it would be good to withhold contraception from people (or decline to provide support for it) because it would increase population size. But utilitarianism with bodily autonomy rights as side constraints would reject this view.
  • Placing a positive value on potential people who may exist in absence of contraception is somewhat like buying into that naive total utilitarian view.

• 28

  In the most recent DHS surveys of Burkina Faso, Chad, the Democratic Republic of the Congo, Kenya, and Malawi, ~20-50% of women and men wanted no more children. (DHS Statcompiler)

• 29

  E.g. 36 LMIC governments “have made commitments to advance family planning in their respective countries. By joining FP2030, they demonstrate their commitment to providing their citizens with the necessary resources and support to make informed decisions about their sexual and reproductive health.” (See FP2030 website) This list includes countries like Ghana with competitive democratic political systems.

• 30

  “WHO is working to promote contraception by producing evidence-based guidelines on safety and service delivery of contraceptive methods and on ensuring human rights in contraceptive programmes. WHO assists countries to adapt and implement these tools to strengthen contraceptive policies and programmes. Additionally, WHO participates in developing new contraceptive technologies to and leads and conducts implementation research for expanding access to and strengthening delivery contraceptive information and services.” (WHO website)

• 31

  The number of unwanted births averted (~0.05) multiplied by our value of lives of early neonates (84) would greatly exceed the benefits of modern contraception (0.05*84=3.9>>0.7). We model this scenario here. Even if we assumed an unwanted birth was ~20% as valuable as averting an early neonatal death, this would imply a year of modern contraception would be a net negative (0.05 * 84 * 15% = 0.8 > 0.7). We model this scenario here.

• 32

  See grant page on MiracleFeet.

• 33

  An experiment that randomized secondary high school scholarships in Ghana found that female recipients delayed their first birth and ended up marrying a more-educated partner (Dupas et al., 2024; Table 1). One could imagine family planning programs→less pregnancies→more schooling→a more educated partner as well.

• 34

  For examples of macroeconomic models, see Jones (2022) and Kremer (1993).

• 35

  See discussion here.

• 36

  For example, one experimental study of contraceptive injections notes that: “However, we do find that these individuals experienced a significant reduction in happiness, health and ease of mind compared to those in the Couple treatment. This suggests a longer-term psychosocial cost to concealable contraceptives that can be mitigated by spousal involvement and is often ignored by programs focused on giving women reproductive control.” (Ashraf, Field, and Lee, 2014)

• 37

  Ben Williamson and Sarah Eustis-Guthrie’s EA forum post discusses other potentially important side effects: “Side effects like inconsistent or excess bleeding may be seen as relatively manageable in Western countries with more accepting cultural norms and practices but can be far more significant in low-income countries (Kulhmann et al., 2017; Polis et al., 2018).
  Cultural practices around inconsistent or excess bleeding can prevent women from participating in a range of regular activities, including prayer, sexual activities, and community engagement (Bradley et al., 2009; Mohammed et al., 2020). Meanwhile, amenorrhea from contraceptive use is frequently interpreted as a sign of pregnancy and therefore potential promiscuity (Mackenzie et al. 2020). Given these consequences, the choice of some women not to use hormonal contraception in order to avoid these consequences is worthy of respect and support.”

• 38

  For example, the Indian government’s aggressive family planning initiatives in the 1970s involuntarily sterilized many people. This is an example of a highly unethical family planning program that GiveWell would not fund. In addition, it reduced trust in and usage of public health services in later years (Pelras and Renk, 2023). This historical experience highlights the necessity of avoiding coercive family planning programs.

• 39

  Senderowicz (2019) interviewed 49 women of reproductive age in an African country in 2017 and found that some women felt pressured by health clinic workers to adopt long-acting reversible contraceptive methods. A systematic review (Boydell and Smith, 2023) also found evidence of coercive practices in relation to encouraging uptake of long-acting reversible contraceptive methods.

• 40

  For more, see this sheet we calculate the units of value per year of modern contraception implied by the 25th and 75th percentile values of several key parameters. Note that we do not simply move all parameters to the 75th percentile to estimate the 75th percentile estimate. That’s because the parameters are unlikely to be perfectly correlated, meaning that it is less likely that they would all be relatively high than that one of them would be relatively high.

• 41

  This definition is from page 3 of Sedgh, Singh, and Hussain (2014).

• 42

  Riley T et al., Adding It Up: Investing in Sexual and Reproductive Health 2019—Methodology Report, New York: Guttmacher Institute, 2020

• 43

  See calculations here. Since, on average, a bit under half of women aged 15-49 in LMICs want to avoid pregnancy (based on Guttmacher Institute estimates), we multiply pregnancy outcomes per year by ~16 to get these estimates. This estimate does not account for the age distribution of women in LMICs since we could not find estimates of the percent of women who want to avoid a pregnancy in narrower age ranges.

