Emily Fair Oster is an American economist and author. She is currently the JJE Goldman Sachs University Professor of Economics and International and Public Affairs at Brown University, where she has taught since 2015. Her research interests span from development economics and health economics to research design and experimental methodology. Her research has received exposure among non-economists through The Wall Street Journal, the book SuperFreakonomics, and her 2007 TED Talk.Oster is the author of three books, Expecting Better, The Family Firm, and Cribsheet, which discuss a data-driven approach to decision-making in pregnancy and parenting.
Richie helps individuals and organizations get better at using data and AI. He's been a data scientist since before it was called data science, and has written two books and created many DataCamp courses on the subject. He is a host of the DataFramed podcast, and runs DataCamp's webinar program.
So one of the things I sometimes say is data is not your boss. Data is not bossy. Data is not going to tell you what to do. Data is an input, and you have to put on top of that input the values or preferences or constraints that you are gonna face individually. And for that reason, in almost all these places, there is no right decision. Very rarely, would the data boss you to a sort of particular decision. It is much more frequent that there is a right decision for you—but that's data plus preferences, values, and constraints.
The main distinguishing feature between the easier things to answer and the harder things to answer is often the quality of the data. So take something like smoking. A lot of the worries with smoking during pregnancy is an impact on birth weight. And so you can actually do randomized trials on smoking in pregnancy. It's not a randomized trial where you encourage people to smoke, obviously, but you can do randomized trials of smoking cessation. So they have trials where they more strongly encourage one group than another to quit smoking. And it turns out when you do that, you can get people to quit smoking to some extent, and you can see that it translates to large increases in birth weight among the babies of moms who were less likely to smoke. So you sort of have something where we can be more confident that there's a causal relationship because the experiment is randomized. The places that I think are most challenging are those where we don't have data like that and where there's kind of reasons to believe sort of, there are reasons that either story is compelling. So alcohol is an example where we know that heavy drinking, drinking a lot during pregnancy, that binge drinking during pregnancy, is associated for some children with negative consequences. So there's an increase in behavioral problems, a condition called fetal alcohol syndrome that can occur with very heavy drinking. And so we know that. But on the other side, there's an enormous amount of data that is just correlations. It's not randomized data, but we have large data sets where we compare people who drink to people who don't. And there's a huge amount of variation in the quality of that data. So a lot of the work of that part of Expecting Better is to dive into that data to say, okay, we actually wanna ask the question, what's the impact of occasional drinking on kids? Is it like what we see with heavy drinking and there are some reasons to believe this might be true biologically, is it the case that drinking at lower levels would not show the same kind of impacts? Learning that out of the data is tough because you have to figure out what are the studies that do a good job of isolating moderate drinking, isolate occasional drinking, and some studies do a better job of that than others. You worry that because it's not randomized, the kinds of people who drink are different from the kinds of people who don't. Some studies do a better job with that than others. This is the thing I am trained in. So I think part of what I am bringing to the table, part of the perspective and sort of expertise that I'm bringing to the table, is my training is in like how do we learn causal relationships out of observational data and how do we evaluate the differences in quality in these different kinds of studies.
Emily's advice for new parents wanting to incorporate data-driven decisions into their parenting is to trust themselves. Believe that you're smarter than you think, and your ability to understand these parenting problems is greater than you might imagine. You should feel empowered to engage with parenting questions and make choices that work for your family.
To be more efficient when finding and exploring data and studies from academic sources, learn to understand and read appendix tables. Using appendix tables will allow you to be able to figure out where the common ground on the topic is among the community, allowing you to figure out who's being cited and whether what you're reading is a good source of information.
When considering the skills needed to go out and make your own data-driven parenting decisions, try to have some basic training in statistics and causal analysis. Try to make sure you're comfortable with research design, research design methodology and understanding where data comes from. Ensure you're in a place where you're asking questions like 'What is the source of this information?', 'What's the source of variation?' and 'How do we think about causality?'
Richie Cotton: Welcome to DataFramed. This is Richie. Before diving into today's episode, I wanted to let you know that DataCamp is hosting a virtual conference on generative AI on June 22nd. It's called Radar ai and it's free to attend. We'll be talking about all the new AI startups, how to gain AI literacy, how to use AI for more productive data science, how to safely use AI in the enterprise, and how to use AI to upskill your workforce.
It's gonna be a lot of fun. Register here. With that out the way, we've been talking about AI a lot, so today we're taking a break and discussing something else entirely for Mother's Day in the USA. Let's talk about parenting data instead. Parenting involves a lot of decision making from whether it's okay to drink alcohol while you're pregnant, to when to wean your baby onto solid food, to how to juggle your time between working and parenting.
You might ask for help with these decisions from friends and family or the internet, but it turns out there's an awful lot of research into parenting that can help. Today's guest is Emily Oster. She's a Professor of Economics at Brown University. A research associate at the National Bureau of Economic Research, and in 2022, Emily is on Time's list of the 100 most influential people.
How does parenting fit into this? You may ask. Well... See more
Richie Cotton: Hi Emily. Thank you for joining us on the show.
Emily Oster: you so much for having me.
Richie Cotton: Just to begin with so you're an economics professor and I'm curious as to how you went from economics professor to writing books about parenting.
