Preschool: I Was Wrong

Kelsey Piper has written an article for Vox: Early Childhood Education Yields Big Benefits – Just Not The Ones You Think.

I had previously followed various studies that showed that preschool does not increase academic skill, academic achievement, or IQ, and concluded that it was useless. In fact, this had become a rallying point of movement for evidence-based social interventions; the continuing popular support for preschool proved that people were morons who didn’t care about science. I don’t think I ever said this aloud, but I believed it in my heart.

I talked to Kelsey about some of the research for her article, and independently came to the same conclusion: despite the earlier studies of achievement being accurate, preschools (including the much-maligned Head Start) do seem to help children in subtler ways that only show up years later. Children who have been to preschool seem to stay in school longer, get better jobs, commit less crime, and require less welfare. The thing most of the early studies were looking for – academic ability – is one of the only things it doesn’t affect.

This suggests that preschool is beneficial not because of the curriculum or because of “teaching young brains how to learn” or anything like that, but for purely social reasons. Kelsey reviews some evidence that it might improve child health, but this doesn’t seem to be the biggest part of the effect. Instead, she thinks that it frees low-income parents from childcare duties, lets them get better jobs (or in the case of mothers, sometimes lets them get a job at all), and improves parents’ human capital, with all the relevant follow-on effects. More speculatively, if the home environment is unusually bad, it gives the child a little while outside the home environment, and socializes them into a “normal” way of life. I’ll discuss a slightly more fleshed-out model of this in an upcoming post.

My only caveat in agreeing with this perspective is that Chetty finds the same effect (no academic gains, but large life-outcome gains years later) from children having good rather than bad elementary school teachers. This doesn’t make sense in the context of freeing up parents’ time to get better jobs, or of getting children out of a bad home environment. It might make sense in terms of socializing them, though I would hate to have to sketch out a model of how that works. But since the teacher data and the Head Start data agree, that gives me more reason to think both are right.

I can’t remember ever making a post about how Head Start was useless, but I definitely thought that, and to learn otherwise is a big update for me. I’ve written before about how when you make an update of that scale, it’s important to publicly admit error before going on to justify yourself or say why you should be excused as basically right in principle or whatever, so let me say it: I was wrong about Head Start.

That having been said, on to the self-justifications and excuses!

1) Head Start seems to work for reasons unrelated to the ones that made people want to do it. Those people were still wrong, and this is still a good example of policy effects being difficult to predict. It seems to have succeeded by coincidence, not because “early childhood education” is a good idea.

2) This probably strengthens rather than weakens the Caplanian case against education, since the studies find that the educational parts of preschool are not useful, and better teachers and curricula do not affect the benefits.

3) This strengthens rather than weakens the case that academic achievement is related primarily to IQ, and that IQ is primarily genetic and difficult to change. An intervention targeted at academic achievement and IQ manages to change everything else except those variables, which remain stubbornly the same. Studies consistently find that IQ is only responsible for about 25% of life outcomes, suggesting that education works on the other 75%.

But on a broader scale, this does lower my confidence in biodeterminism. Preschool is a shared environmental effect; your parents have a big effect on whether or not you go to preschool. Why doesn’t this shared environmental effect show up in studies, which generally find no shared environmental effect matters?

This is the same problem raised by Ozy’s post on lead. We know lead is important. We know it can damage your life outcomes. But we also know lead is related to the shared environment. And we also know studies keep finding the shared environment doesn’t matter. Some studies find the shared environment matters a little, when you make extra-double sure to have very high income inequality in your sample. But other studies find that it doesn’t, and almost all of them find that it doesn’t matter much at the still-high levels of income inequality you get by recruiting a convenience sample. How can this be? We have two really excellent and well-replicated scientific literatures, each proving opposite things. What now?

All I can think of is that maybe shared environment can matter, but is so small in the grand scheme of things that it’s below the threshold where zoomed-out studies of everything can detect it. That would help reconcile the two literature bases. But it doesn’t seem right. The lead effects are huge. The preschool effects, while moderate, suggest that something as minor as “whatever social advantage your family gets from your mother not having to take care of you for part of the day from ages 3 – 5” can have lasting and detectable effects. Surely then we would expect much larger effects from whether your mother is independently wealthy and can do whatever she wants, or whether your family otherwise has the ability to accrue social advantage.

It might also be an effect of what we’re measuring. Although there’s conventional wisdom that shared environment shows little effect in twin studies, there are occasional outliers. For example, studies of crime often find shared environment factors around 15-20%, especially in younger or poorer samples. And some of the studies that found effects from preschool measured crime. These are some inconsistent findings, and 15-20% from everything doesn’t seem consistent with measurable effects from preschool alone, but I’m kind of desperate here.

I guess I will just increase my belief in the studies that suggest shared environment matters a bit more when you limit yourself to non-cognitive factors and include the really poor, and hope that future work confirms this result.

I’ll also increase my political support for programs like these. I think these findings make universal childcare (almost) a no-brainer. They make universal pre-K much more appealing, with the strongest arguments against being inefficiency, eg that universal childcare or basic income are a more effective way of doing the same thing. But given the political realities that make universal pre-K more likely to happen than childcare or basic income, I am now happy to support it.

This entry was posted in Uncategorized and tagged . Bookmark the permalink.

269 Responses to Preschool: I Was Wrong

  1. spineback says:


    On another note, would this be considered worthy enough to make it onto your mistakes list?

    • Scott Alexander says:

      I’ve been trying to figure out how to handle that. As it exists now, it’s just about minor factual mistakes in posts. I should probably update it in a way that covers things like this better.

