GiveWell recently published a cost-effectiveness analysis on contraception: Valuing Contraception in GiveWell’s Cost-Effectiveness Analysis | GiveWell.
At MSI Reproductive Choices, we welcome GiveWell’s cost-effectiveness analysis on contraception in low and middle-income countries. The findings affirm what we’ve long seen: access to contraception transforms lives – not just for individuals, but for entire families and communities – unlocking better health, greater autonomy, and expanded economic opportunities for women and girls.
With nearly five decades of experience and over 220 million people served globally, MSI has witnessed this impact firsthand. Our rich data, drawn from over 15,000 client exit interviews in 2024 alone (215,000 over past decade), shows that 98% of clients report a positive impact of contraception their lives – in health, education, family wellbeing, and more.
We also appreciated the opportunity to collaborate with GiveWell and Rethink Priorities in reviewing our Impact 2 model. While our approaches differ in some areas, we found meaningful alignment in others. These conversations are helping shape the next iteration of MSI’s Impact 2, which we’re currently updating with more dynamic, scenario-based modelling to reflect the latest evidence and counterfactual insights. As an open-source, peer-reviewed tool rooted in widely accepted sexual and reproductive health rights (SRHR) methodologies, we remain committed to strengthening Impact 2’s capabilities to better support the wider SRHR community.
One area we’re especially focused on strengthening is how we account for the social and demographic realities of the people we serve when modeling our impact. Not all clients start from the same place, and the benefits of contraception are not equally distributed. In countries like Nigeria, for example, maternal mortality is up to 80% higher among women living in extreme poverty—raising critical questions about how we measure cost-effectiveness and impact. How do we reflect the greater value contraception provides to women in the lowest income quintiles, who face the highest risks and the fewest alternatives? As we refine our modeling, we’re prioritizing this equity lens to ensure our resources maximize impact. Our clients are disproportionately poorer, younger, and more rural than the general population—and for two-thirds of them, MSI Reproductive Choices is the only known source of contraception. That’s the difference we aim to capture more accurately in our models—and the difference we strive to make every day.
We’ll continue to reflect and engage, as this report has reignited some important questions for us. We're keen to share, debate, learn, and evolve together – with a consistent commitment to grounding our decisions in solid, reliable data and evidence. Cost-effectiveness remains a core part of our project cycle, and we're always looking for ways to strengthen our approach through collaboration and evidence.
Models and estimates are vital tools—they help guide decisions, spark dialogue, and strengthen accountability. But behind every data point is a woman whose ability to choose if or when to have children changes everything. That choice goes beyond just a human right, it’s one of the most powerful, life-altering interventions we can support. It fuels health, opportunity, and economic progress across generations.
If you’ll be attending EA Global London next month in June, our Director of Evidence and Impact, Anisa Berdellima, will be there and would welcome the opportunity to connect and discuss the GiveWell analysis and share more of MSI’s insights.
Thanks very much for sharing this. I appreciated the explicit section on the lives prevented from existing. Having said that, I'm pretty skeptical of some of your reasoning.
As an initial framing, I think this should be a bit of a red flag. Typically in the absence of an expert consensus on the value of some parameter (how much to discount potential lives), we would adopt an intermediate value as our prior, perhaps weighted by the rough distribution of experts. Exactly what is the most appropriate prior is up for debate, but it seems clear it should be somewhere within the range. In contrast, assuming a 100% discount for potential lives, as you do, it taking one of the most extreme positions possible.[1]
You offer some specific reasons for this extreme position, but all seem weak to me.
First reason:
Autonomy might be a good, to be weighed against potential lives - but it doesn't mean you get to just zero-out part of the cost-benefit ledger. Or it might be a side constraint on what we can do - but 'decline to donate to this specific thing' is not the sort of action that would typically be considered to violate a side constraint. (If it was, then you would arguably be guilty anyway, for failing to donate to IVF and other fertility treatments for couples who want them).
This is an especially big deal because you do consider autonomy as a positive contributor, with quite large weight:
The category they are in (admittedly they are not all of this category, but surely a lot of them), accounts for 0.23 'units of value' out of a total of 0.7, or 0.23/0.7 = 33% of the total.
So autonomy is getting double-counted: [at least one aspect of it] both adds to the positive and somehow 'nullifies' a negative.
Second reason:
This seems like a situation where it is important to read the citation... given the text, I was not expecting the citation to say that actually most women and men want more children! But regardless, this argument seems... beside the point? The fact that (a minority of) people don't wish to experience the costs of more children doesn't mean that those children wouldn't have valuable experiences and worthwhile existences. You are taking into account many of the costs that parents experience from having children (e.g. opportunity cost), and then assuming the benefits are zero. In fact, because parents pay most of the costs of having children, but do not capture all or even most of the benefits, which suggests they should under-value having more children.
Third reason:
The fact that someone does an activity doesn't mean they evaluate the costs as being zero - it just means they think the costs are smaller than the aggregate benefits. (Or, perhaps more likely, that they have never done a proper cost-benefit analysis).