Emily Oster: So the short answer is I got pregnant and became apparent, and I found in that experience, I did a tremendous amount of work that was. Really informed by my training as an economist and in particular my training in data analysis but was in the service of my pregnancy and my parenting and the books, the newsletter, all of the sort of public facing stuff I do really comes out of that experience.
And that feeling that other people could use these tools to make better choices in their own in their own lives.
Richie Cotton: That's pretty nice alignment when something you doing at work feeds into something in your personal life, it all comes together. So, you've got a series. Of three books, so can you talk me through what's in each of the books?
Emily Oster: Yeah, absolutely. So, so Expecting Better is my first book. So I wrote this book when I was pregnant with my daughter and it really was a kind of, it like, it's like a labor of love in some way. Like I, I got pregnant and I started having these experiences in which I. I felt like I wasn't. Getting the data and evidence I needed to make the kinds of choices that I knew I had to make.
And so I spent just a huge amount of time trying to analyze the data behind many of these common pregnancy restrictions and pregnancy rules and pregnancy decisions. And so the first book is really that effort. So it's kind of walking through what does the data say about sushi in pregnancy, about prenatal testing, about, epidurals, about different birth choices.
And so it's about pregnancy. My second book is called Crib Sheet. It's about early parenting, so it takes really very much the same frame of kind of database decision making, but two questions like. Circumcision and breastfeeding and potty training and kind of things that come up in, say, the first year, three years of your kids' life.
and like expecting better. It's very like data forward. And the third book is called The Family Firm. And it's about older kids and it has some of the same feel. So there are some data about sleep and nutrition and. School and some of the stuff that comes up a lot when we're dealing with elementary school aged kids.
But that book also has what I like to think of as like a decision tool wrapper. Trying to help people use not only data but really structure decision making, deliberate decision making to try to think carefully about how they want their family lives to look. So I think those tools are pretty relevant even in other stages of life.
They become, Somewhat more relevant as your kids get older and logistics become more front and center. So, that book is about the decision making piece and crafting the family life that you want, but also about, what does the data say about summer camp?
Richie Cotton: Absolutely. So I have to say I've read all three and they're really fun reads and really informative I found. so maybe we'll start with talking about expecting better. And I felt like one of the recurring themes in that book was that you got given a lot of rules to follow by doctors, and you're like, well, why am I following these rules?
I have no idea. So can you maybe talk about how working with data and economic thinking helped you with this process and how to deal with the rules you were given?
Emily Oster: Yeah, so I think one way to summarize the experience and the frustration of this experience was, the first time I had a like serious prenatal visit, I was like maybe 10 weeks pregnant, and I left the visit. And they gave me this list. They ran after me and they were like, oh, we forgot to give you this piece of paper.
And the piece of paper was like, here are all the things not to eat. And it was just incredibly long. It's like hotdogs and deli meat, deli, Turkey and like different kinds of fish and coffee and alcohol and cigarettes and you know, moving up from there. But there was no effort at that point to explain either a y any of these things.
Were restricted and B like, which were worse. Is deli Turkey as bad as smoking? Like not quite as bad as smoking. If I had to pick one, would it be Turkey? And so the, a lot of the book is about then is really about trying to dive into those restrictions and try to answer the question, what can we see from data or from facts about the world that will help us understand why is this thing restricted?
Because that's really then the avenue into thinking about. do I want to change my behavior in this way? And that's one piece of it. And then there's a second piece, which is, you know, the other thing that happened in this very early on for me was the need to make some decisions about about prenatal testing, which was, is like a quite a fraught topic on which really one does have to choose.
And There were real potential consequences in either direction. And yet there I was told, well we have a rule. People who are over 35 do this and people are under 35, do that. And I knew because I'm a person who works with Dayden and like I.
I just knew that there was basically nothing in the world that works like it turns on at 35 and not at 34. I mean, that's just not the way data is, and yet without some framework for trying to understand this, it was really difficult to think about how should I make this choice? What is actually the right way to make this choice?
All I know is the way they're telling me to make the choices wrong. So again, in the book, working through how can you frame that decision, how can you look at the right data to make that choice? So it is the choice that works for you.
Richie Cotton: I think it's been quite comforting to everyone who's just about to turn 35. We think, oh no, the world can run. It's actually doesn't make a difference. It's one day. so, A related thing is you're not just using data, there's a mix of like you start with the data and what's the evidence, and then there's personal decision or personal preferences built into whatever decision you make.
Can you just talk a bit about how those two things reconcile?
Emily Oster: In almost all cases. There has to be some personal preference playing a role in the choices that you make. So I like an example is like an epidural. So an epidural is a form of pain relief in labor. And one starting place for thinking about, do I want an epidural? I. Is, what is the kind of data on the benefits or the downsides, what are the risks of this?
And you can go through the data, and I do in the book, talk about potential differences in recovery or, length of labor, which is the sort of biggest potential trade off there. So you can look at all of that, but at the end of the day, the observation that, you know, having an epidural lengthens the first stage of labor, which is like before the pushing that lengthens the first stage of labor by an average of, 12 minutes or whatever number it is, that's actually not a decision, right?
