Search the Site

Episode Transcript

Hey there, it’s Stephen Dubner, and we’ve got something special for you this week: an episode of Freakonomics, M.D., one of our sibling shows from the Freakonomics Radio Network. Freakonomics, M.D. is hosted by Bapu Jena, who is both a physician and an economist; every week, he tries to answer a thorny question at the intersection of economics and medicine. Recently, he’s looked at why doctors overprescribe antibiotics, whether Americans are going overboard with colonoscopies, and what happens when physician assistants and nurse practitioners replace doctors. You can get all those episodes, and many more, wherever you get your podcasts. As for the episode we’re about to play for you: there aren’t many questions thornier than this one. It’s a classic example of how it can be really hard to prove cause and effect in healthcare. I hope you enjoy this episode, I hope you’ll follow or subscribe to Freakonomics, M.D. on your favorite podcast app, and I hope you’ll stick around to the end of this episode to hear a preview of a special series that we’re about to run right here on Freakonomics Radio, with me back in the host’s chair. As always, thanks for listening.

*      *      *

Emla FITZSIMONS: My interest in breastfeeding, I suppose, came from my own personal experience when I had my first child, almost 14 years ago.

Emla Fitzsimons gave birth on a Saturday afternoon and stayed in the hospital until Monday morning. At first, she didn’t think much about what day of the week it was. Her mind was on something else.

FITZSIMONS: I really struggled in the early couple of days with breastfeeding. I found it very hard to access support in the hospital. But by Monday I just noticed quite a stark difference in the availability of support. So a midwife was able to spend some time with me and to help me, which I hadn’t had available over the weekend. 

Emla happens to be an economist and researcher at University College London. And this experience stuck with her. Surely she couldn’t be the only woman who delivered over the weekend and received less breastfeeding support as a result.

FITZSIMONS: I suppose that’s what resulted in the idea behind this paper, which is looking at how that support goes on to affect breastfeeding and using that to try to establish the causal impact of breastfeeding on children’s outcomes.

What Emla is talking about here is the question at the heart of many of the debates around breastfeeding: What does it do for the child?

FITZSIMONS: There’s lots of studies out there. But the vast, vast majority are mainly associations.  

It’s much easier, in medical research as in other fields, to show that two things are associated — that they tend to go together — than it is to show that one causes the other.

FITZSIMMONS The usual approach is to compare children of mothers who breastfeed with those who did not breastfeed. So one concern is whether these two groups are truly comparable and they’re generally not. And so therefore any difference in their children’s outcomes may be due to breastfeeding, but it may also be due to other factors.

Other factors like socioeconomic status, where a family lives, whether mom goes back to work, how much support there is at home. So, Emla set about to explore the breastfeeding question a bit more rigorously.

FITZSIMONS: First of all, we examine whether babies born at the weekend are less likely to be breastfed compared to those born during the week. And then secondly, it looks at the impact of breastfeeding for at least three months on children’s cognitive development and health. 

Emla’s clever use of this natural experiment offers some of the best evidence to date on a question that parents and health experts have been arguing over — and worrying about — for decades: How important is breastfeeding?

From the Freakonomics Radio Network, this is Freakonomics, M.D. I’m Bapu Jena. Today on the show: what did Emla Fitzsimons learn?

FITZSIMONS: We found that all of the effect appears concentrated on children’s cognition, and we observed strong effects lasting at least to age five. 

But first, economist and author Emily Oster will explain how she tries to help parents make big decisions using data, and why — with breastfeeding — this can be really hard to do.

Emily OSTER: Some of the ways we have been pressuring people to breastfeed have been counterproductive and can actually be quite harmful.

*      *      *

OSTER: I’m Emily Oster. I’m a professor of economics at Brown University.

Bapu JENA: And you’ve written a lot of books, is that right?

OSTER: Yes, I have three books, one on pregnancy and two on parenting at different ages. I’ve kind of made a career out of the fact that a lot of parents want information and they want data.

Emily’s no stranger to the Freakonomics universe. She’s been a guest before on this show, and on other shows on our network, like People I (Mostly) Admire and Freakonomics Radio. She tries to answer some of the biggest parenting questions by using data. Emily’s books about pregnancy and early parenting have been bestsellers, but her research focus is a little different. She studies health behaviors, and also statistical methods. So, basically, assessing how good research methods — and research itself — actually are. This is especially relevant with parenting.