• 44

  Refer to the “Pregnancies averted” sheet here. We are assuming that the projected number of unwanted births averted is equivalent to the reduction in lifetime fertility. This calculation assumes that averting a mistimed birth through modern contraception does not reduce lifetime fertility as we expect mistimed births to occur at a later time.

• 45

  Here we are using non-users to mean women who do not use modern contraception (meaning they could be using no contraceptive method or could be using traditional contraceptive methods such as abstinence or withdrawal).

• 46

  For simplicity, we take the average of the country-level averages rather than weighting by population size.

• 47

  See the list of these countries here. We have chosen these countries as they are ones in which GiveWell has previously considered funding family planning initiatives.

• 48

  Riley T et al., Adding It Up: Investing in Sexual and Reproductive Health 2019—Methodology Report, New York: Guttmacher Institute, 2020

• 49

  Per Riley T et al., Adding It Up: Investing in Sexual and Reproductive Health 2019—Methodology Report, New York: Guttmacher Institute, 2020, “The survey data used in the analysis include DHS, UNICEF’s Multiple Indicator Cluster Survey (MICS), Centers for Disease Control and Prevention’s Reproductive Health Survey (RHS), Performance Monitoring for Action (PMA) survey, Generations and Gender Survey (GGS), Pan Arab Project for Family Health (PAPFAM) survey and other national survey data.” p11

• 50

  Our main reservation is their use of 40% as their prior on the unintended pregnancy rate among non-users with unmet need for contraception. We think this prior substantially influences their final estimates of this parameter. Since their writing suggests that 40% is a best guess rather than based on strong empirical evidence, we are somewhat concerned that it could differ substantially from the truth. More details on their methodology: “For the “initial” pregnancy rate for women wanting to avoid pregnancy but using no method, we used a rate of 40%. The commonly used estimate of 85% represents the estimated pregnancy rate during the first 12 months of couples attempting to get pregnant. The 40% pregnancy rate is likely more realistic for a general population of couples who are at risk of unintended pregnancy but are not using a contraceptive method because it reflects probable lower levels of sexual activity and fecundity among actual nonusers, many of whom have not become pregnant despite being sexually active and not using a method for more than 12 months. Also, based on the DHS approach to categorizing women as having unmet need and using no contraceptive method, nonusers wanting to avoid pregnancy include women who identify their current pregnancy as unintended or are experiencing postpartum amenorrhea after an unintended pregnancy.” p. 62 of Riley T et al. (2020)

• 51

  Refer to here for details on our unadjusted estimates of unintended pregnancies averted per year of modern contraceptive use

• 52

  Another way to put this is that we are trying to address the typical issue with non-experimental evaluations of an intervention, namely that other factors are confounding the relationship between non-users and users. For example, LaLonde (1986) compares experimental and non-experimental evaluations of a job training program. The non-experimental evaluations (e.g. income of program users vs. income of program non-users) are biased due to unobservable differences in the characteristics of program users (e.g. they might have used the program because they were more motivated to find a job than the non-users).

• 53

  This adjustment is partially motivated by a randomized controlled trial of a family planning program in Zambia, which found that women who took up modern contraception due to the new program were more likely to have been using traditional contraception at baseline. (Ashraf et al. 2013, p. 3)

• 54

  We did not apply any adjustment for the type of modern contraception used by women who take up modern contraception due to GiveWell-funded programs compared to typical users of modern contraception.

• 55

  Riley T et al., Adding It Up: Investing in Sexual and Reproductive Health 2019—Methodology Report, New York: Guttmacher Institute, 2020, p. 23, Figure 4.2

• 56

  According to our (unpublished) analysis of DHS surveys in Ghana (2022), Uganda (2015-16), and India (2015-16), 10-20% of women who are experiencing postpartum amenorrhea after an intended pregnancy use modern methods of contraception. We selected these surveys for convenience and have not prioritized further work on this estimate.

• 57

  This is based on our rough assumption, informed by our review of the literature, that the number of women who are experiencing postpartum amenorrhea is ~10-20% of the number of women with unmet need for modern contraceptives, and our assumption that they will be somewhat less likely to adopt compared to women with unmet need (due to lower pregnancy risk making adopting modern contraceptive less attractive). It is also possible that some of the women who adopt modern contraception will not be sexually active and may be adopting for other benefits of modern contraception (such as hormonal control). We did not include an adjustment for this possibility since we did not find evidence of this being an important factor in LMICs.

• 58

  We excluded studies that had little effect on modern contraceptive usage since they would have little statistical power to estimate the effect of modern contraception on unintended pregnancies or births.

