The GiveWell Blog

David Roodman’s draft writeup on the mortality-fertility connection

We’re sometimes asked whether life-saving interventions (such as insecticide-treated net distribution) risk leading to overpopulation. A common response is that (as argued by Hans Rosling of Gapminder and the most recent Gates Foundation annual letter, among others) the reverse dynamic holds: population growth tends to slow as child mortality declines, possibly because parents feel less need to have multiple children in order to hedge against the risk of death.

Empirically, it appears generally true that lower child mortality tends to correlate with slower population growth. However, one might ask whether this relationship is causal (i.e., falling mortality leads to falling fertility, for reasons outlined in the previous paragraph) or merely correlational (i.e., falling mortality generally accompanies a host of factors related to general economic development, and including falling population growth). This question is important when assessing the impact of a life-saving intervention in isolation: it’s possible that health interventions, unaccompanied by improvements on other dimensions, do in fact lead to higher populations.

We’ve long wondered about this question, and we recently commissioned David Roodman – whom we have been impressed with in the past – to examine the rather extensive and complex literature on it. David has recently posted a draft of his writeup, and is seeking feedback.

We haven’t yet fully vetted the writeup, but it appears to me to give a deep and balanced examination of the available literature, along with interesting discussion of the general challenges of isolating causality in cross-country studies. Its conclusion:

I think the best interpretation of the available evidence is that the impact of life-saving interventions on fertility and population growth varies by context, above all with total fertility, and is rarely greater than 1:1 [rarely enough for the fertility decline to exceed the mortality decline in terms of effect on population size]. In places where lifetime births/woman has converging to 2 or lower, family size is largely a conscious choice, made with an ideal family size in mind, and achieved in part by access to modern contraception. In those contexts, saving one child’s life should lead parents to avert a birth they would otherwise have. The impact of mortality drops on fertility will be nearly 1:1, so population growth will hardly change.

Overall, it appears that life-saving interventions unaccompanied by other improvements, where access to contraception is weak, are likely to lead to some acceleration of population growth. With that said, we wish to note the following:

  • No intervention takes place in isolation, and we expect population growth to slow in the future in most low-income areas as poverty falls.
  • Acceleration of population growth should not necessarily be associated with overpopulation and its connotations of a net decline in standards of living.
  • GiveWell’s current top charities are all focused on improving living standards rather than averting deaths, and Against Malaria Foundation (a former and potentially future top-rated charity) aims to avert deaths and bring about other benefits by reducing the burden of malaria.

We encourage our readers to check out the writeup and send in their thoughts. It’s a thorough review of the question and a generally interesting read.

David Roodman’s draft writeup, “The impact of life-saving interventions on fertility”

More on why we commissioned this writeup

We have previously made some preliminary attempts to review the literature on the connection between mortality and fertility. However, we’ve found this literature to be (a) challenging to synthesize, as most of the key papers use relatively complex methodologies that would take significant work to interpret; (b) only a moderate priority for our research, since it has limited relevance to which charities we recommend and how we rank them.

Recently, we decided to explore the idea of working with David Roodman (whom we’ve generally been impressed with) as someone who might increase our capacity to produce thorough reviews of literature on key giving-relevant questions. For our first project together, we chose to work on this question because David had done substantial work on it in the past, and felt he could produce an initial writeup relatively quickly.


  • Jon Claerbout on April 20, 2014 at 3:05 pm said:

    Since it’s just a draft, I suggest Roodman find a better way to present the information on his fourth illustration — or direct us to a table.

  • Colin Rust on April 21, 2014 at 2:24 pm said:

    (adapted from an email to David)

    I learned a lot from reading David’s careful piece. Kudos for commissioning it. This is an issue I’ve long wondered about: I had a vague sense that the consensus view was that philanthropic interventions to reduce mortality actually reduced population growth (and I tentatively accepted this), but I didn’t have a sense of how well-grounded that was. It’s good to have that corrected and clarified!

    This does make me wonder anew about family planning as a philanthropic area.

    Does it make sense as an individual donor to fund family planning as an intervention? Is it a better area to donate to than interventions that, like GiveWell’s top charities, are primarily about reducing morbidity and mortality (though of course the two can overlap, e.g. for providing condoms for people with AIDS). My vague feeling is the answer likely partly depends on moral judgments. But I don’t feel I even have a clear frame of what a dollar can plausibly do compared to what a dollar given to say one of GiveWell’s top charities would do, so I could even start to come up with my own judgment for weighing it.

    And what are the specific charities that are highly effective in this area, if any? It seems to me that family planning is an area with relatively few donor dollars, at least until the Gates’ Foundation’s recent push.

  • Colin Rust on April 21, 2014 at 4:28 pm said:

    Acceleration of population growth should not necessarily be associated with overpopulation and its connotations of a net decline in standards of living.