  2. Elisha says:

    I kind of feel that beyond things like lead in the water supply, there really is no such thing as a shared environment, definitely not when looking at single house holds as ‘shared environment’.
    Each individual experiences the environment differently and feels different pressures. For instance, three siblings in one household would each have a different experience:

    The first-born is on one hand ‘clearing the path’, doing everything first – from school and social activities to finding hobbies and developing an independent personality.
    Second born has an example to follow, and can either confirm or diverge, but it will always be with the anchoring of what the first-born does. So it’s a different type of skills developed.
    Third already has two examples to look at…

    Take into account the counter trends of parents gaining experience but also growing older and more tired down the line – the whole concept of shared environment seems to me to require so much nuance that it’s almost irrelevant, any effects will be masked by birth order effects.

    Perhaps my definition of ‘shared environment’ is not rigorous enough, and birth order effects can be controlled for?

    (I’ve heard that’s it’s accepted among the sociologists that birth order effects don’t exist… but that seems strange to me. I admit it’s not my expertise, but from my experience, they seem real.)

    • Alkatyn says:

      Shared environment doesn’t have to be absolute to be absolute to be useful. Children raised in the same household will have more shared factors than next door neighbors, who will have more in common than people in another neighborhood, etc. Degrees of similarity can still be useful statistically when you have large enough samples

    • notpeerreviewed says:

      Even household lead exposure might not show up as a shared environment effect. One of my uncles had a bedroom where he was exposed to lots of lead as a child, but neither of his siblings was exposed. Based on his life outcomes compared to theirs, it’s likely that the lead exposure had a big effect on him, but it would show up as “nonshared environment.”

      • Scott Alexander says:

        The correlation is less than 100%, but surely greater than 0.

        • notpeerreviewed says:

          Might it be low enough that 15-20% shared environmental effects seems like the right number? That’s what I’m getting at.

      • Doug says:

        Alternatively lead exposure might have a genetic component, even when two siblings share the same parents. I’m guessing the propensity to eat paint chips probably is inversely correlated with childhood IQ. If Bob inherited a lower polygenic score from the random shuffling of his parents’ genes than Alice, he’s more likely to have higher lead exposure.

        • Hoopyfreud says:

          I’m guessing the propensity to eat paint chips probably is inversely correlated with childhood IQ.

          Why on Earth would you guess that?

        • rlms says:

          That seems rather implausible. If anything, I would expect the effect to be the other way round, where more intelligent children explore their environment more and are thus more likely to encounter lead. But either way, I’m pretty confident the effect of genes on lead exposure is very small in comparison to the effect of whether your bedroom has lead paint or not.

          • albatross11 says:

            Don’t forget about parental effects. My genes came from my parents, and so there will be correlations between my genes and the nature/quality of choices made by my parents in my upbringing. That could involve lead paint exposure in a variety of ways.

          • rlms says:

            Sure, I wasn’t very clear. I agree that there is probably a correlation between genetic IQ and probability of lead exposure in the population at large. The implausible claim that there is a significant correlation within groups of siblings.

  3. bobbertonian says:

    Interventions will always show an effect on some variables if you can keep sifting through outcomes until you find an effect.

    • albatross11 says:

      Yeah this is my concern. I haven’t read these studies, but it would be awfully easy to propose a policy (universal pre-K, say) on the theory that it would improve academic success, implement it, and then after the fact find some variables other than academic success where it seems to improve outcomes. But in order to evaluate that, we need to know something about how many different choices there were for choosing those variables. The garden of forking paths is always lying in wait for us when we try to interpret this kind of data. That’s why people require pre-registration of studies to really nail down effects.

      • JulieK says:

        It’s hardly surprising that providing a benefit to needy families will make them materially better off. It would be pretty pitiful if it didn’t. The real question is, is providing $X worth of preschool better than providing $X in cash?

        • albatross11 says:

          Well, there’s also the issue of selling a policy on the basis of its actual benefits and costs, rather than imaginary ones that don’t hold up.

          It’s entirely plausible that there are benefits to families from offering them free daycare for their kids, or benefits to kids from really awful home environments when they get to spend most of their time in the presence of functional, well-behaved adults who feed them and don’t beat them up. But I think I’d like to see those benefits studied directly (state up front what you think the effect will be and how you’ll measure it), and I’d like to see the policy sold to the public on the basis of its actual effects.

          • Deiseach says:

            It’s entirely plausible that there are benefits to families from offering them free daycare for their kids, or benefits to kids from really awful home environments when they get to spend most of their time in the presence of functional, well-behaved adults who feed them and don’t beat them up.

            The families need not even be abusive, just not educated middle-class professionals who know all the ropes and have the college little Tamsin is going to attend already picked out when she’s in the eighth week in utero. To take an example from the Irish preschool guidelines:

            Learning experience 3: Helping me to learn

            Theme: Communicating, Aim 3 and Learning goal 2
            Age group: Young children
            Setting: Home and infant class (primary school)

            Kara (4 years) is in junior infants. Her parents left school early. They have difficulties with literacy and know this is a disadvantage. They really want Kara to do well in school and to get a good education. But Kara says she doesn’t like school. Kara and her family have the support of a Home School Community Liaison co-ordinator, Betty. Betty encourages Kara’s parents to talk to her teacher, Ms. Nugent, and she suggests some questions they might ask. Ms. Nugent encourages them to help Kara in whatever way they can. She suggests that they use a picture book to read a story or to tell her stories themselves about when they were children. They can draw pictures together at home and talk about them. If they have time they can come in some days and help out in the classroom.