• 59

  Joshi and Shultz, 2012, p. 1

• 60

  We rely on two papers to estimate the effects of the Matlab experiment, Joshi and Shultz (2012), an observational study on the program, and Cleland et al. (1994), a World Bank report on the program.

• 61

  Joshi and Shultz, 2012, “These Community Health Workers (CHWs) advise women on the use of birth control, provide supplies (including the pill and injectable) as well as follow-up services, and refer women to local clinics or hospital when necessary (Phillips et al. 1982, 1988). After 1981 additional maternal and child health services were added to the program, such as tetanus toxoid immunization for women, measles immunization for children 9 months to 5 years, and then other EPI childhood vaccinations, oral rehydration therapy (ORT) for diarrhea, vitamin A supplements, and antenatal care, etc. (Phillips et al. 1988; Fauveau 1994). By the 1990s, the Government of Bangladesh provided some of these vaccinations but their adoption is not recorded (LeGrand and Phillip 1996, p.58.1” p. 4

• 62

  Joshi and Shultz, 2012, “The sample size is 282 from combining two cross sections of villages, and the GLS estimates are reported in the two columns of Table 1 for the two different post-program census or survey years, 1982 and 1996.” p. 7

• 63

  Joshi and Shultz, 2012, Table 4

• 64

  Source: Cleland et al., 1994, Figure 4.3

• 65

  For example, Barham et al. (2012) find evidence that it was the child health intervention aspect of Matlab that caused effects on physical and cognitive development of children. These improvements in the health of existing children could influence a family’s decision to have another child. “One limitation of this study is that the interventions were not randomly introduced in a cross-over design, making it difficult to determine the separate effects of the interventions. An important contribution of this paper has been to provide some evidence of the separate effect of the child health interventions. First, estimation by age group finds a large and significant effect of the child health program on children born after the child health interventions were introduced, but a small, negative, and statistically insignificant effect of the program for children who were born before the child health interventions began, but whose mothers were eligible for the family planning and maternal interventions. While this analysis is by itself insufficient, since family resources differ for earlier- and later-born children, adding a control for the family planning program in the ITT and TOT models still yields no evidence that the family planning program improves children’s cognitive function. Lastly, using the phasing-in of the measles vaccine within the treatment area to hold constant the effect of the family planning program and other interventions, the ITT effect was similar at approximately 0.33 standard deviations. While any one of these analyses is insufficient, taken together they suggest the child health interventions likely played a large role in improving cognitive functioning.” (pp. 270-271)

• 66

  Karra et al., 2022, p. 2: "However, findings from the Matlab program have been extensively debated, with critics noting that the bundling of FP with other MCH services makes it difficult to disentangle the independent impacts of FP (13, 25). In addition, the potential nonrandom selection of intervention and comparison areas has sparked questions about the extent to which causal inferences can be made from the program"

• 67

  The fertility rate in Bangladesh was 6.6 in 1977 compared to 4.6 in low-income countries as of 2022.

• 68

  For example, female educational attainment has risen over this time period (UNESCO Institute for Statistics, 2024). Duflo et al. (2021) randomize secondary school scholarships and find that female scholarship recipients have fewer unwanted births. The evidence in this paper is consistent with the reduction in unwanted births being driven by increases in the opportunity cost of bearing and raising children, the ability to make better choices thanks to better decoding of information, and changes in the type or preferences of the partner (p. 18), which suggests that it is not entirely driven by increases in modern contraceptive usage.

• 69

  For example, using an estimate of .5 implies that modern contraception results in 8.3 fewer unintended pregnancies over a woman’s lifetime (assuming women want to avoid pregnancy for ~16 of their 34 child-bearing years based on the Guttmacher Institute’s estimate of % of women who want to avoid becoming pregnant and are fecund and sexually active in 17 LMICs; details here) and ~4 unintended births, which seems high relative to the total fertility rate in low-income countries with low rates of contraception. For example, Chad’s total fertility rate was 6.2 in 2022 ( United Nations Population Division. World Population Prospects: 2022 Revision) with only ~6% of women 15-49 years old using modern contraception based on Guttmacher Institute estimates (see here). However, it is important to note that the modern contraceptive usage rate might understate the influence of contraception on the total fertility rate if modern contraceptive use is concentrated among those with high pregnancy risk.

• 70

  Traditional methods have a failure rate of ~14% (interquartile range of ~7% to ~21%; Polis et al., 2016, Figure 3), which means that inducing traditional method users to adopt modern contraceptives (failure rate of ~4% in Polis et al., 2016) would lead ~.09 fewer pregnancies. Thus, many non-users of modern contraception would need to be using traditional contraceptives for 0.2 to be the effect on modern contraceptives. The Guttmacher Institute’s estimates imply that roughly 10-20% of non-users of modern contraception who want to avoid pregnancy are using traditional contraceptive methods, though we have not looked into these estimates in-depth.