    Sure, but “necessarily” is a pretty high standard of proof. It would be good to review the empirical evidence, but my prior is pretty strong that at least in the parts of the developing world with high fertility rates, increased population growth is negative for future standards of living. There are plenty of reasons to think this should be so: If there are many children, then land must be divided up into smaller pieces with each generation. Natural resources, such as firewood and water become harder to source. Etc., etc. Holden, are there some positive network effects or something that you think may offset the obvious negative effects?

  • Holden on April 21, 2014 at 6:16 pm said:

    Colin, we share your uncertainty about the area of family planning. There is a fair amount of philanthropy in the area, and our impression is that there is significant unmet need, but we haven’t developed a strong sense of whom one should fund in this area or what results-per-dollar one might expect. For more, see our our conversation notes with the Hewlett Foundation on this topic; we also have a private draft writeup on this area, though completing it is a low priority.

    To your question about positive effects of growing population growth, I think it just depends on whether a given region’s productivity is more bottlenecked by natural resources or by labor. It isn’t clear to me that one or the other should be thought to be overwhelmingly more common. An additional point I should have made clearer in the above post is that many of the regions served by life-saving interventions will, in fact, have access to family planning services, making it less likely (according to David’s writeup) that saving lives leads to population growth.

  • David Roodman on April 22, 2014 at 9:20 am said:

    First, thanks *very* much for your careful reading of the draft and comments in private (the last round of which I still need to incorporate).
    Second, I haven’t done the cost-benefit analysis any more than GiveWell, but for what it’s worth, I am convinced that family planning is in area in which governments, donors, and philanthropists have made a big difference, by making contraception more widely available and by shifting public ideas about ideal family size. For more, you can see the Bongaarts et al. (2012) study cited in the draft review, starting around page 36.

  • David Roodman on April 22, 2014 at 9:21 am said:

    Jon, for the final version, I think we should post an Excel file with all the graphs and data.

  • Colin Rust on April 26, 2014 at 10:45 am said:

    David and Holden, thanks for the readings. One point that struck me about the Hewlett Foundation conversation:

    Geographically, the area most in need of family planning services is Western Africa, which has the highest maternal mortality rate and the highest fertility rate of any area in
    the world. Relative to their needs, Western Africa has too little funding, so a new donor could make an impact in this region. One reason for underfunding is that the largest funders are English-speaking and Western Africa is largely French-speaking.

    It wouldn’t have occurred to me that the language spoken in an area would be a material issue.

    Holden, I’m not sure I understand your argument that if a region’s productivity is “bottlenecked” by labor resources, population growth might be expected to increase standards of living. Could you flesh that out a bit? Productivity is the economic output per worker. It seems to me that with high population growth, there will tend to be a less educated and therefore lower quality work force; that’s a reason on the labor-supply side to expect lower productivity. On the labor-demand side, to the extent that anything other than labor (e.g. arable land) is a constraint at all, I would expect the productivity of an additional marginal worker to be less than the average productivity. Am I missing something?

  • Colin Rust on April 26, 2014 at 12:15 pm said:

    PS Of course, nothing happens in isolation. So e.g. reducing malaria through bed nets should significantly boost productivity more or less immediately because of reduced illness (there are hundreds of episodes of malaria per death). I would guess that effect on productivity would overwhelm the presumably negative effect via (presumably) increased population.

  • Peter Singer on May 1, 2014 at 6:20 pm said:

    This may be a bit pedantic but perhaps it is worth correcting this:

    “rarely greater than 1:1 [rarely enough for the fertility decline to fully offset the mortality decline in terms of effect on population size].”

    The passage in square brackets would be an accurate gloss on “rarely equal to 1:1” but not on “rarely greater than 1:1” If you want to gloss that, it would be better to write “to exceed the mortality decline…”

  • Holden on May 2, 2014 at 9:57 pm said:

    Colin, I agree that the productivity of a marginal additional worker will be lower than the productivity of the average worker. However, marginal people may still carry positive externalities for other people (e.g. by diffusing the cost of public goods over more people). Overall, I’d intuitively expect a community to benefit from population growth in many cases (and think this has been the case for the world as a whole), though I agree that the factors you name could push in the opposite direction.

  • Holden on May 3, 2014 at 1:58 am said:

    Peter, thanks – I’ve made your suggested change.

  • Colin Rust on May 11, 2014 at 12:05 pm said:

    Holden writes:

    [M]arginal people may still carry positive externalities for other people (e.g. by diffusing the cost of public goods over more people).

    Agreed. But it seems quite implausible to me that would be anywhere near sufficient to offset the negative impacts in areas with a high population growth rate. Many cases that come to mind here e.g. around transportation are more about density than population anyway, pointing to benefits of urbanization more than of population growth. In some cases, say hydro dams, higher population might point to a negative externality to additional population (because hydro resources are limited). Some other cases, e.g. around media, culture or inventions of new products and practices may have pretty clear positive externalities related to scale. But like I said it seems intuitively clear to me that such effects are overwhelmed by the negative effects on standards of living. (As an aside, I’d add that my sense is that high population growth also tends to have negative distributional consequences.)