Because people will view that differently because of course there's this other thing on the other side, which is labor's really painful and you might not want to experience that. And so ultimately, like all other examples, what you can do there is you can look at that data and you can say, okay, let me combine that data with my preferences.
And that's preferences about. Pain potentially, but also preferences about what kind of birth do I wanna have? What is the experience that I imagine, you people are gonna have different views about that. So one of the things I sometimes say is data's not your boss. Data's not bossy, data's not gonna like tell you what to do.
Data is an input and you have to put on top of that input the values or preferences or constraints that you are gonna face individually. And for that reason, in almost all these places, there is no right decision. Very rare the data would boss you to a sort of particular decision. It is much more frequent that there is a right decision for you, but that's data plus preferences, values, constraints.
Richie Cotton: Absolutely. And it seems like maybe sometimes a lot of the. Arguments around parenting seem to be more about preferences than about data. Have you experienced that?
Emily Oster: I have experienced that. I think part of what it is, is that as parents, we really want, we wanna do the best thing for our kids. I mean, I think this sort of comes from a good place, which is like, people want, they wanna do what is best for their kids. They want their choices to be right. And I think that they want their choices to be so right, that they are right for everyone.
Right? There's this sort of I feel so strongly about this choice, or I made so many sacrifices to make this choice that It has to be that this is not just right for me, but right for everybody. And one of the things I push, particularly in encrypt sheet actually, where I think a lot of these kind quote unquote mommy ward debates come up.
A lot of what I talk about there is the idea that like a decision can be right for you and not right for other people. And that, just accepting that is part of both pushing back on the external pressure that you feel in general, but also a reminder that like we probably shouldn't be telling other people what to do quite so frequently.
Richie Cotton: That seems like great life advice in general
Emily Oster: like,
it's a hard one. It's a hard one because I'm like, you know, What's interesting is that of course I'm a very bossy person. And I work in these spaces and so I exercise a tremendous amount of self-control in trying not to tell me what to do. And one of the things that came out of writing crib sheet was I tried much harder to be like, I'm not gonna tell you exactly what I did, or I'm not gonna give you like my seven bullet point sleep training plan because I'm not sure that is what you want to do.
So occasionally I'll slip into that, with my siblings, but I try hard not to.
Richie Cotton: I think like messing with siblings is absolutely fine in like pretty mature culture, so,
Emily Oster: They're definitely doing it wrong. It is just, that's it. That's what your siblings are doing.
Richie Cotton: So actually it seems like there were some cases where you got like a fairly clear cut answer. So the sort of smoking and pregnancy thing was like, well, there's pretty much only one way to do it. But something like when you were talking about drinking alcohol during pregnancy, that took a lot of effort to get to some kind of conclusions it felt like.
And so what made one problem harder than the other?
Emily Oster: The main. Distinguishing feature between this easier things to answer and the harder things to answer is often the quality of the data. So take like something like smoking, A lot of the worries with smoking during pregnancy is an impact on birth weight. And so you can actually do randomized trials on smoking in pregnancy.
It's not a randomized trial where you encourage people to smoke, obviously, but you can do randomized trials of smoking cessation. They have trials where they like more strongly encourage one group than another to quit smoking, and it turns out when you do that, you can get people to quit smoking. To some extent and you can see that it, it translates to large increases in birth weight among the babies of moms who were less likely to smoke.
So you have something where we can be more confident that there's a causal relationship because the experiment is randomized. The places that I think are most challenging are those where we don't have data like that and where there are reasons that either story is compelling.
So alcohol is an example where we know that heavy drinking, that drinking a lot during pregnancy, that binge drinking during pregnancy is associated in children with negative consequences. So it has, there's an increase in behavioral problems, a condition called fetal alcohol syndrome that can occur with very heavy drinking.
And so we know that. But on the other side, there's an enormous amount of data that is just correlations. It's not randomized data, but we have, large data sets where we compare people who drink to people who don't. And there's a huge amount of variation in the quality of that data. So a lot of the work of that, part of expecting better is to dive into that data to say, okay, we actually wanna ask the question, what's the impact of occasional drinking on kids?
is it. What we see with heavy drinking or is it, and there are some reasons to believe this might be true biologically. Is it the case that drinking at lower levels would not show the same kind of impacts? Learning that out of the data is tough because you have to figure out what are the studies that do a good job of isolating moderate drinking, isolate, occasional drinking and some studies do a better job of that than others.
You worry that because it's not randomized, the kinds of people who drink are different from the kinds of people who don't. Some studies do a better job with that than others. This is the thing I am trained in. so part of what I am bringing to the table, part of the perspective and expertise that I'm bringing to the table is my training isn't like, how do we learn.
Causal relationships out of observational data and how do we evaluate the differences in quality in these different kinds of studies. I mean, that's my job. And so that's a lot of what I try to do in the data and try to translate out to people who are not trained in doing that.
Richie Cotton: Maybe we can nerd out for a bit about experimental design then. So, It seems like.
Emily Oster: love that. That would be amazing.
Richie Cotton: Yeah, so you've got like these randomized control trials is like the gold standard if you want to detect some kind of causal relationship, so A causing B or whatever. So when you don't have that, like how do you figure out something's causing something?