• 71

  Examples of panel data sets from LMICs include the Performance Monitoring for Action dataset and the MLE component of the Urban Reproductive Health Initiative.

• 72

  Minor sources of uncertainty include:

  • How precise is our weighted average of study estimates? We do not estimate confidence intervals around our weighted average of empirical study estimates. The wide variance of estimates (0.6 to -0.6 unintended pregnancies per year of modern contraception) and few women who adopt modern contraception per trial (162; refer here for this calculation.) suggest that the confidence intervals would be relatively wide. In the future, we could use meta-regression analysis to quantify the confidence intervals on this estimate and adjust our belief in the estimate accordingly.
  • How should we weigh the empirical studies? We do not adjust the weight on each empirical study based on whether treatment was assigned individually or to clusters. This could cause us to put too much weight on clustered trials such as the Matlab experiment. We also do not adjust the weight on a study’s estimate for the internal or external validity of the study. This could mean we are putting too much weight on studies such as Kosgei et al. (2011), which was a retrospective cohort study rather than randomized experiments.
  • What would be the failure rate of modern contraceptives among women who take up modern contraception due to GiveWell-funded family planning programs? Our estimate of failure rates comes from studies such as Polis et al. (2016) which survey current users of modern contraception. We adjust our estimate downward to account for the fact that women induced to use contraception by family planning programs might have higher failure rates, but we are uncertain about the direction and magnitude. We might change our opinion if we investigated it further.
  • What types of modern contraception will women who take up modern contraception due to GiveWell-funded family planning programs use? The failure rate of the types of modern contraceptives used influences our estimate of unintended pregnancy averted. That’s because higher failure rates→more unintended pregnancies among users→less unintended pregnancies attributed to non-users. Polis et al. (2016) finds that there are higher failure rates for some methods (condoms and birth control pills) compared to others (IUDs or injectables). If we have evidence that a specific grant will disproportionately drive adoption of certain types of modern contraception, we could adjust failure rates accordingly. However, we are concerned about potential coercion into using certain methods (e.g. LARCs), so we plan to be wary of programs that push certain methods rather than allowing women to choose the contraceptive method that works best for them.
  • How accurate are the Guttmacher Institute’s estimates of unintended pregnancies? We did not review their methodology in detail, so it is possible that they are making an error that would shift our estimates.
  • Should we incorporate evidence from high-income countries? We did not look at studies from high-income countries. Given the sparse evidence from LMICs, incorporating high-quality evidence from high-income countries could shift our beliefs about this parameter, even after accounting for the fact that these results may not fully generalize to LMIC contexts.
  • How should we account for postpartum insusceptibility? Two of the programs (accounting for ~5% of the weight in our estimate) likely oversampled women who were postpartum insusceptible. Removing these two studies would increase our estimate of unintended pregnancies averted from .30 to .31. In the model using survey data, we account for women who are “experiencing postpartum amenorrhea after an intended pregnancy” potentially adopting. However, the Guttmacher Institute classifies women “experiencing postpartum amenorrhea after an unintended pregnancy” as women in need of modern contraception (Riley et al., 2020, p. 23). We think they account for the fact that these women have low pregnancy risk when they calculate their unintended pregnancy rates among women in need, but we have not looked into this in-depth.

• 73

  Bearak et al. (2020) provides more recent estimates, but they do not include separate estimates for estimates of unwanted and mistimed births. For reference, Bearak et al. (2020) estimates that 37% of unintended pregnancies in sub-Saharan Africa end in abortion, so they are not far off from Sedgh, Singh, and Hussain (2014).

• 74

  For example, Bearak et al. (2020) estimates that the proportion of unintended pregnancies that end in abortion rose between 1990-1994 and 2015-19. While we use Bearak et al.’s estimate for abortion, we do not have up-to-date estimates for the proportion of unintended births that are mistimed vs. unwanted.

• 75

  Refer to What about lives that don’t occur due to contraception? section for more information on this decision.

• 76

  This assumption could be wrong, but it is based on women’s reported fertility intentions (Sedgh, Singh, and Hussain, 2014; pp. 8-9). Averting an unwanted birth means that a woman does not want to have any more children, so their intention is to have 0 more children regardless of whether or not the unwanted birth is averted. Whether or not an abortion or miscarriage is averted does not affect a woman’s number of children so likely does not affect the number of children they intend to have.

• 77

  It is possible that women who avert a mistimed pregnancy and intend to have a later birth will never have that birth due to changing their mind or fertility issues.