OSTER: The kinds of data that are easiest to access when we have questions about parenting are data where you have two groups of families and one group engages in one parenting practice and another group engages in another parenting practice. For example, something I was writing about this morning is Montessori school. So you can think about how you would evaluate the quality of Montessori school by looking at parents who send their kids to Montessori school and comparing them to parents of kids who did not go to Montessori school.

Remember the “associations” issue that Emla Fitzsimons mentioned earlier? Here’s why it can cause problems in research.

OSTER With an analysis like that, you very rapidly run up against the problem that the kinds of families that are sending their kids to Montessori school are really different than the kinds of families who are not. They’re more likely to be white, they tend to live in higher income areas. There’s just a bunch of other things which are really difficult to separate from the impact of the kind of school. That problem is inherent in almost every one of these parenting questions that we ask, which means that much of the work in trying to help people learn what they can from that data is understanding where are the better studies and worse studies within a set of methods, all of which usually have some problems.

Breastfeeding is another example of a parenting question facing the inherent problem Emily describes. So let’s take a step back and try to understand why. Breastfeeding rates in the U.S. have ebbed and flowed. In 1972, following decades of decline, only around 22 percent of women breastfed their babies. This decline has been attributed to the aggressive manufacturing and marketing of infant formula, as well as to certain societal shifts, like more women entering the workforce full-time. That number has since steadily crept up, though, for a few reasons. Cultural changes led to more emphasis on natural approaches to birth and childrearing. Education efforts have let mothers know about the apparent health benefits of breastfeeding like decreased risk of asthma and obesity for babies, and decreased risk of breast cancer and high blood pressure for moms. Guidelines from professional groups have changed too. In the late 1990s, the American Academy of Pediatrics started recommending that women exclusively breastfeed their babies for at least six months. In 2019, more than 80 percent of infants started out receiving some breast milk — a massive spike from nearly fifty years earlier, when just 22 percent were breastfed. But even with so many more women choosing to breastfeed their babies, it’s hard to know if breastfeeding is causing improved health outcomes in children, or if the improvements are due to factors that researchers can’t see or control for. Still, women feel a lot of  pressure to breastfeed their babies.

OSTER: Probably the central question that people have is what are the benefits of breastfeeding and are the things that they are telling me out in the world about the long-term benefits of breastfeeding, are those really all that they’re cracked up to be?

After the break: what factors do women consider, when they consider whether to breastfeed? I’m Bapu Jena, and this is a special episode of Freakonomics, M.D.

*      *      *

The Brown University economist Emily Oster writes about using data to inform how we parent. I asked her what drives women’s decisions on breastfeeding.

OSTER: We know that there’s clearly very large sociodemographic differences in both desires and outcomes. So, women with more education, higher income women, tend to be more likely to want to breastfeed. They’re also more likely to be successful with breastfeeding, and that is likely partly because of the kinds of support that you get either in the hospital or after. It is because of maternity leave. You know, if you only have two weeks of maternity leave, that actually isn’t really enough time for most people to establish a good breastfeeding schedule that would be able to sustain time away from the baby.

JENA: What is your assessment of the prior literature on this? I mean, it’s been going back decades. What has it shown in terms of the effects of breastfeeding on health and cognitive development?

OSTER: When we look at sort of the better studies of this, I would say there’s two categories. There’s one large randomized control trial, run in Belarus in the 1990s, called the PROBIT Trial. And the PROBIT trial uses what’s called an encouragement design. They encourage the treatment group to breastfeed. They didn’t do as much encouragement with the control group. They have differences in breastfeeding rates at three months, at six months. And if you dig down into that data, you actually see some impacts on eczema and sort of allergy kind of reactions, on gastrointestinal illness in the first year. But you don’t really see anything in terms of long-term health, in terms of long-term cognitive development. So that is one piece of the better literature. And the other piece is sibling studies. So those are a bit easier to run. It’s not a randomized trial. You want to think about, there’s two siblings, same family. One is breastfed, one is not. Those kinds of studies tend to show quite limited impacts of breastfeeding. Again, maybe something on the gastrointestinal illness, maybe something on ear infections, but not these kind of long-term impacts on cognitive development or weight or height or the kinds of things that are often cited when people are told, “Breast is best, don’t you want to give your kid the best start?”

JENA: So maybe explain to me what the intuition is behind doing a sibling study. What are you trying to account for that would otherwise be difficult to do?