Emily Oster: I mean, I think it's tough. I mean, one, like very general. statement that I try to make to people is you're only gonna be able to learn a causal relationship if you have some effectively random variation in the treatment. So whatever is the technique that you're using, you need to think that, like you're isolating something random about the treatment.
So I think that's why I'm often quite skeptical of these like just observational papers where they like do what we call a selection on observables, where you just put in whatever controls you happen to see, like control for whatever demographics you happen to see. and then you say, well, that's like probably enough.
Because it's probably not enough. And there almost certainly are still, residual, confounding issues, concerns that there are other differences across groups. sometimes we can do better with things like sibling analysis, you know, within family analysis, which controls for more differences across families.
Sometimes we can do better with Policy variation, not so much in something like alcohol, but sometimes we'll have like policies which turn on and off and try to get a sort of do better with observational data. But I dunno, I think it's hard and I think a lot of what gets published is really poor.
Richie Cotton: It's one of those things where you think, okay, science is someone's published something that must be like decent quality research cuz at least it's been through a peer review process. And it turns out actually quite a lot of science is a little bit junk.
Emily Oster: Yeah, I think, I think that's absolutely right and I think it, it happens, partly because you'll read these papers and there'll be like a limitation section. Like, Well, this might not be a causal relationship. It's like, oh, it's definitely not a causal relationship. For sure not. I don't know. I find this kind of thing very frustrating because it, it does feel like it gets translated out. Like the incentives for people are almost messed up by what media is interested in covering. And so you'll have, one day coffee makes you live forever, and then the next day it kills you.
And it's like those cannot possibly both be true. And the answer is neither of them is true. They're both just terribly wrong.
Richie Cotton: So I guess one, one thing that strikes me with this is that once you're trying to figure out some kind of causality and you've not got this kind of really fancy randomized control trial, you've just got, you need some kind of domain knowledge there. So you need to understand maybe the underlying processes.
So how important is this domain knowledge?
Emily Oster: I think it's often important. I mean, I think If we think about our approach to data as more Bayesian as more about having a prior, set of beliefs that are informed by other aspects of the world or other evidence we have and so on, and that then every individual study is gonna add and move our prior in some way.
it can be valuable there to think about some relationships seem more plausibly true than others based on biology. And so there are more, there's like your prior should be in. Is like in a different place. And that is part of, where the belief that you should end up with at the end.
And I mean, priors, they're not in like things I think or wish were true, but Pryor is in like a sort of formal sense of like informed by existing evidence of various types. I think the other piece of this is that sometimes even though we don't have randomized trials, we do have better kinds of data or other things we can do with data as a test.
So like an example in the. In the context of of alcohol and consumption and pregnancy is a lot of which you worry about as demographic differences in people who drink and don't drink. And one of the things you can do is look at paternal drinking. We sort of thinking about like dads as a kind of a placebo test where The sort of demographic features are similar, have a similar kind of bias. But of course the sort of plausible relationship with infant health is isn't, is not there. And so there are some like, clever things people have done with data to try to get at some of these questions.
Richie Cotton: So one of the things in crib sheet you were talking about is recently become a hot topic in the uk.
So, the NHS has just launched this big promotional campaign about weaning children onto solid foods. Apparently, like a lot of first time parents don't know this big campaign about, well, when should you give your child solid food
Emily Oster: So one of they, I, so I, okay, so tell me more. This is, we don't have this in America. I mean, we have solid food in America, but what is the concern? Is it that people are giving their kids solid food too early or too late, or never.
Richie Cotton: I think both.
Emily Oster: Okay. Just, it's just wrong in all the directions. Okay, got it.
Richie Cotton: So, yeah. I know you talked about bit, a bit in cry, so can you tell me what the data says about
Emily Oster: So, so babies should be introduced to salad food sometime between four and six months. And that used to be like, there used to be more of a push for sort of four months. And increasingly in the US at least, people have tried to push it back to six months. And the reality is that like the. Some babies are ready for solid food at four months and some babies are ready at six months.
And some of it is about figuring out like, does your baby seem interested in food? Are they able to sit up and hold their head up in a reliable way? And so there's a little bit of parenting, parenting guesswork. One of the other things that comes up a lot is a question of what's the right first food?
There's a sort of traditional, in the US like the traditional first food is like rice cereal, and people would give their kids rice cereal and they would move up to purees. Now there's a push in some circles for what we call baby-led weening. Where you start by just giving your kid the food that you eat like in smaller pieces.
It's not quite that simple, but it's some version, some effective version of that. And you'll get people on both sides as an example of something where people say, well, if you don't, if you do baby-led weening, your baby's gonna choke immediately. And then on the flip side, people say, if you don't do baby lid weaning that your kid will like, Never learned to eat a variety of foods and it's gonna contribute to obesity and all kinds, and all of this stuff is not correct.
There's very little evidence that either thing is good or not good. So some, if really this is like something where the data says do what you want, and then the answer to what you want is what are your preferences? And people have different views about the ways that they want to they want to introduce foods, but outside of the very broad scaffolding of don't do it before four months.
You do wanna start introducing solids by about six months. And giving your kid a access to a variety of different foods is broadly good. Other than that, it's like a very kind of choose your own adventure experience,
Richie Cotton: So again, it's one of those things where there's maybe a lot of panic and there's probably like a few parents who I know they're feeding their children potatoes at two months or something.
