Bryan Caplan  

Breast Milk, Twins, and Outcomes: A Difference-in-Differences Approach

PRINT
Bob Murphy's Critique of Bill ... Baiting Garett Jones...
Does breast-feeding really give your kids a leg up in life?  It's an important question, and there's a lot of research on it.  But most of the research is, at best, moderately convincing. 

The key weakness: If parents falsely believe that breast-feeding is good for kids, then the (genetically and environmentally) "best" parents will be more likely to breast feed.  When their kids turn out well, breast milk will undeservedly get the credit. 

Sure, you can control for parents' genetic and environmental quality.  When you do, the effect of breast-feeding on e.g. IQ shrinks but doesn't disappear. But you always have to worry that the lingering correlation reflects unmeasured differences in parental quality.

I've figured out a more convincing way to measure the effect of breast-feeding, using what economists call a difference-in-differences approach.  The starting premise: Breast-fed twins get less breast milk because they have to share it.  They probably get more than 50% of a normal dose, because milk production responds to nursing.  But when nursing time doubles, milk production less than doubles.  As a result, moms who want to breast-feed their twins are more likely to supplement with formula. 

Given this fact, it's tempting to compare the outcomes of twins and non-twins, then chalk up the difference to breast milk.  But that's no good.  Different doses of breast milk are just one of many differences between twins and non-twins.  (For starters, twins have lower birth weights).  I propose a more sophisticated approach:

1. Measure the differences in outcomes between breast-fed twins and breast-fed non-twins.

2. Measure the differences in outcomes between bottle-fed twins and bottle-fed non-twins.

3. See if (1)>(2), and if so, by how much.  The key intuition: If breast milk matters, the outcome gap between breast-fed (twins and non-twins) should exceed the gap between bottle-fed (twins and non-twins).

Of course, this only measures the marginal effect of breast milk, not the total effect.  Perhaps a small dose of breast milk is all a baby needs to realize all the benefits.  But to the best of my knowledge, any study along these lines would be a big advance over previous research. 

P.S. Any researcher who wants to steal my idea is welcome to do so.


Comments and Sharing





COMMENTS (7 to date)
Bob Murphy writes:

Bryan wrote: Sure, you can control for parents' genetic and environmental quality. When you do, the effect of breast-feeding on e.g. IQ shrinks but doesn't disappear.

I know I say this every time Bryan posts on this stuff, but hey someone has to do it: This is yet another case where Bryan admits PARENTING DECISIONS AFFECT HOW KIDS TURN OUT.

If Bryan merely said, "Parenting decisions don't have nearly the influence that most parents think," OK fine I would give him a high-five. But that's not at all how Bryan sums up the literature.

I blame his mom.

Alex Godofsky writes:

Of course, this only measures the marginal effect of breast milk, not the total effect.

Don't you mean to say it the other way around?

Rachel writes:

It's a tempting idea in theory. But it wouldn't work in practice. The only rigorous evidence (randomized trials) for breastfeeding is premature babies. They seem to do much worse on formula. Twins are much more likely to be premature or have other health problems. Therefore, the effect of breastfeeding is probably higher.

If medical ethics weren't so complicated, it would be really easy to test breastfeeding. Just give half the mothers expensive breast pumps and half the mothers nothing. Then follow the kids. But that's somehow forbidden.

A more serious methodological problem with measuring "breastfeeding" is that formula has gotten a lot better since the 1920's. So, any pre-existing studies comparing breastfed babies to formula fed babies are basically worthless. Furthermore, the pressure to breastfeed means less attention is devoted to improving formula. Maybe we only need to tweak formula a little bit and there will be no difference at all.

Piotr Pieniążek writes:

@ Alex,
I suppose it's correct since the scheme of the research is to measure how the difference between twins and non-twins in the quantity of breastmilk given (say 300 ml to 400 ml = 100 ml at margin) matters.

JayMan writes:

@Bob Murphy

I know I say this every time Bryan posts on this stuff, but hey someone has to do it: This is yet another case where Bryan admits PARENTING DECISIONS AFFECT HOW KIDS TURN OUT.

No, it doesn't, because you're missing a key point: parents who breastfeed may be (and almost certainly are) systematically different from parents who don't breastfeed. As such, any difference in the outcomes of their children could be entirely due to heredity.

If smart, health conscious parents are the ones most likely to heed advice to breastfeed, then their children will be on average smarter and physically healthier. This is a general weakness of all parenting studies.

Arin D writes:

I have a different suggestion ... Something I have thought to do myself but never got around to it. Use the presence of inverted nipples as an instrument for breast feeding ... Moms with inverted nipples have much harder time breast feeding but this is uncorrected with any other biological or sociological differences that may violate the exclusion restriction.

mobile writes:

Bryan has twins, right? He already missed a golden opportunity to contribute to the social science community when he refused to give one of them up for adoption to a family with a very different environmental. It's a shame he also missed this much less onerous silver opportunity to breast feed exactly one of his twins.

Comments for this entry have been closed
Return to top