OSTER: Let me tell you about one of these kind of studies. So there’s a data set called the National Longitudinal Survey of Youth, which has a lot of kids in it, and it follows the starting point as women. And so then they sometimes have multiple kids. And so you can analyze the impacts of breastfeeding on cognitive development in that sample. So you can look at the IQ of kids and how it relates to whether they were breastfed. And so if you just compare the raw means, you see that the kids who are breastfed tend to do better on IQ tests than the kids who are not. You can then start by putting in some basic controls. So sort of some basic adjustments for maternal education and maternal income and race and so on. And that lowers those effects a lot. So the differences are much smaller, once you adjust for some of that. But you then worry, well, there’s a bunch of other stuff still left. For example, what about mom’s IQ? Okay. So then you can do a version of the analysis where you control for mom’s IQ. That makes the effects even smaller. But there are still potentially unmeasured differences between women that are both contributing to their decision to breastfeed and contributing to their kids’ IQ. The idea behind the sibling effect is to say, well, let’s just hold the mom constant. Let’s take the same mom, and then we’re going be able to hold constant everything about her, including the things that we don’t directly see in the data. It’s literally the same person. When you do that in that study, when you sort of control and compare two kids with the same mom, you don’t get any impacts on IQ.

JENA: How do those studies deal with the problem then that if a mom has two kids, the decision to breastfeed one child and the decision to not breastfeed the other child, that isn’t random, right?

OSTER: I think that people who run these studies would tell you, well, maybe there is an element of randomness. You know, it’s like, you’re busy or it’s harder to breastfeed the first kid. But I actually agree that with almost any sibling study you run, you’ve got to ask the question, well, like, this seems important. Why did you do it differently? And I think this kind of study’s not as good as a randomized trial, for sure.

JENA: Knowing the literature as you do, how much would you say that the literature should inform the decision of a mom to breastfeed or not?

OSTER: My general takeaway is that the effects are relatively small and as a result, a big part of this decision should be about, you know, what works for you, what works for your family.

JENA: What’s the pushback that you get about these views?

OSTER: I don’t get as much pushback about this as I thought that I would. I think that there is an increasing recognition in the world that some of the ways we have been pressuring people to breastfeed have been counterproductive and can actually be quite harmful, particularly to maternal mental health. You know, to the extent there is pushback, I think it’s absolutely around the idea of, well, what if there are benefits, or what if we’re missing something, or what if there could be some benefit for a small number of people, and we’ve kind of missed out on it, and so we should, you know, get everybody to do this. And I don’t find that surprising. I don’t agree with it.

JENA: How valuable do you think it would be to run a randomized trial to help answer some of these breastfeeding questions?

OSTER: I think it would be very valuable.

The British economist Emla Fitzsimons didn’t exactly run a randomized trial — but she came close. After the break, she’ll share the findings from her new study — and tell us who stands to benefit most.

FITZSIMONS: I think it speaks to the importance of the early years in terms of their influence on later outcomes. Absolutely vital.

I’m Bapu Jena, and this is Freakonomics, M.D.

*      *      *

FITZSIMONS: I’ve long been interested in studying different aspects of early parenting behaviors and the early childhood environment. And as an economist, I’m particularly interested in measuring causal impacts, so that we can think of the best possible policy levers to help improve people’s well-being. 

That’s Emla Fitzsimons again, from University College London. As she told us earlier, she was drawn to study breastfeeding when she gave birth to her first child on a weekend and noticed she received less breastfeeding support as a result.

FITZSIMONS: I was aware that there was evidence out there showing associations between breastfeeding and children’s outcomes. I was aware that it was perceived as a positive thing to do. But I wasn’t overly familiar with the evidence. I really, really wanted to understand what we really knew and what was actually causal versus more sort of observational, association-type studies. 

Emla also happens to lead the Millennium Cohort Study, which follows over 19,000 children born in the U.K. at the turn of the millennium. Using this group and its data — combined with her own experience — Emla tried to answer one question about breastfeeding, and then another.

FITZSIMONS: We know more disadvantaged mothers are less likely to breastfeed. That was an important motivation behind our research. So first of all, we examine whether timing of birth affects breastfeeding. And in particular, whether babies born at the weekend are less likely to be breastfed compared to those born during the week. And for a certain group of mothers, those who left school before age 17, we find that it does. So to give you an example, a baby born on a Saturday is six percentage points less likely to be breastfed compared to a baby born on a Monday. So 27 versus 21 percent. And we find evidence in our study to suggest that it’s because breastfeeding support is less available in the hospital at weekends. So that’s the first aspect. And then secondly, it looks at the impact of breastfeeding for at least three months on children’s cognitive development and health and noncognitive outcomes for this group of mothers. It finds benefits for cognitive development lasting at least to age five, and it doesn’t find effect on health or non-cognitive outcomes.