Emily Oster: right.
Richie Cotton: Someone like not.
Emily Oster: Yeah, people will get, there's like what used to be this advice to put rice cereal in the baby's bottle like almost immediately. So it would, they would sleep better. That's not right. Don't do that. So there are a few like sort of old wives tales that are like, don't do it. I mean, it's interesting you think this is a place where parents find the freedom sometimes, not reassuring, but terrifying, right?
This sort of being like, just do whatever. It feels well, I, but I don't know what to do. I haven't done this before. I just, I want you to tell me like, here's what you do day one, day two, day three. And so you get these you can get to a place where people are relying very heavily on some particular piece of advice, partly just because they otherwise have no.
No special guidance. And that's fine as long as we recognize it. There's something special about that advice. There's a lot of good ways to do it, and it, having guidance is helpful
Richie Cotton: Okay. Yeah. It does seem like cadence is a good idea in general, as long as you comfortable with a bit of.
Emily Oster: as long as you give it with nuance. Like, this is one suggested way to do this.
Richie Cotton: And so a related kind of thing seems to be around eating peanuts. And this is one where there's like advice in opposite directions. Like some people are like, well, you should give babies peanuts, cuz they'll get used to it and it'll help with allergies. And other people are like, well, oh sh.
Avoid peanuts. They're very dangerous. Cause if your child does have an allergy, then they're gonna die. And how do you reconcile these sort of opposing points of view?
Emily Oster: One of those is right and one of those is not right. Um, So there was a, a sort of a view, a sort of a bunch of advice that GE was given generally until about, well, until exactly 2015 which was avoid exposure to peanuts, peanut butter allergens in before the kid turns one, because you don't want them to have an allergic reaction.
And these allergens can be dangerous and you want a way to introduce them till later. Then there was some evidence that came out actually out of a researcher in the UK named Gideon Lack So they, a few different papers, but the really key one is a randomized trial of early exposure to peanuts.
Not to peanuts specifically. You cannot give nuts to a baby, but peanut protein, peanut powder and it turns out that exposing kids to peanuts early in the four to six month range lowers the risk of developing a peanut allergy by 70%. So this effect is not just. There, but it's huge.
It's like huge. It ha it works even for kids who are already showing some sensitivity to peanuts. So it doesn't work for a hundred percent of kids, but it is really big effect. And it seems to translate over to other common allergens like eggs. Milk, there's a few allergens that are responsible for almost all childhood allergies.
And so the advice switched from wait to introduce peanuts until later to introduce them. as soon as you're doing oth other things and I think that advice is correct. The whiplash in the advice is hard for people, right? So the idea of you told me this one thing before and now you're telling me this other thing, why should I believe you?
This is why it's really important to explain to people, well, the reason is, we were wrong before and here's the study that, it's not hard for people to understand this. Here's the study that showed us that this is the right thing to do, and so here's what you should do now.
Richie Cotton: Again, it's about updating your beliefs based on new data and new evidence.
Emily Oster: Yeah, so for my kids, like when I was a kid you could have peanut butter and jelly sandwiches at school and, but my kids who are like 11 and seven there is no. Peanut butter at school. It's no nuts. Nu free school. My daughter is in the peak of the kids who are in the space of having peanut allergies because the advice we got when she was born was like, don't expose your kid to peanuts until they're due.
But my son is like a tw born in 2015, so he's like the first cohort of kids where they were like, give them peanuts right away. So actually nobody in his class has a peanut allergy.
Richie Cotton: That's absolutely
Emily Oster: It's amazing.
Richie Cotton: Wow. I hadn't realized it was such a strong effect. So, I'd also like to chat a bit about developmental milestones. And I have to say, I was a lazy baby. You missed every milestone. I didn't walk till I was nearly two, and so I'm kind of like, well, do developmental milestones really matter?
That's just like a sample size of one person. So, the data say? Are there any important milestones that you need to watch out for?
Emily Oster: So we talk about sort of two categories of milestones. so one is about physical milestones and one is about like talking milestones. We talk about physical milestones first. And there, I think that there the main. Takeaway for many parents is there is a wide range of what is normal.
So for something like walking the range of ages at which kids start walking, kids who are developmentally normal, who will end up walking fine, who do not have, a physical limitation, that range is like seven months to 18 months. So that is a very big range. if you have been a parent, the difference between a seven month old and an 18 month old is really big.
And so we get into this thing where we tell people, well, kids walk on average at one year, like at 12 months. That's true, but it misses a huge amount of nuance. And that's true for almost every one of these milestones. Something like crawling, actually, like kids never crawl. so again, there's a, for the physical milestones, there's a really big range.
Some kids won't hit certain ones of them basically ever. And that's okay. This is a place where pediatricians play a really important role because there are things, conditions that will show up after birth that you can only diagnose by recognizing a missed milestone. But there are things where your pediatrician really has to see your kid like in person, right?
And they, and once they see them in person, then there are some ways that they can interact with your kid that will give 'em a sense of is there a low tone issue? Is there some other reason they're not doing this. But most of the time it's just like your kids. Mean, I like to call babies lazy.