  • Holden on May 15, 2014 at 2:17 pm said:

    Colin, my intuition differs from yours. To give a bit of an outline of where I think my intuition comes from … I think that the gains to phenomena like “unusually successful people migrating and being able to send money and provide connections for the people they grew up with” and “small/subtle, or major, innovations that benefit a community” can be quite large. I think it’s also relevant (though not conclusive) that the world as a whole seems much better off today than it would have been with slower population growth; there are certainly reasons to expect different dynamics for the world as a whole vs. the areas in question, but I still find this observation to provide an important outside view. I feel the burden of argumentation is on one who would argue that population growth is highly likely to be net negative in a given area.

    Just as a point of clarification, this doesn’t mean I am negative on family planning interventions. I find the idea of meeting demand for contraception (and thus bringing people’s actual family sizes into line with their desired family sizes) to be extremely appealing, from the perspective of empowering people and raising their quality of life.

  • Colin Rust on May 17, 2014 at 2:27 pm said:

    Holden, I think our intuition on the world as a whole differs also. I tend to think that slower population growth worldwide would be good for standards of living (and zero or modestly negative growth still better at current population levels). I’d say it’s plausible that very negative population growth (as we see e.g. in Japan with very low fertility combined with negligible immigration) is bad, though even there I’d say the case is not as strong as usually stated. For one thing, when the demographic shift older is discussed, the greater number of elderly to be cared for inevitably (and appropriately) comes up, but on the other side of the ledger the lesser number of young people who are too young to care for themselves or are in full-time education is usually omitted.

    I’m glad you agree that meeting demand for contraception (which is really what we’re talking about) is a good thing. Presumably though, views on the effect on population growth are still important in weighing funding contraception compared to other philanthropic opportunities.

  • Holden on May 19, 2014 at 11:12 pm said:

    Colin, agreed – this sounds like a substantive disagreement and a longer conversation.

  • Colin Rust on May 25, 2014 at 1:58 pm said:

    Holden, to reply further to your earlier comment.

    Re “unusually successful people migrating and being able to send money and provide connections for the people they grew up with”: With more population there will surely be more such people, but it’s far from clear to me that there will be more such people (or money or connections) per capita.

    I do agree we expect more beneficial innovations with more population growth and hence more people, and further that though innovation is difficult to anticipate it can be very significant. On the other hand, for example, the risk of war is higher.

    On your point about burden of proof, even if we accept for the sake of argument that the world’s population growth rate at around 1.2% is too low, I don’t think that gets you very far in thinking about cases where the growth rate is double or triple that. The average case doesn’t constrain extreme cases that much.

  • Holden on May 30, 2014 at 6:55 pm said:

    Colin, thanks for the further thoughts.

    Regarding emigration: my point was that the big net gains to a community from a small number of emigrants can make up for the net consumption of other community members, with the result that the total resources per capita available to a community fall less with population growth than they would without emigration. If this point held, I think the externality-related benefits of population growth would likely make it a net positive. (In addition, the information/connection provided by emigrants shouldn’t necessarily be thought of in “per capita” terms.)

    Regarding total world population growth: my point wasn’t that the outside view is definitive. However, it seems to me that most of the common arguments on the “faster population growth is worse” side don’t clearly apply more at one level of population growth than at another. So if population growth is beneficial for the global case (the one that seems easiest to assess), one must specify why the dynamics are sufficiently different in a different area.

  • Randall on June 16, 2014 at 12:57 pm said:

    Finally finished it. I don’t think these comments really add anything, but here’re my reactions:

    – Wow, it’s really fundamentally hard to tell how one big, hairy variable with lots of causes and effects impacts another. Maybe I should have expected it after seeing lots of flawed scholarship in rich countries that, say, doesn’t successfully address income/socioeconomic status’s correlation with the variables being studied. But it’s another thing to see a lot of researchers try and not exactly succeed at clearly answering a question I wasn’t initially aware was so hard.

    – Wow, people publish lots of pretty sketchy claims based on country- or region-level aggregate data anyway. And this report was helpful in understanding how microdata studies are promising for doing things we can’t confidently do with country-by-year data or the like.

    – Huh, some methods aren’t as advanced as I thought–like, we still eyeball a graph to look for the “fingerprints” of a discontinuity in quasi-experiments, though apparently there are some numerical methods too.

    – As much as anything else it may be hard to find the right questions, i.e., ones that are both answerable and have answers you can act on.

    And my reaction about what this means for concerns about philanthropy and “overpopulation” essentially match Holden’s–fertility and mortality *can* both fall in the next decades, even if the mortality decline doesn’t cause the fertility decline; it’s not obvious that some population growth would lead to starvation and so on; and real life-saving interventions improve the world in a lot of different ways.

Comments are closed.