JENA: What is it exactly about the lactation support that is different on the weekend in U.K. hospitals versus the weekdays?

FITZSIMONS: It’s just less available. So, in the U.K., midwives and nurses in maternity wards generally, as part of their regular duties, provide lactation support in the very early postpartum period. Weekend working hours are more expensive for the National Health Service. So they tend to have lower capacity at weekends compared to during the weekday to sort of focus on the non-core aspects of maternity care. So they’re more focused on the aspects around delivery and ensuring that everything runs smoothly. And they have less time really available for other non-essential services or what are considered non-essentials, such as providing lactation support to new moms.

JENA: A critical assumption that you make here is that if you show that moms who deliver on the weekend are less likely to breastfeed than moms who deliver on the weekdays, that is due to differences in lactation services as opposed to differences in the characteristics of the moms or other services that the hospital may provide, for the weekend versus weekdays. Can you explain to me what you did to make sure that the moms were essentially randomized or quasi-randomized to these two different levels of lactation support?

FITZSIMONS: We focused on a sample of low-risk, natural deliveries. That was the first thing. So we excluded elective Cesareans, which may be timed in a particular way. In the U.K., at the time of our study, around 98 percent of births were through the public health system, which I know is quite different to the U.S., and so there was no choice, practically no choice, about when a woman gives birth, so it’s a spontaneous act. The data set we use is extremely rich in lots and lots of background characteristics. So, we could then show that if you compare the women who gave birth at the weekend versus weekdays along many, many dimensions, they were absolutely comparable. We didn’t observe any significant differences along lots and lots of observable characteristics that we looked at. So that was one really important aspect. We also observe some information about the birth, about the type of care they received in hospital, about their experience of breastfeeding support. So we use different sources of data, again, to show that when you compare the births, at the weekend and during the week, with the exception of breastfeeding support, we didn’t observe any differences in the core services that were being provided. So the type of delivery, the type of pain relief being used, and so on. That was a really important dimension to show as well. But when we looked at people’s experience of breastfeeding support, we found that they were significantly more likely to report that they were less satisfied with the support they received at weekends. They had less access to midwives to help them establish early breastfeeding. So, I suppose, together, this is all really, really important, like you say, Bapu, to show that essentially this is pretty much comparing two groups who are absolutely comparable. The only difference is just the sort of lottery of birth due to that one happens to give birth during the week and the other group happens to give birth during the weekend. And so we can then go on to compare their children and compare their development over time and attribute any differences in their development to breastfeeding.

JENA: You focused on economically disadvantaged mothers. Was there a particular reason that you looked at that group?

FITZSIMONS: We needed some variation in breastfeeding that was due to exogenous or random reasons. So when we studied that first stage, whether timing of birth affects breastfeeding, we found that it did affect breastfeeding, but only for the sample of relatively low-educated moms. So, a baby born on a Saturday versus a Monday was less likely to be breastfed. When we looked at that for high-educated moms, we didn’t find any difference in breastfeeding by day of the week.

JENA: I see. And let me just make sort of a statistical point. It’s nuanced, but I think it’s important. In your design, which is sort of this natural experiment, you’re relying on something that is plausibly random with respect to the timing of birth, which is the availability of a lactation consultant. But for that to, allow you to study the question that you want to study, it has to be the case that that service affects the likelihood of women to breastfeed or try to breastfeed. And maybe for women who are already planning on doing that, maybe economically advantaged moms, doesn’t really have an effect. But there are some women who are at the margin, for whom that sort of access would lead them to do something that they otherwise might not do. And then in this case, that’s breastfeed or try to breastfeed their kids. Is that the right way to think about it?

FITZSIMONS: I think that’s right, and because they’ve received this extra support, they’re then able to breastfeed compared to had they not received that support.

JENA: And what were the outcomes that you studied specifically, and how large were the effects?

FITZSIMONS: We looked at children’s cognitive development and mainly focused on expressive language, nonverbal reasoning, and problem-solving. We looked at a range of maternal reported measures of health of the child, including infections, skin problems, and persistent vomiting and diarrhea. And also we looked at measures of obesity as well in childhood. And then we looked at maternal-reported measures of children’s socio-emotional and behavioral development. And we looked at all these measures up to age seven. We found that all of the effect appears concentrated on children’s cognition, and we observed strong effects lasting at least to age five. In terms of the size of the effect, we found the effect to be mostly concentrated, I would say, on expressive language. We found that breastfeeding for at least 90 days versus less than that would shift a child up 15 percentile points in that distribution, so the effect was large. By age seven, it appears that the effects are a little smaller and possibly fading out, but we can’t say for sure and we need to study this further to really understand the longer-term outcomes. And I would say this is an absolutely vital and open question. But luckily for this study we now have data on educational attainment up to age 16, which we’re now planning to look at next.