They're working on other skills other than walking, and then later they will learn to walk. And then there's the sort of developmental milestone piece where like talking is the most significant. And again, there's a lot of variation across kids in how quickly they learn to talk. Some of that is a little teeny bit predictive of later like school performance, but only in a very, Moderate way.
And again, then there's a sort of a lower bit of tail at which you do wanna say, okay, this is like outside of the range of normal. And then there's a bunch of things that, parents can do to try to understand better what's going on.
Richie Cotton: Okay, so it seems like in general there's gonna be a wide range. Don't freak out like some,
Emily Oster: The people who f you know, there's also gender differences, particularly in the like, language stuff. So the people freak out the most tend to be the people with an older girl and a younger boy, because boys take a longer time to develop language on average than girls at these young ages.
It's not that they don't eventually learn to talk, it's just that they're slower. So sometimes people will be like, my daughter was talking like a mile a minute, and my son has four words. It's yeah. He's coming. He's coming. He's gonna get there.
Richie Cotton: Yeah, I mean, I suppose one you see adults, eventually, most adults do. So,
Emily Oster: Most adults do learn men, even men, if sometimes too much. I would say honestly, sometimes they talk too much.
Richie Cotton: One of the issue that sort of crops up, in fact across all three books, but I think it pays most of the, the last book is around sleeping As I figure this is like relevant adults as well. It's do adults get enough sleep? Do your children get enough sleep? So, any sort of insights you bound around?
Like what's a good amount to sleep?
Emily Oster: I find sleep to be like the most fascinating thing because it is. Really important. Basically, all animals, like all mammals, at least sleep. And many non mammals, even like animals that like swim, so like dolphins will like, like they can rest half of their, I realize this is not relevant to parenting, but dolphins will sleep half their brain at a time.
So you have like half of it awake so you can remember to breathe and the other half is on awake and then they, and they can switch. So it's very clear that sleep is like incredibly important, but we have no idea why. And it must be very important because everybody does it and it's tremendously maladaptive.
If you think about the idea that you have to lay down for eight hours in the middle of every day, like this is like a great time for, the tigers to get you right. So it must be that sleep is so important cuz the evolutionary pressure would be so great against it.
so sleep is really important. that's one thing. we think about kids, Like, most kids are probably not getting as much sleep as they should. So the kind of average sleep needs vary across age groups. But you mean get to like elementary and middle school kids need nine to 11 hours of sleep a night.
And I think for little kids, a lot of the barriers to sleep are just like, kids have a hard time going to sleep. And the, there are some things there. But with little kids, they tend to be able to nap. There's like a lot of ways for them to get a lot of sleep in with older kids, I think we start getting into a place where kids are just not sleeping enough.
And you see changes in sleep, not have, not having enough sleep translating to like lower performance on tests, like worse behavior. And so we see this actually in randomized trials. So you can see an adults where if like you bring adults into a sleep lab and you keep them up all night and then you give them a test you do this.
The college students they think that they did great of a test, but actually they do very poorly. So it's like That resonates. But with kids, you can't leave them. Can't keep them up all night, obviously. But when they do experiments where they have like elementary school kids and they have their parents keep them up like four days in a row for one extra hour.
So pretty small manipulations. And then at the end of the week, they'll have them come and do tests. Actually, that lost hour for four nights translates to like worst performance on these tests. And their parents saying that they're acting out and like all kinds of stuff.
So really, like Sleep for Kids is something that I think people don't prioritize in enough. And then people ask me, well, how do I know how much? Like how do I know my kid's getting enough sleep? And that turns out to be one of the easier questions to answer. So one thing is they shouldn't be sleepy.
So if your kid is like, Falling asleep in class or says that they're super tired at school. Then they are not sleeping enough. And the other test is what we call weekend oversleep. So most of us wake our kids up for school in the morning, or there's like some way in which they're woken up for school artificially.
And then on the weekends, maybe that doesn't happen as much. If you let your kids sleep in on the weekend and they're sleeping like an extra two hours. They're not getting enough sleep during the week, but basically you should get up. You should be like, your body should be waking you up at about the same time every day if you are a well-rested person.
Richie Cotton: So a lot of that sort of ring through like the idea of sort of college people just like staying up all night trying
Emily Oster: And then trying to cram for their test and then being like, I'd probably, I probably killed it. I
Richie Cotton: And in fact, actually a lot of data scientist as well, same working professionals yeah. Trying to stuff done.
So actually related to the, to this idea of finding enough time for sleep is finding time for things more generally. And I think it's particularly true for working parents is basically where do you find the time for everything? Maybe we'll start off with is there any data on just having two working parents have an effect on the child?
Emily Oster: No I mean there is data, but it doesn't have an effect. Like when we look at things like, people ask is it better to have one parent work and one parent stay home, people work part-time, et cetera. Like that data, it's not very good because it's not very easy to learn from, let's say because.
The characteristics of families where both parents work or where, versus parents, where families where one parent stays home. Those tend to be very different on a bunch of dimensions. But to the extent we have links, we have data that tries to link these things. We really don't see much in either direction.
Just like any effects that you see are small. So they're probably actually zero or they're like, they're swamped by every other thing. But to the extent that we, that you even believed what we saw, It would not be like, they're not important. They're not like, magnitude wise, important important effects.