            Ms. Nugent also encourages Kara in school by asking her what kind of books she likes to look at and read. Kara replies, Books about babies are good and books about dressing up and going to my friend’s house. Ms. Nugent regularly uses books on these topics when reading stories to Kara and her friends. She puts dress-up clothes and props such as tiaras, dolls, buggies, and hand-bags in the pretend play area. Ms. Nugent regularly talks to Kara’s mam to see how they can continue to work together to support Kara at home and in school. Betty also liaises with Ms. Nugent and Kara’s parents regularly to ensure Kara and her family have positive school experiences.

            Reflection: What can I do to give extra support and encouragement to some parents?

    • Edward Scizorhands says:

      If you do the study and then keep on sorting the data until you find some effect and declare that effect as victory, that’s bad. A lot of things will show up by chance.

      But if you do the study, and then find some effect, and then do a study to measure if that effect is real, that’s good. (Assuming the protections we want from modern science, like publishing negative results and pre-registration of studies.)

  4. entognatha says:

    I basically agree with this post 100%. It also explains the contradictory finding that countries like Finland that don’t start formal education until 7 actually have better educational attainment.

    To whit, free early childcare is very beneficial for children of poor families, but the educational element in particular is not, and the focus in pre-school and kindergarten should be play-based. Unfortunately the current trend in the US (and the UK) is to push the formal schooling elements earlier and earlier.

    In the UK, the gov’t funds 15 hours of childcare for 2 year-olds if their parents are on benefits. It has universal half time (15 hours) childcare for all 3 year olds. Parents can use these hours at a traditional daycare or at a pre-school as needed, and many public (state) schools have a pre-school within the school. Formal schooling begins at 4. (IMO this is much too early)

    I think this would be great if the US adopted something similar, minus starting formal school at 4.

    • James Green says:

      Yeah, I feel like this does nothing for rich kids. It would be better to just stop allowing there to be poverty, but that is just far, far outside the Overton window.

      • albatross11 says:

        We don’t know how to make there stop being poverty, we only know how to hand money out. Those two aren’t the same. The children of a married couple who are both in grad school are being raised without much money, but probably won’t show many of the marks of having been raised in poverty.

        • thomasthethinkengine says:

          Are you getting poverty mixed up with good parenting? because we can easily fix poverty by handing out money (it is definitional), but it doesn’t necessarily create good parenting.

          There are rich kids whose parents don’t read or talk to them. And poor kids whose parents read and talk to them.

          The grad student parents are in the latter cohort, probably. they confound the overall finding on poverty because they are in a very transient state of depressed income (and honestly they are not really poor in a consumption sense. They are engaging in consumption smoothing, consuming a shitload of education now in order to consume a lot of other things later.)

          • albatross11 says:

            Right. My point (perhaps made too pithily) is that a lot of the bad things that come from poverty turn out not to be about not having enough money, but rather about other things about the parents/siblings/neighbors that wouldn’t be made better even if everyone involved got handed some extra cash.

            I think one part of that is that there’s a lot of life-dysfunction that leads you to both be poor and be a lousy parent–for example, if you’re an alcoholic, or have a serious mental illness, or a serious chronic physical illness, that probably contributes to both poverty (making it hard to hold down a job) and to being a crappy parent (because you’re too drunk/depressed/ill to do much parenting).

          • Tarpitz says:

            Anecdote pointing in the direction of what Albatross is getting at:

            I have a friend who is the eldest of an ungodly-large number (I think 9?) of children of a single mother, by several fathers. They grew up on a council estate with sod all money. They’re all also extremely smart and high achieving in their various middle class fields, and obviously code middle class culturally. Why? Because their low income single mother is an Oxford graduate hippie dropout and an excellent parent, so they haven’t actually missed out on any of the potentially developmentally important things poor kids usually do – and presumably they have pretty good genes to boot. My friend’s two sons are also going to grow up without much money: their parents work in theatre. I wouldn’t bet on that being a barrier to their long term prospects either.

        • James Green says:

          It’s easy, just tax everyone’s income at 50%, then take half that money and evenly redistribute back to the population. Spend the other half on building and operating infrastructure, part of that infrastructure would be offering to teach new parents how to best care for infants.

      • Cliff says:

        After-transfer poverty is down to 5% in the U.S. so it doesn’t seem that outside of the overton window. Nor does it seem likely to be a panacea

    • b_jonas says:

      Wait. “15 hours of childcare for 2 year-olds”? Does that mean 15 contiguious hours during a day? That is a way longer period than what’s generally offered, and I find it really unusual for children as young as 2 year old.

  5. Unirt says:

    But – these are correlative studies? Can’t they imply just that parents who worry about their kids’ education are more likely to send them to preschool, push them to go to college and tell them not to commit crimes? Did I miss something? The sibling study (siblings who attended pre-K were more successful than their not-attending siblings) can also be non-causal: perhaps parents decide whether their kid is ready for pre-school based on its behaviour and social skills, so that the ones who attend are the ones who have better skills already. The comparison of poor counties with or without the program is somewhat more commpelling. And the Abecedarian Project was a randomized trial, though the randomization was poorly done?

    • niohiki says:

      I’d say that if it was so obvious whether a person is going to succeed or not at pre-pre-school ages that average parents can already reliably guess, with nothing more than their trustworthy human heuristics, whether it is worth to pre-school them (and assuming parents are rational utility maximizers and will not just send both of their kids into the same opportunities because something something love) then… it would be so obvious that such detailed metastudies should not be necessary.

      • Unirt says:

        No, I didn’t mean it like this… I’m a parent and my kids are totally different, and I assume that one of them would have been much more happy in preschool than the other. I avoid, when possible, sending them to places where I think they’d be unhappy.