JENA: So, it felt to me that the effect was large. Do you think it’s too large? Does it make sense, sort of the magnitude of the effect?

FITZSIMONS: I mean, is it too large? It is a statistically significant effect. We’re not picking up any effect on health or on non-cognitive outcomes. Of course, absence of evidence is not evidence of absence. But I think the fact that it’s concentrated on cognition is an important finding. But this is one study, using a very robust and new methodology to study this, and hopefully we’ll see more studies.

JENA: I’d like to understand a little bit more about the mechanism here. What I would like to understand is whether or not it’s the breast milk, I mean literally the breastfeeding that’s occurring, versus the mother-child interaction, or both. And is there a way to separate those things?

FITZSIMONS: Yes. Well, in our study we can’t look at the extent to which it’s the actual composition of the breast milk, which is one mechanism that has been put forward in the medical literature. But the other mechanism that you mentioned, Bapu, and that we do look at in the paper, is around maternal attachment and whether there’s any evidence that moms who breastfeed versus those who don’t are displaying higher levels of maternal attachment. We can look at that using maternal-reported measures of how close they feel to their children, and when we do that, we find absolutely no evidence that maternal attachment is any better or worse depending on how a baby is fed.

JENA: Does your research say anything about the role of breastfeeding and, and maybe more specifically, breastfeeding support as a way to address inequality and, and perhaps cognitive outcomes of children?

FITZSIMONS: I suppose the fact that we’re observing potentially large effects on cognition for a particular group of disadvantaged mothers speaks to that directly. And we do find evidence suggesting it may be contributing to persistence in intergenerational transmission in inequalities. I think that makes the case all the more compelling that it is a group of disadvantaged moms that it seems to be having an impact on, and that speaks, to me, of the importance of providing policy support all the more.

Emla’s work was published in July of this year in the American Economic Journal. I wanted to cover it on the show because it’s one of the most creative studies I’ve seen that tries to isolate the causal effect of breastfeeding on children’s health and cognitive outcomes. It can be hard to find a way to bring rigor to questions about private choices like breastfeeding. Emla’s study isn’t the final word, but it gets us closer to some clearer answers about breastfeeding and its potential to address larger issues, like equity. Ultimately, though, regardless of these or any findings, we need to be supporting moms more if they want to breastfeed, for any reason. Here’s Emily Oster again.

OSTER: One of the real reasons to encourage breastfeeding is because people might like it, and because it might work for your family. That is such a strong argument for helping everyone have access to this if they want it. Part of what’s so, so tricky about this is I simultaneously think that we should dial down the intensity of the messaging about how you have to breastfeed, because otherwise you’re a bad parent. And we also should way dial up how supportive we are for people who want to try this, who want to do this.

That’s it for today’s show. I’d like to thank my guests Emily Oster and Emla Fitzsimons. And here’s an idea to leave you with based on today’s discussion. You may have heard that earlier this year there was a large shortage of baby formula in the U.S. This got me thinking about moms who might’ve decided to breastfeed or continue breastfeeding because formula was, all of a sudden, less available. Could that be another natural experiment to study the impact of breastfeeding?

*      *      *

Hey there, it’s Stephen Dubner again, and that was Bapu Jena, the host of Freakonomics, M.D. If you liked this episode, there are a lot more waiting for you.  Just follow or subscribe to Freakonomics, M.D. in your favorite podcast app. And we’d love to know what you thought of this episode. Send an email to

Freakonomics Radio and Freakonomics, M.D. are produced by Stitcher and Renbud Radio. This episode was produced by Julie Kanfer. Our staff also includes Neal CarruthGabriel Roth, Greg Rippin, Rebecca Lee DouglasZack Lapinski, Morgan Levey, Ryan Kelley Katherine Moncure, Eleanor Osborne, Jasmin Klinger, Jeremy Johnston, Daria Klenert, Emma Tyrrell, Lyric BowditchAlina Kulman, and Elsa Hernandez. Our theme song is “Mr. Fortune,” by the Hitchhikers; the rest of the music this week was composed by Luis Guerra. You can follow Freakonomics Radio on Apple PodcastsSpotifyStitcher, or wherever you get your podcasts.

Read full Transcript




Episode Video