What this covers

Emily Oster, an economist at Brown University, discusses her book Cribsheet with Russ Roberts in a conversation that maps the actual evidence underlying major parenting decisions. Oster uses the tools of empirical economics to examine common advice on breastfeeding, sleep arrangements, discipline, vaccination, and food allergies—areas where parents face urgent choices but rarely know which research is trustworthy. The core insight is methodological: many widely accepted claims rest on observational studies that confound parental traits with parenting effects, so that randomized trials and sibling studies consistently reveal smaller real impacts than popular messaging suggests.

The conversation ranges across the full landscape of parenting science and its blind spots. On breastfeeding, randomized control trials show genuine but modest benefits—early reductions in gastrointestinal illness—not the transformative effects often claimed; the gap arises because mothers who breastfeed differ systematically in education, income, and family stability, factors that independently drive child outcomes. By contrast, early peanut exposure emerges as a rare case where the evidence is decisive, nearly eliminating allergy risk compared to avoidance. On sleep and co-sleeping, Oster calibrates actual risk against daily hazards and emphasizes parental wellbeing as a legitimate consideration. The conversation also covers why minor choices—formula versus breast, co-sleeping or separate rooms, timing of potty training—likely matter far less individually than parents fear, and why framing decisions as tradeoffs between child outcomes and family happiness, rather than seeking the one correct answer, reflects both the evidence and parental realities. Vaccination receives direct treatment: Oster defends the safety record while arguing that engaging skeptics with actual evidence, rather than dismissing them, proves more effective.

Sharpest takeaway

Oster argues that much popular parenting advice (on breastfeeding, co-sleeping, discipline) rests on weak observational evidence confounded by family characteristics, so parents should weigh the modest real effects against their own family's preferences rather than treating each choice as decisive for the child's outcome.

  • Observational studies comparing breastfed and non-breastfed kids are biased because breastfeeding mothers differ systematically; randomized and sibling studies show much smaller effects
  • Most minor parenting choices don't individually determine a child's success, so decisions should account for parental wellbeing
  • Where evidence IS decisive (vaccines, early peanut exposure) it rests on large, carefully constructed studies and should guide action

The argument · threads6 threads · 31 claims
0.75

Randomized trials reveal many claimed parenting benefits are small or nonexistent when confounding is properly controlled.

4 pointscentrality 5/5
  • The original peanut finding—comparing allergy rates between kids in Israel and Jewish kids in the UK—was a cross-country observational comparison, the kind of weak evidence economists distrust, and only became compelling when followed by a randomized control trial.

    “cross-country studies--it's like, you know, this is like we are in the 1980s. You know, we're not into that. But then they actually went a step further and they ran a randomized control trial”

  • Observational studies comparing breastfed to non-breastfed children are heavily biased because mothers who breastfeed differ systematically—more educated, richer, more stable family structures (in the US)—and these traits independently drive better child outcomes, so simple comparisons confound the effect of breastfeeding with the effect of family characteristics.

    “moms who breastfeed tend to be different from those who don't, and in particular they tend to more educated, they tend to be richer, they tend to have more stable family structures”

  • When you focus on randomized control trials (the Belarus study) and sibling studies (which hold the mother constant), breastfeeding's benefits are much more muted than public claims—statistically significant but small early-life reductions in gastrointestinal disease and rashes, with no support for claimed effects on IQ, obesity, eczema, or long-run allergies.

    “you find much, much, much more muted benefits than the sort of grandiose claims that are made in a lot of the public discussion of breastfeeding”

  • A randomized control trial showed that exposing infants to peanuts early dramatically reduces peanut allergy risk—about a third of high-risk unexposed children developed an allergy versus roughly 10% of early-exposed children—overturning prior advice to avoid peanuts, which likely made the allergy epidemic worse; similar early-exposure benefits now appear for other allergens like wheat, eggs, and milk.