        I’m not saying that the careful meta-analysis is unnecessary or that the conclusion is wrong or anything; it’s just my impression that such correlations in social interventions often turn out to be not causal.

      • I’d say that if it was so obvious whether a person is going to succeed or not at pre-pre-school ages

        It doesn’t have to be “so obvious.” It only has to be obvious enough to result in a non-random selection biased in the right direction.

        That’s the usual problem with non-controlled studies. Even a weak selection effect can give you a strongly significant result if the sample size is large.

      • JulieK says:

        A parent deciding to place one child in preschool but not his sibling doesn’t necessarily mean the parent thinks child 1 is smarter.
        Maybe child 1 is more challenging and parent wants a break; or maybe the reason has nothing to do with the child’s personal qualities, but instead has to do with the parent having a new full-time job, or a new baby, or whatever. (Those were some of the reasons for why my children entered daycare or preschool at particular ages; with one child, I was recovering from mono and needed to rest.)

    • JulieK says:

      The sibling study raises a big question for the hypothesis that preschool helps by enabling the parents to get better employment, since one would expect the better employment to also benefit the other children in the family.

      • baconbits9 says:

        The main question about all of these studies is if the effect is partially coming from a lowering of achievement in the “control” group, or a relative performance gap that can’t be generalized which is a sticky wicket to test for.

    • Doug says:

      > And the Abecedarian Project was a randomized trial, though the randomization was poorly done?

      The Abecedarian started out as randomized but didn’t stay that way:

      > First, 11% of families (7 families) assigned to the Abecedarian group dropped out of the study after learning their group assignment, and their outcomes were not tracked. The same was true of only 2% of control group families (1 family). Such self-selection out of the Abecedarian group, a violation of the “intention to treat” principle, could have distilled the group down to those families most committed to their child’s education, and thereby undermined the equivalence of the Abecedarian and control groups in their degree of such commitment. This difference in family commitment between the two groups, rather than the Abecedarian program itself, may at least partly explain the superior outcomes observed for the Abecedarian group.

      › Second, six months after random assignment, 8 additional families were recruited to replace those who dropped out of the study. It is not clear how these families were recruited, but they were disproportionally, and we presume non-randomly, assigned to the Abecedarian group (7 to the Abecedarian group vs. 1 to the control group). Such non-random allocation of the additional families may have further undermined the equivalence of the Abecedarian and control groups.

  6. Ketil says:

    Could somebody elaborate on the term preschool, and how (whether) it differs from other kinds of daycare or kindergarten? It seems to me that preschool emphasizes academic learning. If it is true that the long term benefits are from learning social behavior, we should see similar benefits from daycare which doesn’t try to be particularly educational. I couldn’t find this explicitly explained in the article (beyond at least one study using siblings that were at home as the control group).

    • JulieK says:

      “Preschool” and “Daycare” are not strictly defined; daycare for 3- or 4-year-olds probably includes some educational activities, and preschool programming varies.

  7. Dermot Harnett says:

    The reality is also that older twin studies make assumptions about genetic (and probably environmental) architecture we now know aren’t valid. For instance, The classic twin study simply assumes that the difference in similarity between MZ and DZ twins is purely determined by the difference in relatedness between them, but this is only valid if genetic effects are additive. New studies show that even simple departures from additivity like dominance effects create problems. To the extent that epistasis and dominance matter, DZ twins will be less similiar than their 1/2 relatedness would otherwise suggest, while to the extent that environment matters, they will be more similiar. These two effects therefore cancel out in simple twin study designs. (note that the assertion is often made that almost all genetic effects are additive – this is only true in the sense that in e.g. GWAS, most allele’s effects are well captured by an additive model, but this is an artifact of low allele frequencies and how regression works, and doesn’t invalidate the above.)

    • Freddie deBoer says:

      How does any of that contradict adoption studies?

      • Dermot Harnett says:

        That specific issue doesn’t, but adoption studies are difficult and often underpowered, because adopted twins are in low supply, and they suffer from other issues (e.g. adoptions are not to randomly selected members of the population, instead are often in the same town etc., twins still share environment up until adoption)

        • Freddie deBoer says:

          And the genome-wide association studies?

          • Dermot Harnett says:

            If you mean GWAS in the strict sense, then this is for finding effects of specific SNPs. This never works that well (see scott’s omnigenic post). If you mean generally studies that use mass-genotyping to look at heritability between individuals (e.g. the one I just linked), then these are generally turning out to be compatible with the above – there is room for shared environmental effect, albeit something like half the magnitude of the additive genetic.

      • INH5 says:

        Read about the process followed by any actual adoption agency, and it’s clear that the adoption matching process isn’t even remotely random, with lots of effort put into matching children with “compatible” adoptive families. Prospective adoptive families are extensively screened with the specific intention of filtering out poor home environments. The process by which a child up for adoption is also highly non-random, and some studies have found that adopted children have symptoms of fetal alcohol exposure at rates more than 9 times higher than the general population.

        With all of these confounds, I don’t put much stock in most adoption studies.

    • lunawarrior says:

      So it has been a while since I have taken Biology or anything, but wouldn’t simple Dominant/Recessive genes actually make a pair of DZ twins MORE similar than 1/2 relatedness would indicate? I brute forced through all the posibilites on a simple pair (16 Punnett squares) and 48 out of 64 times (75%) the child would end up with the dominant gene.

      I am pulling on old High School biology though, and I know that Dominant/Recessive only rarely works so precisely, so I could totally belive this doesn’t transfer to real life. But it would surprise me a little if it were so far off.