• 78

  This is the moral weight of averting a maternal death that we’ve used previously in our BOTEC of facility-based maternal and neonatal health interventions. Our reasoning is, “This estimate has been derived by assuming a distribution of the age of pregnancy, and additionally making an assumption that maternal deaths are likely to have substantial impact on a child's health and development. We estimate a direct value from preventing maternal death of 110 and an indirect value from other effects of maternal death of 15.” For more on our moral weights, see this write-up.

• 79

  This is the average of a UN estimate for low-income countries from 2020 and an IHME estimate for low SDI countries in 2021. See here for UN estimate for low-income countries in 2020 and here for IHME estimate for 2021 for low SDI.

• 80

  Based on Adding it up 2019 methodology report Table MA3.4: The % of maternal deaths attributable to abortion in sub-Saharan Africa (7%) and miscarriages/ectopic pregnancies (5%; we include ectopic pregnancies in the miscarriage category since ectopic pregnancies prevent the pregnancy from being brought to term). We use sub-Saharan Africa as a proxy since there was not a number for all LMICs. Refer to “Maternal mortality calculations” in Valuing contraception model for the details on these calculations.

• 81

  A secondary uncertainty is whether we should adjust the maternal mortality ratio to account for the women affected by the program being different than the average woman. We chose not to make any adjustment based on a study on the characteristics of women with unmet need for modern contraception in LMICs (Anik, Islam, and Ramik, 2022). This study did not find a clear pattern in terms of the relative socioeconomic status of women with unmet need, making it unclear how we should adjust the maternal mortality ratio estimate in terms of direction or magnitude.

• 82

  This is the moral weight of averting a maternal death that we’ve used previously in our BOTEC of facility-based maternal and neonatal health interventions. Our reasoning is, “This estimate has been derived by assuming a distribution of the age of pregnancy, and additionally making an assumption that maternal deaths are likely to have substantial impact on a child's health and development. We estimate a direct value from preventing maternal death of 110 and an indirect value from other effects of maternal death of 15.” For more on our moral weights, see this write-up.

• 83

  We rely primarily on Cleland et al. (2012) because it is the most reliable estimate we found in a shallow review of the evidence. The other estimate we found Bauserman et al. (2020) only focused on the effects of improved birth spacing in observational data which would not capture the overall benefits of modern contraception on the safety of a birth for the mother (e.g. averting pregnancies by young mothers). Refer to the “Mortality benefits due to improved birth timing” sheet.

• 84

  p. 5 of appendix of Cleland et al. (2012): "We conducted a panel regression analysis of change in the MMRatio as a function of change in the contraceptive prevalence rate (CPR) and baseline MMRatio and CPR in these 40 countries...Thus for each percentage point gain in the CPR, the MMRatio declines by 4.3 deaths per 100,000 births."

• 85

  See calculations here.

• 86

  We are not discounting the benefits for time between a potential mistimed birth and the later, intended birth. That's because we do not think it would make a large difference. We would only be discounting for temporal uncertainty (1.4% of discount rate) and the gap between these potential births is likely <10 years.

• 87

  Brown and Eisgenberg, 1-2995, p.1-2 “In addition, an unintended pregnancy is associated with a higher probability that the child will be born to a mother who is adolescent, unmarried, or over age 40—demographic attributes that themselves have important socioeconomic and medical consequences for both children and parents. Pregnancy begun without planning and intent also means that individual women and couples are often not able to take full advantage of the growing field of preconception risk
  identification and management, nor of the rapidly expanding knowledge base regarding human genetics.”

• 88

  Maternal Health Initiatives CEA includes postpartum anemia, Cesarean section for obstructed labor, fistula, postpartum depression, and chronic hypertension. Refer to the “Maternal morbidity” sheet here for details on the calculations.

• 89

  In Table 2 of Higashi et al., 2014, they estimate 34 deaths due to abortion in LMICs and 2128 DALYs lost. Assuming each maternal death loses 35 DALYs, this implies that 45% of the burden of abortion is attributable to morbidity rather than mortality.

• 90

  A WHO report from 2000 notes that “No attempt has been made to quantify the burden of disability and deaths due to spontaneous abortion for the GBD 2000” (p. 1). Spontaneous abortion is another term for miscarriage. We think that recent versions of the GBD have also not attempted to quantify the burden from miscarriages. The most recent version (GBD 2021) appears to bucket abortions and miscarriages together.

• 91

  WHO report from 2000: “Spontaneous abortions seldom have severe complications and are rarely fatal” (p. 1)

• 92

  See “Mortality benefits due to improved birth timing” sheet in our Valuing contraception model for more details on the infant and under-5 mortality rate estimates.

• 93

  This is based on the 0.25 ratio of value from projected income effects to value of lives averted used in our vitamin A supplementation CEA.

• 94

  E.g., we apply a 74% internal validity discount and a 30-50% external validity adjustment in our vitamin A supplementation CEA.