    “The kids who did not get exposed, but a third of them ended up with a peanut allergy. Versus 10% in the group that had early exposure. So, that's like a huge effect.”

0.70

Parents must decide major issues with incomplete data; the same evidence rationally supports different family choices.

3 pointscentrality 5/5
  • Many parenting choices are binary and unavoidable (e.g., baby in your bed or not), yet you will never have unlimited perfect data to evaluate their full consequences, so you must make the choice with limited information—as is true of much of parenting and life.

    “You have to make these choices with limited information... you are not going to have unlimited amounts of perfect data to evaluate the full consequences of that. But you still need to make the choice.”

  • People often try to convince others that their parenting choice is right for everyone, but the same data can rationally lead different families to different choices—something can be the right choice for you and not the right choice for someone else—so we should resist universalizing personal decisions.

    “Something can be the right choice for you and not the right choice for other people. And so, we can all see the same data and make different choices, and they can all be the correct choice.”

  • Once a child's broad circumstances are set, minor parenting choices—breastfeeding vs formula, co-sleeping or not, potty-training timing, preschool choice—likely do not individually matter much for outcomes; what matters is that parents make these choices in a structured way they feel good about, since being a happy, non-guilty parent is itself valuable and probably makes you a better parent.

    “I can more easily imagine that those kind of choices do not individually matter very much... most of these choices are not the most important thing that you will do”

0.66

Parental wellbeing and realistic risk assessment matter; intense pressure on mothers and catastrophizing over low-probability events harm families.

5 pointscentrality 4/5
  • Many parenting choices involve only small costs or benefits to the child, so decisions should be structured to take those seriously while also weighing the preferences and wellbeing of the parents and family—rather than investing in the child at infinite cost to the family, since a miserable parent is not the best parent.

    “there is so much emphasis put on investing in the kid, and thinking about what is right for the kid at infinite cost, sometimes it seems like, to the family”

  • A pediatrician's advice—'I'd probably try just not to think about that'—captures that you cannot obsess over every unlikely catastrophe (e.g., a bee sting); since it will probably be fine, anxious parents must learn to let go of low-probability worries.

    “she said, 'Hmm. I'd probably try just to not to think about that.'”

  • The assumption that breastfeeding is easy combined with messaging that it is 'the best start' creates intense pressure on new mothers, so that women who struggle—which is common, especially first-timers—feel they are failing their baby, when in fact no single early choice determines whether a child becomes a successful adult.

    “It's the first thing I can do for my baby. And I'm failing them. I'm just a huge failure.”

  • Co-sleeping carries SIDS/suffocation risk, but the danger is concentrated in cases where parents smoke or drink heavily; for parents who don't smoke or drink and remove extra pillows, the increase in death rates is very small—small relative to everyday risks like car travel—so it is a reasonable choice either way provided one understands and mitigates the actual risks.

    “the evidence suggests that that combination--that sort of drinking and smoking and co-sleeping together is a very dangerous combination”

  • The intense emotionality and crying in the first few days after birth ('baby blues,' driven by hormones) is near-universal and not cause for worry, but should be distinguished from postpartum depression, which is marked by persistence and, notably, lack of interest in the baby; depression is hard to recognize from the inside, so partners learning to spot the signs can be valuable.

    “as, you know, if that sort of continues and people don't have interest in the baby, in particular, that's, then we're sort of moving into post-partum depression”

0.56

Parental circumstances profoundly shape child outcomes; universal appeals to ancestral practice or easy dismissal ignore evidence.

3 pointscentrality 4/5
  • The claim that parenting and circumstances don't matter for outcomes contradicts a huge body of social science on the importance of circumstance; persistent patterns of inequality in the United States make it implausible that being born into very different circumstances has no effect on outcomes.