      • Dermot Harnett says:

        So in general, the dominant phenotype is going to be the one that’s common in the population. “dominant” and “recessive” are descriptions of genetic phenotypic segregation patterns that originate from before we had molecular biology, but they usually reflect the fact that you only need one copy of a gene for it to work, so most recessive genes are rare, deleterious, broken genes (though rare and often interesting exceptions exist, e.g. signalling proteins gone hyperactive).

        Think of it this way: if your identical twin differs from the population because of two broken genes, then you do as well. But if your DZ twin has two broken genes, you must share not just one (p=0.5) but two (p=0.25) broken alleles with them before you share the effect, so even though you are half as similar genetically, you are only 25% as similar phenotypically – essentially, flip 1 and 0 in that punnet square.

        • lunawarrior says:

          Ahh, essentially the “Average Person” would express all dominant, not just because if they have both they express the dominant, but for other reasons as well. Combine this with the fact that (essentially) the differences from average are what is interesting, and for any recessive gene you are only 25% likely to share it with the DZ twin.

          Makes sense. Thanks!

        • Michael Watts says:

          So in general, the dominant phenotype is going to be the one that’s common in the population. “dominant” and “recessive” are descriptions of genetic phenotypic segregation patterns that originate from before we had molecular biology, but they usually reflect the fact that you only need one copy of a gene for it to work, so most recessive genes are rare, deleterious, broken genes

          What point are you making here? To me, this doesn’t look like information that matters, but I’m probably missing something. I’d expect the variant common in the population to be the one that was more advantageous, without much regard to whether that variant is classically dominant or recessive or other.

          What are the implications of the well-known exceptions to your rule of thumb, like number of fingers? (Having 5 fingers is classically recessive; having 6 fingers is dominant.)

    • Doug says:

      The upshot is that classical twin studies (which already estimate genetics to explain 40-50% of population variance) are significantly underestimating the magnitude of genetic heritability.

      Since twin studies typically find shared environment to make up 0-5%, most of that effect has to be upwardly biasing the importance of non-shared environment. If genetics are even more important than we thought, and total non-shared environment much less important, than our priors about the impact of targeted intervention should be even smaller than before.

  8. TomGrey says:

    There needs to be comparisons between state support for pre-school / day care vs the same state support for mothers staying at home with their own kids.

    The article says “birth to age 5, spending about $20,000 a child in today’s dollars.” So about $4k per year per child.

    Instead of Uni Basic Income, wouldn’t it be better to give each parent some $4k per year of cash to spend? For responsible parents, almost certainly the cash would be best. For irresponsible parent, perhaps only child care/pre-school would be best, where they have the opportunity to choose among many local “qualified” centers.

    • Alkatyn says:

      On a national level there’s benefits of keeping women in the workforce, therefore getting better economic growth, productivity etc. (Both the net number of workers is increasing, and they can be assigned to jobs which are their comparative advantage.) So whether or not its a net gain to a particular family if we’re setting top level policy it is probably a better option.

      • Why should I care about the nation except to the extent I care bout its people?

      • Statismagician says:

        Goodhart’s Law detected – doubling GDP and then requiring half of the new total to be spent on childcare, time-saving devices, and increased housing costs thanks to supply/demand is only a net good if you think higher GDP is a worthwhile thing outside its role as a proxy for net resources/person.

      • Michael Watts says:

        On a national level, are there any effects of discouraging women from reproducing by keeping them in the workforce? What does economic growth look like over the course of, say, four generations?

        • Statismagician says:

          This either is what’s causing the demographic transition, or would be masked by whatever is, on a very gross level.

          EDIT: I misread your question, sorry. In any case GDP is probably uniquely badly suited to addressing this, since it would by definition privilege economically-legible activity over, say, homemaking/child-rearing.

    • JulieK says:

      $4K is not going to be enough to pay for childcare (though obviously it helps).

      the average monthly cost of child care in the U.S. – including at-home and in-center care – is $1,385 a month. Nationally, the median rent payment is $1,500 a month – only $115 more than the monthly cost of child care.

      • christianschwalbach says:

        Isnt that an argument for State provided child care then? Lower costs? If money is being given to parents to fund childcare, providers will just raise costs , will they not?

        • sandoratthezoo says:

          Why would that happen? Supply for childcare is not inflexible. If everyone else in the world is squandering $4k per child per year on cocaine and fancy cars, why wouldn’t I open a preschool/daycare, squander a mere $2k per child per year, and create vacancies in other people’s preschools?

          • The Nybbler says:

            They’re not squandering it on cocaine and fancy cars; it’s going into regulatory compliance, directly or indirectly.

        • pontifex says:

          Minimum wage in San Francisco is $15 an hour. (Starting in 2022 it will be $15 an hour for the whole state). So hiring a nanny for 8 hours of childcare a day for 4 weeks a month is already $2,400, just in labor costs alone. That is ignoring profit, ignoring healthcare, ignoring everything else. So yes, $4,000 a year is about enough for less than two months of childcare.

          Isnt that an argument for State provided child care then? Lower costs?

          It’s mind-boggling that you think getting the state involved would *lower* costs. Look up how much schoolteachers in California make and what their benefits are.

  9. IrishDude says:

    All I can think of is that maybe shared environment can matter, but is so small in the grand scheme of things that it’s below the threshold where zoomed-out studies of everything can detect it. That would help reconcile the two literature bases. But it doesn’t seem right. The lead effects are huge.

    This could be analogous to minimum wages. If you’re a marginal worker, an increase in minimum wage puts you out of a job. However, only a couple percent of people earn minimum wage, so at a zoomed-out view there’s little effect on overall employment, even while for those affected the impact is huge (job versus no job).

    I’ll also increase my political support for programs like these. I think these findings make universal childcare (almost) a no-brainer.