• 95

  See calculations here.

• 96

  We do not use these as our main estimates because short-timed births (often defined as <18 months between pregnancies) are relatively infrequent even in countries with relatively low modern contraceptive usage rates. For example, in Chad (6% of women 15-49 use modern contraception according to Guttmacher Institute; see here), only 11.2% of women’s last births were short-spaced (see DHS StatCompiler; Chad 2014-15 DHS; short-spaced defined as a previous birth interval of 7-17 months). However, we should note that the modern contraception usage rate may understate the influence of modern contraceptives on birth spacing if the women who are using modern contraception are at high-risk of short-spaced pregnancies.

• 97

  For example, see the Additional Benefits section of our seasonal malaria chemoprevention intervention report.

• 98

  Barham et al., 2020 writes that “Relatively few other studies have demonstrated long-term effects of family planning programs, in part due to the difficulties of longitudinal follow-up, biased self-selection into treatment, and the lack of appropriate comparison groups.” (p. 2)

• 99

  See description of Matlab program here.

• 100

  Refer to SI_Table_4 and SI_Table_5 in Barham et al. (2020).

• 101

  Refer to the “Income effects” sheet of our Valuing contraception model for more details on these calculations.

• 102

  Miller (2010), p. 781

• 103

  This is a rough assumption but accords with our understanding that all areas were increasing contraceptive use over this period but that the program substantially accelerated this process.

• 104

  Refer to the “Income effects” sheet of our Valuing contraception model for more details on these calculations.

• 105

  See detailed calculations here.

• 106

  Refer to the “Household size” sheet in our Valuing contraception model for details on the household size estimates and “Pregnancies averted” for the projected unwanted births averted over a lifetime.

• 107

  More on GiveWell’s discount rate here.

• 108

  See Ontiri et al., 2023 p. 4-5 for qualitative evidence that this is a primary motivation of women.

• 109

  E.g. Alano and Hanson (2018): “The women’s experiences also indicated that contraceptive use improved the educational status of their daughters.” In addition, Family Empowerment Media,an NGO working on family planning in sub-Saharan Africa, reported that their surveys of women showed that primary drivers of interest in contraception were “Better ability to pay for necessities (e.g. food and clothing) for children” and “Better ability to educate/train their children”.

• 110

  “Subfecundity is defined as having fewer children than desired for reasons that are not in any way linked to motivations or behaviors” (Bougma et al., 2015, p. 5)

• 111

  One reason for why the subfecundity instrument might be flawed is that subfecundity depends on the desired children of the woman which could be correlated with socio-economic status. The first child gender instrument might be flawed because first child gender depends on whether the family uses sex-selective abortion which is likely correlated with socio-economic status.

• 112

  Refer to the “Benefits for existing children” sheet here for details on this estimate.

• 113

  Refer to the “Income effects” sheet for more details on these calculations.

• 114

  Our CEA on deworming has a note describing why they estimate 8 years between the intervention and the child working for income. We added 5 years to this estimate because, in this context, the child could be 0-16 years old rather than 5-16 years old.

• 115

  We use the total fertility rate in low-income countries from the United Nations Population Division. World Population Prospects: 2022 Revision.

• 116

  Refer to the “Pregnancies averted” sheet.

• 117

  We use the under 20 mortality rate in Low SDI locations from the IHME.

• 118

  Note that it would not make sense to add together this effect and the effect on existing children’s education, because investing in existing children’s education (through paying school fees or providing better nutrition) would mean that less of the additional resources would go to consumption for other family members.

• 119

  Refer here for details on this calculation.

• 120

  For example, we assume that a household member is a drain on household resources from ages 0-13, are net-neutral from ages 13-20, contribute to household resources from ages 20-54, are net-neutral from ages 55-64, and then become a drain again from ages 65-85. If we instead assumed that household members are only a drain from ages 0-10, then our estimate would fall to 0.14. There is evidence that children aged 10 to 15 years contribute to farm productivity in rural Tanzania (Andre, Delesalle, and Dumas, 2020), though we are uncertain whether this evidence would apply to most settings where GiveWell would consider funding family planning programs.

• 121

  “First, estimation by age group finds a large and significant effect of the child health program on children born after the child health interventions were introduced, but a small, negative, and statistically insignificant effect of the program for children who were born before the child health interventions began, but whose mothers were eligible for the family planning and maternal interventions. While this analysis is by itself insufficient, since family resources differ for earlier- and later-born children, adding a control for the family planning program in the ITT and TOT models still yields no evidence that the family planning program improves children’s cognitive function. Lastly, using the phasing-in of the measles vaccine within the treatment area to hold constant the effect of the family planning program and other interventions, the ITT effect was similar at approximately 0.33 standard deviations. While any one of these analyses is insufficient, taken together they suggest the child health interventions likely played a large role in improving cognitive functioning.” pp. 270-1 of Barham et al. (2012).