So I think, that says to me like there's a lot of good ways to organize your family. You gotta figure out what are the ways that work for you? And I also, I mean, let me say one other thing about this, which is that when this is framed, at least when we talk about this in the us it's often framed as should there be a stay-at-home mom or not? I really don't like that framing because I think both it is like incredibly heteronormative. It's really gendered. Why is it not a stay-at-home dad? Why are there not two moms? But also I think it, it neglects the fact that actually there may be more choices than are recognized by the idea of one parent works and one parent doesn't work.
You know that there are ways to work part-time or, there may be ways to adjust on, on, on the intensive margin that are really productive to think about. And then I'm not sure everybody is considering when we set it up in this very bifurcated way.
Richie Cotton: so, you went into, I can't speak there, so something about the bifurcated margin. Sorry,
Emily Oster: Like the intensive. Okay. Just that like, fair enough. Um, So I guess my point is just that That we tend to think about, if we were gonna have 40 hours of work in the family, we tend to think about it. Like the only way to achieve that is one person works 40 hours and one person doesn't work at all.
But another way to achieve that is both people work 20 hours. And I think when we have these conversations, we sometimes miss the possibility of dialing down work a little bit at certain times in our life without completely eliminating it. I think that's something where I would like. Employers to have a better handle on, how can we retain talent by recognizing that there are stages of people's lives in which they would like to, not work 90 hours a week, but they may eventually be interested in returning to 90 hours a week.
We don't have that much of that set up.
Richie Cotton: Yeah, so, a lot more flexibility in the workplace seems like a great idea. And just related to this, I mean, so you're a professor of economics, you've written three books and you've got your own parent data organization, and you're doing work with a National Bureau of Economic Research. So you got a lot of things going on as well as being a parent.
So I'm hoping you have some good productivity tips if you are juggling all these different things.
Emily Oster: Not really. I don't know. I mean, most of my productivity tips involve outsourcing. I have a partner although he also works a lot. We have. Really some really amazing childcare, which I always think is important to acknowledge because, somebody else is doing at least some subset of my childcare and other things in my in my house.
then we try to, like every, I try to have everything be really regimented, which is boring and I think makes me a somewhat boring person. But ultimately ultimately is probably the key to like making anything work.
Richie Cotton: I mean, so I think boring's fine, but it's interesting that like being regimented and doing things in order to a schedule is actually the secret source to productivity.
Emily Oster: I mean, I think not for everybody but for some, I think being regimented and being able to do many things at the same time are my key things. Like I can write while my kids are yelling at me, and if I wasn't able to do that, I don't think I would be able to write as much as I do.
Richie Cotton: Alright, so, one thing I noticed about like the last book you wrote, so, the family firm is once you start getting into older children, it's a lot more like cultural issues rather than the biological stuff. So With pregnancy, it's well, okay, the biology just says smoking's bad.
But when you're trying to raise a child, it's more about, oh, how do I feel about this culturally? So, I guess, how does this affect like the data side of things to begin with?
Emily Oster: The main challenge when we get into these areas is that the effects are more. Heterogeneous, they're more different for different kids, and that's gonna be partly because kids are more different, but partly because their cultural context is more different. And so it's harder to use the data, even the data we have to draw conclusions about.
The population we might be interested in. So with something like some of these questions in pregnancy, it's a little bit easier to say, look, the biology's kind of working the same way. So even though this data is from, the Netherlands I feel like confident that, we could use it to speak to the US or the UK or whatever.
When we're talking about schools, I don't feel confident that I could say, we studied this school, different kinds of schools in the Netherlands and we're gonna use that to speak to, charter school quality or something in the us Like the context is just way, way too different.
And so I think that limits the scope of the data we can bring to bear on these questions. And at the same time, I think a lot of the kinds of choices that parents make are gonna matter really differently for different kids. And. That won't be captured on by the average. So even though you had a great study on the right population that told you the average effect, it wouldn't necessarily tell you the effect for your kid.
And it could be very different. And I think that's to some extent true in some of these earlier things, but just to a much, much lesser extent.
Richie Cotton: That's very interesting. So in general, like the cultural stuff is much more localized and well, obviously it doesn't translate across cultures. So, I guess from a decision making point of view, as a parent, like how does this sort of different data, like the more cultural data affect the decision making process?
Emily Oster: I mean, I think part of what I emphasize in the third book in the family firm is the idea that like, you need a decision making process. And so some of what you're gonna bring to bear on your decisions is different kinds of evidence. And that could be data from other cultural contexts. It could be data.
That's not quite so good, but it's from your context and can be information about your kids. Like there's many different pieces of information you wanna bring to bear. Like the most important thing is to think deliberately about the choices that we're making. And so, sometimes we have to recognize I don't know what the right.
Decision is here. all I can be confident about is that I've made the decision in the right way, that I've thought about the options that I've clearly articulated, like option A versus option B, that I've gotten the information I need, that I've made the decision thoughtfully.
And at the end of the day, you're often not gonna really know if you're right in the moment. You may never know if you're. right, which is different from little kids, right? So sometimes with little kids is like sleep training. You make a decision to sleep train, you can see how it goes.
You're basically gonna find out if like it helps your kids sleep or not. You're gonna find that out real fast. With bigger kids, you can make some choice and you just be like, well, I don't know. Like I make a choice about give my kid a phone, when to give my kid a phone. I'm never gonna learn if that was the right choice.