    Your post discussed benefits, but didn’t delve into the magnitude of the costs. Shouldn’t how much programs like this cost factor into whether the program is a no-brainer? There’s many things I could spend money on that would benefit me that I choose not to pursue because the costs are higher than the benefits.

  10. HeelBearCub says:

    Studies consistently find that IQ is only responsible for about 25% of life outcomes, suggesting that education works on the other 75%.

    Is this a typo of some sort? Your steadfast conclusions don’t seem to follow from this statement. Even if it’s an inversion, your steadfast conclusions don’t seem to follow.

    • AlexanderTheGrand says:

      “Works on”, not “determines”. As in, if education doesn’t increase IQ, then it improves outcomes through non-IQ means.

    • idontknow131647093 says:

      Its either a typo or Scott picking the worst studies on the topic.

    • Eponymous says:

      He means that the correlation between g and outcome X is about 0.5 on average (so explains 25% of variance in X). Then postulates that if you measure an effect of education in X, that doesn’t mean it’s working by raising g — it could be working by changing the stuff that explains the other 75% of variation in X.

      (Of course, a decent amount of that 75% is probably just luck.)

    • cactus head says:

      I interpreted “Works on the other 75%” to mean that it explains a small fraction within that 75%, as opposed to the same small fraction within some other percentage.

  11. HeelBearCub says:

    This probably strengthens rather than weakens the Caplanian case against education, since the studies find that the educational parts of preschool are not useful, and better teachers and curricula do not affect the benefits.

    1 – I don’t think you have proven that the education isn’t useful. Consider that the education might be useful, but the difference in educational attainment isn’t distinguishable from IQ effects. The education may be part of what is generating the effect.

    2 – Note how you have immediately forgotten the earlier point you made that there is evidence that better teachers have an effect.

    • Freddie deBoer says:

      No contradiction in point 2. He explicitly says that the supposed gains from teacher quality are non-academic in nature.

      • HeelBearCub says:

        That seems to require a very stretched reading of this:
        “better teachers and curricula do not affect the benefits.”

        If better teachers affect outcomes, but not academic “achievement”, that would seem to indicate that better teachers affect the benefits.

    • Scott Alexander says:

      My evidence here is the preschool studies I discussed immediately preceding.

      • HeelBearCub says:

        I think we we would all agree that being able to read and write is useful and requires education (of some sort.)

        Now try and measure the effect of that in the US and try and separate that from the covariates. You will find it extremely difficult to do so, because literacy education is in the water. But it would be a mistake to conclude that the education in literacy isn’t useful.

        Thus, I think it is a mistake to say that studies of pre-school education support the premise that education is not useful. They don’t attempt to measure this.

        • Watchman says:

          What is this education that is not measured? How is it to be measured. It seems to me education is an abstract value, with the only attempted assessment methods, qualifications, not being independent of this abstract nature but instead measuring achievement of whatever criteria within the abstract system that is education that are socially and politically desirable.

          You need to define education outside of the prism of its self-justifying qualification structure to have an argument that education is useful. To look at your analogy the literacy rate is a measurable thing, which can be counted, which is maybe related to education but is not a direct dependent: some people learn to read outside of education. You need to produce some way of measuring education itself or refocus your discussion on an output as Scott has done, his being a basket of measures about life outcomes.

          • HeelBearCub says:

            some people learn to read outside of education

            I can’t tell if this is a willful mis-reading of what I wrote or not.

            People don’t learn to read without being taught to read. If you misunderstood that this is what I meant, then you aren’t understanding the thrust of my comment.

          • People don’t learn to read without being taught to read.

            That is not the case, unless you count “taught themselves” as being taught.

            My wife taught our daughter, who was five, to read. Her younger brother observed the process and taught himself, with no input from us.

            Learning a language is surely harder than learning to read. But it seems to be a common pattern for the children of monolingual immigrants to become fluent in the local language, not because someone is teaching it to them but because they have learned it by interacting with people who speak it.

            My uncle was about ten when the family immigrated to the U.S. from western Europe. He apparently learned English initially from sailors on the boat.

          • acymetric says:

            It seems like this is a semantic argument about the word teach. I would suggest that you did (inadvertently) teach your son to read while you were actively teaching your daughter.

            I would suggest spoken language is quite a bit different from reading in terms of how it can be learned, regardless of which is more difficult. In your father’s case, he was “taught” by sailors (both passively and I suspect to some extent actively), same for children of immigrants and the local language speaking people in their community.

          • The claim I was responding to was:

            People don’t learn to read without being taught to read.

            If you define “being taught to” so broadly that it includes a child’s casual conversation with native speakers or a child observing that a sibling is being taught and teaching himself, the statement reduces to “People don’t learn to read without learning to read,” which is not very interesting.

          • Igon Value says:

            I would even say that the difficulty of learning a language is nil for a kid since all kids learn at least one language without having to be taught.

            Not all kids learn how to read so I conclude that learning how to read is more difficult than learning how to speak.

            Speaking is an inherent aspect of the human species, writing/reading is not.

          • Not all kids learn how to read so I conclude that learning how to read is more difficult than learning how to speak.

            I don’t think that follows. Kids have much more reason to learn to speak than to learn to read.

          • Igon Value says:

            Kids have much more reason to learn to speak than to learn to read.

            When I am so motivated to learn some skill X that I do it naturally without thinking about it at all, I don’t call it “difficult”. Maybe the intense motivation is part of why it is not difficult. And if everybody else is equally motivated and acquires skill X with no need for coercion or nudging or even guidance, yeah I conclude that X is easily acquired. If it were difficult, some people couldn’t do it on their own.