    “the claim that, sort of, parenting and the king of circumstances that people are raised in are not important for their outcomes belies a huge amount of social science about the importance of circumstance to outcomes”

  • The attachment-parenting appeal to how things were 'always done' in ancestral cave-dwelling life is of limited relevance because modern society and the skills children need differ greatly from those conditions, and it's unclear what specific outcome the argument is even meant to evaluate.

    “there are a lot of things that are different about modern society than when people were living in one cave together... the kind of skills that people need are different from living in caves”

  • It is a mistake for the pro-vaccination community to dismiss vaccine-hesitant people as crazy lunatics; engaging seriously with the actual evidence (as in the IOM report) is more persuasive than appealing to authority, because the anti-vaccination side can otherwise appear more evidence-based by citing papers.

    “it is a mistake to just dismiss people who don't believe in vaccines, as crazy lunatics. Which I think is sometimes the attitude of the pro-vaccination community.”

0.51

Specific parenting practices work best when executed consistently and credibly; discipline evidence converges on core principles.

2 pointscentrality 4/5
  • A core discipline principle is to never make a threat you are not willing to follow through on (e.g., 'I'll leave you on the airplane'); children detect empty threats, so consequences must be credible and the parent must be willing to act on them.

    “'If you keep doing that, I'm going to leave you on the airplane.' That is like a classic example of a threat that you shouldn't make. Of course you are not going to leave the kid on the airplane.”

  • Randomized evaluations of discipline programs (1-2-3 Magic, Positive Parenting) converge on the same principles: be very consistent, don't get angry, don't negotiate or explain (children are not adults reasoning the way you assume), and apply a structured set of consequences like warnings followed by time-outs.

    “the main things they are emphasizing are consistency and not getting mad--that adults should try not to get mad in the moment”

0.44

Current evidence on major parenting decisions is weak; stronger randomized research would better inform families.

6 pointscentrality 2/5
  • It is surprising that there is only one major randomized control trial on breastfeeding, since running another—particularly in the US where breastfeeding rates are not very high—does not seem especially difficult, raising the open question of why our evidence on important parenting decisions isn't better.

    “this actually doesn't seem that difficult, to run another trial... And I'm always sort of curious why people haven't pursued that. So, why isn't our data better than it is? I don't know.”

  • Parents asked whether they are happy in the moment with small children often are not—kids are tiring and frustrating—yet when asked about the overall experience of having children, most say it is the best thing they do, illustrating that momentary and retrospective happiness measures diverge sharply.

    “in the moment of being with their kids they are not necessarily that happy... But, then when you ask people about sort of the experience of having kids, most people will tell you, like... 'This is the best thing that I do.'”

  • Reading to children appears beneficial—evidence from birth-spacing (more concentrated parenting time leads to more reading and better reading readiness) and small FMRI studies showing more imagery-related brain activation in frequently-read-to children—but the data are heavily confounded by parental education and income, so the cleanest case for reading is that the parent enjoys it.

    “reading to your kids is really heavily associated with other kinds of features of parents like education or income”

  • Vaccines are safe and effective: a roughly 980-page Institute of Medicine report reviewing thousands of studies finds that identified risks are either extremely rare (e.g., only for immune-compromised children, in a handful of episodes) or minor (e.g., fever after MMR), and there is abundant evidence that vaccines do not cause autism—the original link came from a single discredited, retracted paper.

    “vaccines are safe and effective”

  • This parenting work has shifted Oster's academic approach toward more descriptive research ('here's an interesting relationship we could learn from') rather than strong causal policy prescriptions ('do x and y will happen'), even though the prescriptive style is more successful at attracting attention.

    “my work, I think, has become more sort of saying, 'Look, here is an interesting relationship in the data that I think we could learn something from.'”

  • Vaccination rates have fallen so low in some areas that there are real risks of outbreaks, and a large measles outbreak is currently occurring because not enough people have vaccinated their children.

    “the vaccination rates are so low in some areas that there are all real risks of outbreaks, and we are having, currently, in the middle of quite a large measles outbreak. Because people haven't vaccinated.”