I'm not sure why we should be deferring to these groups here. To the extent that other groups have specific arguments for their position, we should consider them. But the mere fact that some other funder is supporting a program... should surely update us against the program, because it reduces neglectedness, not in favour of it.
Finally, I don't think you would make this argument in other contexts. You call out the government of Ghana specifically as a positive example worthy of deference to their moral views - would you also exhibit significant deference to them on the issue of LGBT rights, where they recently unanimously passed a bill imposing prison sentences for same-sex intercourse or promoting LGBT? (The bill seems to have not gone into effect for procedural reasons, but might be re-introduced). I suspect not - I think you would assign this basically zero moral weight when considering an AIDS charity - which suggests to me that your moral view came first, and then you looked for congenial authorities to defer to.
And the question of moral weight on potential lives turns out to matter a great deal! I think you deserve a lot of credit for making this explicit:
I encourage you to take this seriously. If you are uncertain, probably you shouldn't set the parameter at one of the most extreme values possible!
Yes, technically you could assign values less than 0% or more than 100%, but I don't think this is very credible, especially as we generally consider young children to have positive lives in expectation.
One reason to be suspicious of taking into account lost potential lives here is that if you always do so, it looks like you might get a general argument for "development is bad". Rich countries have low fertility compared to poor countries. So anything that helps poor countries develop is likely to prevent some people from being born. But it seems pretty strange to think we should wait until we find out how much development reduces fertility before we can decide if it is good or bad.
I agree that all else equal it is an argument against development, though I still think development is good overall on net, and I think the extent to which development specifically is the cause of reduced fertility has been overstated.
More generally, I am very skeptical of arguments of the form "We must ignore X, because otherwise Y would be bad". Maybe Y is bad! What gives you the confidence that Y is good? If you have some strong argument that Y is good, why can't that argument outweigh X, rather than forcing us to simply close our eyes and pretend X doesn't exist?
Is not every moral theory based on assumptions that X must be better than Y, around which some model is built?
Yes, I have an intuition that development is good, just like I have an intuition that ice cream is good. That doesn't mean that the price of the ice-cream should be ignored and assumed to be zero when deciding when to buy it, and nor should the costs of development be ignored and assumed to be zero.
Yes, but if at some point you find out, for example, that your model of morality leads to a conclusion that one should kill all humans, you'd probably conclude that your model is wrong rather than actually go through with it.
It's an extreme example, but at its basis every model is somehow an approximation stemming from our internal moral intuition. Be it that life is better than death, or happiness better than pain, or satisfying desires better than frustration, or that following god's commands is better than ignoring them, etc.
"More generally, I am very skeptical of arguments of the form "We must ignore X, because otherwise Y would be bad". Maybe Y is bad! What gives you the confidence that Y is good? If you have some strong argument that Y is good, why can't that argument outweigh X, rather than forcing us to simply close our eyes and pretend X doesn't exist?"
This is very difficult philosophical territory, but I guess my instinct is to draw a distinction between:
a) ignoring new evidence about what properties something has, because that would overturn your prior moral evaluation of that thing.
b) Deciding that well-known properties of a thing don't contribute towards it being bad enough to overturn the standard evaluation of it, because you are committed to the standard moral evaluation. (This doesn't involve inferring that something has particular non-moral properties from the claim that it is morally good/bad, unlike a).)
A) feels always dodgy to me, but b) seems like the kind of thing that could be right, depending on how much you should trust judgments about individual cases versus judgements about abstract moral principles. And I think I was only doing b) here, not a).
Having said that, I remember a conversation I had in grad school with a faculty member who was probably much better at philosophy than me claimed that even a) is only automatically bad if you assume moral anti-realism.
Quick note I think this is GiveWells reasoning not @Meghan Blakes the OP. She might want to respond though regardless.
Thanks Meghan I'm keen to meet up with Anisa in EA Global! Our OneDay Health centers provide some contraception in remote areas in Uganda. Although its not our no.1 focus we're keen to improve our family planning coverage.
I think for many orgs, including all programatic costs ,$20 for a year of family planning (the preliminary GiveWell bar) might not be the easiest bar to meet. What are your thoughts on this?
Hi Nick! We'd love to connect, while I won’t be at EA Global myself, our Director of Evidence and Impact, Anisa, will be attending and would be happy to set up a meeting via Swapcard once it launches. We’re really interested in learning more about OneDay Health’s work.
On the $20 GiveWell benchmark, it's important to note that this refers specifically to $20 per additional year of family planning that wouldn’t have happened without the intervention. Across MSI’s “last mile” delivery channels, Outreach and Public Sector Strengthening, our average global cost per year of protection is $4.88 (this includes all CYPs). We're currently analyzing which of our programs meet the $20 per additional CYP threshold and expect that many, though not all, will. At MSI, cost-effectiveness is a key focus, and we recognize that this can be a high bar, especially when all programmatic costs are included, but we think there may be select other organizations that can meet it too.
Thanks Meghan yes I'll definitely organise a meeting with Anisa!
It seems to me like it would be challenging for a big international organistion to manage $5 for a year of protection including all management and overheads (often the biggest cost for NGOs), but if that's in the ballpark that's really impressive.
Do you have a public calculation you can share at all? All good if not!