            I suppose it depends on what you mean by “difficult”. You could argue that the motivation is innate but the learning process objectively difficult (for example it requires large amounts of energy or time or whatever your metric for “difficulty” is).

  12. edgepatrol says:

    I’ve always thought that the point of preschool (and maybe also kindergarten) wasn’t to improve intellectual skills, but to ease you into the formal learning environment. If you jumped straight to first grade, you’d be shell-shocked. That would inhibit your learning; you’d be distracted by all of the social stuff, and you wouldn’t know the format or expectations. (?)

    • Mark V Anderson says:

      Yes this is a very good point. I don’t know a lot about Head Start, but my kids had a public school program called Hi5 in their year before Kindergarten. My impression was that it wasn’t very academically oriented at all. It was mostly play based. But the key learning experiences was getting along with other kids and being comfortable in a school environment. So at least in my experience, they aren’t teaching academics so much as social behavior. That way, when the kid does start Kindergarten or 1st Grade they aren’t distracted by the new social environment, and can immediately engage in the academics then. That is my understanding of how most educators view pre-K, and thus probably Head Start, but I may be wrong. Maybe they do teach them their letters and numbers or something.

      Of course, if this socialization really does add to their academic learning in 1st Grade, I would think this would up their IQ a bit. But, as Scott says, the evidence doesn’t support this. I suppose just decreasing the stress these kids have when they start school proper might have some benefits down the road.

      Although I’d really like to see substantial evidence for this. Of course I haven’t read Scott’s links yet, so I am kind of talking up my butt. For others that have read these, how many studies are we talking about to show benefits, even if not IQ benefits? I don’t believe any new finding in social science until there have been several confirming studies.

  13. Andrew Hunter says:

    This has been carefully designed to maximize the probability of obtaining a p-hacked result. We tried something, it didn’t work, we measured it again, it didn’t work, we looked at a new endpoint, that didn’t help, and the fourth one stayed up and now we found some way to measure a good outcome. As measured by people deeply committed to spending ungodly amounts of money on the policy.

    I really think you’re over-updating.

    • wysinwygymmv says:

      Would you have raised this objection if the result was consistent with your biases? Be honest now…

      (Obviously, I mean in a counterfactual where most of the evidence was against your biases heretofore. I don’t really agree that most of the evidence flatters your biases in this case — I think you’re filtering — but I’m asking if you’re really as willing to relinquish your biases as you think your ideological opponents should be. I’m skeptical.)

      • Alex Zavoluk says:

        That doesn’t make any sense. Their critique only applies to positive results. You almost never p-hack until you get a null result; you p-hack until you get a significant result.

    • albatross11 says:


      It’s possible this research really did happen upon a completely different positive benefit than the one they originally expected from these programs, but I think there are probably about a zillion ways to measure benefit, and my guess is that you could look at any of these programs and find equally headline-worthy bad effects of preschool if that was what you were looking for.

    • matthewravery says:

      I’m confused. Are you referring to a specific study that you think has been p-hacked? The Vox article talks about a ton of different studies, and it wasn’t clear to me that any of them had been walking down the garden of forking paths.

      P-hacking is an accusation you can lob at a paper on an individual level, but claiming that the whole literature is p-hacked is pretty wild if you don’t provide any evidence beyond, “Well, it sure sounds to me like they could’ve!”

      Regardless, claiming that “Kids who are socialized early tend to act more pro-social later in life” doesn’t seem like a very wild claim. In fact, it’s where my prior would’ve been anyways.

    • Scott Alexander says:

      Did you read the studies in the article? There are lots of them and they’re pretty good. This goes against what I was expecting, and I tried pretty hard to see if there was any way to dismiss them, and I couldn’t.

      The first time you get a study like this, yeah, it’s probably p-hacked. After that when people specifically start looking into it because of previous results, you can’t say that anymore.

      • j1000000 says:

        Wouldn’t the p-hacking be happening outside of the written up results of the studies? This data takes decades to collect — wouldn’t we need Arthur Reynolds to preregister a hypothesis that pre-school graduates would be arrested less often?

        • idontknow131647093 says:

          Exactly. They took non-randomized trials (already fishy to the max) and applied retrospective hypothesis. In addition this result is well in line with what academics line up with politically.

          It is easy to show and example and see how rigorous you have to be to have a chance if you go against the grain: The Seattle minimum wage law. University of Washington pre-registered their proposals and methods, and their working paper showed that the minimum wage law was likely having negative effects. Within a few days of the working paper Berkley had a paper that came out criticizing it saying that UW should have used OTHER methods and variables, even though UW got its initial choices peer reviewed and approved ahead of time.

          If I combed through all the available pre-K studies and found a half dozen that showed pre-K increased some variables, like criminality or obesity such a meta study would never be published. Or, even if it was, there would be a rebuttal co-published.

      • David Conroy says:

        Talking about the potential of education for significantly increasing academic achievement and even significantly increasing IQ scores (at least temporarily for children), without mentioning — or possibly being unaware of — Project Follow Through and Siegfried Engelman’s instructional methods, is like talking about the futility of developing an atomic bomb because of the failure of the German programme without having heard of the Manhattan Project. It’s criminally under-reported on both sides of the political divide, and nobody anwhere ever should make sweeping conclusions about “what the evidence says” without having critically engaged with the voluninous and impressive literature around Direct Instruction.

        2 very good starting points are:
        1) A 50 years of research meta-analysis
        2) His magnum opus delineating his theories (right up your alley I would imagine) He later realised for instance that he had almost exactly reformulated John Stuart Mill’s 5 methods of induction.

      • Doug says:

        > This goes against what I was expecting, and I tried pretty hard to see if there was any way to dismiss them, and I couldn’t.