• 122

  Table 7 of Maggio, Karra, and Canning (2023) has the mediation analysis. In Maggio, Karra, and Canning (2023), “Women assigned to the intervention arm received a comprehensive family planning package of services over a two-year period, which included the following: (1) an information brochure and up to six counseling visits from trained family planning counselors; (2) free transportation to a high quality private family planning clinic in Lilongwe; and (3) financial reimbursement for family planning services, including for the treatment of contraceptive-related side effects.” (p. 1673). It is possible that these other aspects of the program drove the effects rather than the 5.9 pp increase in contraceptive usage. (p. 1677)

• 123

  See our discussion of the evidence on parental reports here.

• 124

  It is exactly equivalent to the average of averting mild anxiety (.05) and moderate anxiety (.15) in the 2019 IHME GBD study. It is slightly higher than the average for GBD 2021 which estimates a disability weight of .03 for mild anxiety and .13 for moderate anxiety.

• 125

  Since the drafting of this page, this report has become inaccessible online. The report was initially available at this link and may become available in the future.

• 126

  “The range of acceptable prices for Protector Condoms is UGX 750 lowest and UGX 1800 highest, for Injectaplan the lowest is UGX 1500 and highest is UGX 5000, and Pilplan Plus the lowest is UGX 1300 and highest is UGX 4550.” (Since the drafting of this page, this report has become inaccessible online. The report was initially available at this link and may become available in the future.) We use the following exchange rate: ~3600 UGX = 1 USD. Estimated years of contraception per device come from this USAID page.

• 127

  “We find that discounts for LARCs had a large effect on their adoption: while 32% of all eligible clients adopted a LARC at full price (CFA 4,000 or approximately USD 7.25), this share increased to 45% at mid prices (USD 1.8 or 3.6), and 48% at low prices (free or USD 0.25).” (Athey et al., 2021, p. 3)

• 128

  “The high upfront costs of LARCs, along with high cost of removals, contrast with the per-month cost of modern contraceptive use from unintended pregnancies. The IUD can be used for up to 10 years while the implant lasts three-to-five years. In contrast, the injectable (Depo-Provera) provides protection for only three months, while the pill needs to be taken everyday” (Athey et al., 2021, pg 3, footnote 1)

• 129

  “On average, respondents were willing to pay 11 birr ($0.65 USD) per injection” (Prata et al., 2013, Findings section).
  “Approximately 3% of women were willing to pay between 50 and 200 birr” (Prata et al., 2013, Figure 2)

• 130

  “Based on 2017 estimates, the annual cost of providing contraceptive services to 4.6 million women aged 15–19 who use modern contraceptives in sub-Saharan Africa is $58 million. This averages to $13 per user annually.” (Adding It Up: Costs and Benefits of Meeting the Contraceptive Needs of Adolescents in Sub-Saharan Africa, ‘Cost of meeting contraceptive needs’ section)

• 131

  This is what we assume in our GiveDirectly CEA.

• 132

  Calculations here.

• 133

  See calculations here.

• 134

  Calculations are here.

• 135

  This figure implies that our target population would be willing to pay ~37.5% of their income for modern contraception. (calculation: 1/1.6 = 0.625, implies a 1-0.625 = 0.375 = 37.5% reduction in income)
  We acknowledge that this figure appears quite high, and significantly exceeds the figures implied by willingness-to-pay studies (~2%-20%; see discussion here). However, as discussed here, willingness-to-pay studies have significant limitations (e.g. individuals may underestimate their risk of unintended pregnancy without contraception) that could cause them to understate the value of modern contraception.

• 136

  IHME describes this condition as “feels persistent sadness and has lost interest in usual activities. The person sometimes sleeps badly, feels tired, or has trouble concentrating but still manages to function in daily life with extra effort.”

• 137

  IHME describes this condition as “feels anxious and worried, which makes it difficult to concentrate, remember things, and sleep. The person tires easily and finds it difficult to perform daily activities.”

• 138

  Consistent with our model for valuing contraception (see here), we apply a discount rate of 4% to income gains.

• 139

  The Rethink Priorities’ replication was of a simplified scenario in the Impact 2 model (e.g. it does not consider interaction effects between methods). Their replication does not include several additional downstream effects of averted pregnancies such as maternal and child DALYs averted, direct healthcare costs averted, environmental impact, and other outcomes, such as long-term child health or economic benefits to the families. A list of all benefits considered in Impact 2 can be seen in section 3.3 pg 27 of their methodology paper.

• 140

  See page 36, section 3.3.5 of their methodology report.

• 141

  See Annex 9 of the Impact 2 methodology for details on how they estimate child mortality.