I'm literally never gonna find out if I made the right choice or not.
Richie Cotton: Yeah, it's like even if you have twins and give a phone to one, that shouldn't be a weird situation.
Emily Oster: then they, yes, active experimentation. The sample's too small. Also forget it. You need it up. Like many sets of twins.
Richie Cotton: Alright, so I think I love our listeners. They're interested in working with data or figuring things out using data. So a lot of the stuff you did was, it seemed like you were doing a lot of literature reviews and going through vast amounts of evidence. So, how do you get started trying to understand things using other people's data?
Emily Oster: Usually the key to understanding other people's papers is to read the appendix tables. I find I think that when I. Do this kind of work. Some of it is you just learn to do this better over time. I mean, this is my job. And then this sort of particular version of it I've done a lot of times I think it's often about finding, finding the, where the thread starts and then pulling it, saying I'm gonna find what's the best paper on this?
What's the like thing everyone talks about? And then let me pull on that thread and figure out, who are they citing and who's citing them, and how is this literature kind of evolving? And then you can dive into okay, like among this sort of set of papers, like which of these are doing a good job?
Which of them in ring less of a good job? And that's where I like to look at the appendix tables and see what's what. That's usually where people put their, that's where they put the bodies.
Richie Cotton: That's kind of cool. So rather than reading things in audience, jump straight to the end, this is where like the ju. See data exists right in the appendix.
Emily Oster: And that, that is a thing that takes time to, to learn. So one of the things is that because I look at so many papers, particularly in this space of I'm correlating this with this and like this behavior is linking to this outcome. I. Most of these papers have very similar structure, so you always know, like somewhere there's gonna be a table where they compare the features of the two groups and you can read the tables.
I can read the tables pretty fast.
Richie Cotton: And are there any good sources of data you recommend around parenting?
Emily Oster: I mean, I, I think my, I think my, my website's a good source of data for parenting. you know, I mean, I think there, there are more technical and less technical versions of this. I think in general, when people ask me like, who should I trust? I think, you know, usually I don't always agree with them, but usually a source like WebMD, Or the American Academy made in pediatrics is that's like a place to start.
And that's gonna give you at least and informed set of information rather than, Dr. Google.
Richie Cotton: Since you mentioned it do you wanna tell us about what's on parent data?
Emily Oster: Yeah, so parent data is really a sort of extension of the work that I do in the in the books. It is a a newsletter on subs and I write about many of these same issues. Pregnancy, parenting decision making data. I try to unpack new papers. I try to. Flesh out broader topics that I talk about in the books, but not as much.
And then there's a lot of fun stuff. There's q and a, there's like people weighing in with their own parenting problems. So we're trying to build a little bit of a community around parents who love data.
Richie Cotton: That's pretty cool. Data and community together is brilliant stuff. So, just while we're talking about. Researching things. What sort of skills do you think are useful if you want to go in and investigate some of these ideas from your books? More?
Emily Oster: I mean, I think an ability sort of some basic training in statistics and causal analysis is key. So that's the research design methodology. Understanding a little bit where data comes from. I think those are the key elements.
Understanding like how you do the analysis, sort of less important than understanding like conceptually. What is the source of this information? What's the source of variation? How do we think about, how do we think about causality? I think those are the key elements. And then just a lot of practice.
Fortunately in the, like this medical space, the papers all look basically exactly the same. And so once you learn how to read them, you can digest them pretty fast.
Richie Cotton: So it seems like we had a little nerd out about experimental time before. It seems like that's the kind of the key then, like understanding how the experiment was set up is more important than maybe just, well, yeah, I know the details of this particular model type.
Emily Oster: Yeah, I mean, I think that's the, like at the core, all of these papers are gonna rest on, what is the source of variation? What is the claim to exogen the claim to what is the claim to random variation in the treatment? I think that's always what we're trying to dive into in understanding these papers.
And that's not really about understanding the mechanism by which they're doing the analysis. It's really about understanding like what is the conceptual idea behind the analysis.
Richie Cotton: And are there any common mistakes you think people make if they're going off trying to research things on their own?
Emily Oster: think mostly like overly trusting the vetting process of. Referee journals. I mean, I think a lot of times people will come to me and they'll say, well, this was published in a journal. Yeah. Okay. That doesn't necessarily mean it's right. And I think that's an easy mistake.
Sometimes it's right, I think that's an easy thing to get dug into.
Richie Cotton: Maybe slightly more trustworthy than. Like random forums on the internet.
Emily Oster: Better than Reddit. Better than Reddit, but not perfect. Or at least like having people understand that there's quite a lot of variation within this space of published work in like how high quality something, something gets.
Richie Cotton: fantastic. Do you have any final advice for parents wanting to make more sort of data-driven decisions with their parenting?
Emily Oster: I think mostly to Trust themselves. I think that you're smarter than you think and your ability to understand these problems is greater than you might imagine. And so I, I would like people to feel empowered to to engage with these questions themselves and, make choices that work for them.
Richie Cotton: Fantastic. Alright thank you very much for your time, Emily.
Emily Oster: Thank you so much for having me.
How is AI Transforming Data Management?