• 142

  See Annex 9 of the Impact 2 methodology (p. 89).

• 143

  See here for sources we use for under-5 mortality rates.

• 144

  See DHS StatCompiler for 17 LMICs here. About ~10% of women are pregnant each year and the median duration of postpartum insusceptibility is ~1-2 years.

• 145

  Specifically, we take a weighted average of the best guess for low-income countries (.5 weight), best guess for LICs multiplied by ratio of the relevant country's unintended pregnancies per woman to the median country of the 17 LMICs (.25 weight), and best guess for LICs multiplied by ratio of the relevant country’s unintended pregnancies averted estimate to the median country in "Adjusted pregnancies averted model using surveys" (.25 weight). We put weight on unintended pregnancies per woman because of uncertainty around whether using a fixed failure rate for modern contraceptives across countries is a valid assumption. We do not simply use the country-level estimates because we have reservations about the accuracy of these non-experimental estimates (refer to the model using survey data from LMICs section). Refer to sheet “Adjusted pregnancies averted model using surveys” here for observational data estimates. Refer to sheet “Heterogeneity across locations” here for updated estimates.

• 146

  Data can be accessed here. Their methodology is described in Bearak et al. (2020).

• 147

  The maternal mortality ratio averages the estimate for 2021 from IHME and estimate for 2020 from WHO, UNICEF, UNFPA, World Bank Group, and UNDESA/Population Division. The neonatal and under-5 mortality estimate averages the UN IGME estimate of neonatal and under-5 mortality in 2022 and the IHME estimates in 2021.

• 148

  IHME under 20 mortality rate estimates from 2021. We did not see an estimate from the UN though we did not search very long since this is not a highly influential parameter.

• 149

  The total fertility rate comes from the United Nations Population Division 2022 estimates. The household size estimates come from a 2017 UN report.

• 150

  We chose Nigeria, Burkina Faso, Togo, the Democratic Republic of the Congo (DRC), and Chad because these are countries that we work in frequently. We included Malawi and Indonesia because they were the highest and lowest in our estimated number of unintended pregnancies averted per year of modern contraception in our initial estimates from “Adjusted pregnancies averted model using surveys” sheet here.

• 151

  Refer to the “Heterogeneity across locations” sheet here for more details on these calculations.

• 152

  Refer to the “Heterogeneity across locations” sheet here and rows here for a full breakdown of the drivers of differences between the low-income country estimate and DRC/Indonesia estimates.

• 153

  See here for Guttmacher estimates by age group. See here for their data visualization tool with 15-19 years old and 15-49 years old age groups.

• 154

  Darroch et. al 2016: "Fifteen million of these adolescent women use modern contraceptives, thereby preventing 5.4 million unintended pregnancies. Of these pregnancies, 2.9 million would have ended in abortion at current rates."

• 155

  Data can be accessed here.

• 156

  For older women, preliminary evidence from Karra, Maggio, and Canning (2023) suggests short-run benefits of family planning on household income for older women so, once this paper is finalized, it could be used as a benchmark to estimate economic effects among older women.

• 157

  In their supplemental materials, Nove et al. (2014) provide estimates of maternal mortality for 144 countries in five year age ranges. These estimates are from 2012 so they would need to be adjusted to reflect any recent changes in maternal mortality.

• 158

  Finlay et al. (2011) estimate associations between maternal age and child outcomes across 55 low- and middle-income countries. We could adjust the population-level child mortality data based on the magnitude of the effect sizes estimated by Finlay et al. (2011). We would want to adjust Finlay et al. (2011)’s estimates to account for confounding since they are estimating associations in observational data.

• 159

  This blog post provides more details on how we incorporate the value of information into our grant-making.

• 160

  By decision-relevant, we mean when we think these adjustments could change our decision on whether or not to investigate the grant opportunity further.

• 161

  “We estimate that at least 37,000 additional women were using modern contraception because of the pilot mass media campaign, suggesting an annual cost per additional woman using modern contraception in the pilot of US$ 42.5. Under reasonable assumptions, this annual cost dropped to US$ 7.7 when the media campaign was scaled nationwide. We estimate this scale-up led to 225,000 additional women using modern contraception in Burkina Faso and roughly 10,000 fewer births a year. While rigorous data on the cost per Couple Year of Protection (CYP) achieved through other approaches is limited, estimates range from US$30 to US$60” (Glennerster et al., 2023, p. 5)

• 162

  0.7/$30/0.00335 = 10x, 0.7/$60/0.00335 = 3.5x. Our “10x bar” means “10 times more valuable than cash transfers from GiveDirectly according to 2023 cost-effectiveness analysis.” In the past year, we have reevaluated GiveDirectly, but we are still deciding how to update our bar.