        CPC Longitudinal – CPC eligibility requires “parents agree to participate regularly.” Massive selection bias. This is not randomized or controlled for whatsoever.

        Abecedarian – The initial group was randomized. But the Abecedarian program had five times the dropout rate as the control group. The dropouts were not tracked, and the results don’t include this. In addition the kids who were selected to replace the dropouts were not randomized, and they are included in the dataset. Massive selection bias.

        Brookings Sibling – As others have pointed out, siblings aren’t a randomized control group. Any parent can give you a good estimate of their kids relative intelligence. Also parents are going to be much more likely to put brighter siblings in more intense education programs. Brookings makes no attempt to quantify or even address this issue. Massive selection bias.

        NBER Regression Discontinuity – This study uses kernel regression to estimate impact of Head Start. But deep in the paper the authors admit they selected the hyper parameters for the kernel bandwidth arbitrarily. They admit that using a formalized, unbiased cross-validation method would have selected very different hyperparameters. When the unbiased hyperparameter used, the results are no longer statistically significant. Reeks of p-hacking.

        • Aapje says:


          I updated to being more uncertain than my previous conclusion that pre-school programs are useless, but I’m not convinced yet.

  14. Alkatyn says:

    I would lean towards the idea that socialization makes a big difference to children’s outcomes. With the caveat that this is all anecdotal and idle theorizing, my experience teaching* has been that consistently the kids who are the most difficult to deal with are the ones who spend the most time with their parents rather than other adults.

    My feeling is that because parents naturally prioritize their child and judge them more positively in everything they do than an impartial observer. Which isn’t inherently bad, but means that kids are less prepared for circumstances where they’re not “special”, and not the main focus of attention, so find it harder to operate in a classroom setting with impersonal rules and a need to share resources and the teachers time with other students. I’d guess this also generalizes to other outcomes like behavior in the workplace.

    This isn’t quite the same as the conventional idea of a “spoiled” child, which might be the extreme of this issue. As these kids can be perfectly nice and intelligent but have trouble with the idea of being an equal member of a group of peers.

    * My experience has generally been in the 12-18 range and high SES, so may not generalize.

    • christianschwalbach says:

      Do you happen to have worked in a higher-income/wealth area? I grew up in such a neighborhood, but finished my high school years in a lower income region (but low poverty rates) , and from my admittedly juvenile impression, what you stated was true in my Middle School and 1st year of High School, but less relevant my last 3 years of HS, where the “problem kids” were the ones with the least parental influence and support. (Granted, my 2nd High School had a strong pipeline to the local alternative school, so kids that were acting up or having trouble at mainstream HS were shunted rather quickly to the Alternative School). At my first High School, there certainly was an aura of academic competition and arrogance among some students and family, so perhaps that fits well with your experience.

  15. Zephalinda says:

    Yeah, in discussions like these, it’s worth remembering Freddie DeBoer’s “Selection Bias is the Most Powerful Force in Education”. Given that the methodological norms in education research seem uniformly bad, it makes no sense to string together a broad argument from top-level conclusions of studies (as Vox seemingly does) without first doing a really thorough review of all the actual data-gathering and analysis methods involved.

    The very first link I clicked– Vox/Piper’s support for the splashy statement “Kids who enter intensive preschool programs are less likely to be arrested, more likely to graduate, and less likely to struggle with substance abuse as adults”– was to an interview (!!) about this study ostensibly finding better outcomes from full-day vs. half-day preschool. But looking at the study reveals that the kids were not assigned randomly to one or the other type of preschool; instead, the full-day group was deliberately selected on the basis of factors including “parental employment and education”. (Analysis claimed good matching of parental employment between the two groups, but did not differentiate between part- and full-time employment). And surprise, surprise, children of parents with full-time educational or job commitments turned out to have better outcomes at the end of the study than kids whose parents were free to pick them up in the middle of the day.

    I’d like to give Scott the benefit of the doubt for doing a more critical review of the literature than that, but this subject is really more suited to a “much more than you wanted to know”-style deep dive than to a breezy half-page hot take.

  16. DeservingPorcupine says:

    How about instead of funding a massive entitlement so that a few kids whose lives are so bad that a few hours out of the home provides them with measurable lifelong benefits we just take kids away from those parents?

    • broblawsky says:

      That’s the kind of policy that definitely never gets applied disproportionately to people of disenfranchised ethnic backgrounds. Yep, definitely a concept that has never been tried before at the cost of thousands of children being ripped away from their parents in a quiet attempt at genocide.

    • wysinwygymmv says:

      We already do that. Are you saying we need a special program that goes around spying on families instead of waiting for reports of abuse? Or do you think that the process for deciding who gets to keep their kids is too generous to bad parents? What should that process look like?

      • DeservingPorcupine says:

        I’m saying it doesn’t seem like we do it well enough if this part of Scott’s hypothesis is true. Again, if home life is so bad that merely removing these kids from it for 8 hours per day and putting them anywhere that passes the “first-world sniff test” measurably improves their long-run life outcomes, then I’m inclined to believe the parents are garbage. I don’t think it’s a matter of letting them have too much screen time or giving them too few chores. I’d readily wager it’s more of the lock-them-in-a-dungeon variety.

        • JulieK says:

          A better explanation is not that the home life is bad, but that once the child is in daycare, the parent can earn more money, so the household is materially better off.

        • albatross11 says:

          I seem to recall that one of the early justifications for universal public education was the need to basically do this–to take the kids from immigrant familes (who often really were backward and poor, since they were often subsistence farmers who came here to avoid starving) and put them in the presence of properly-behaving adult role models who’d impart the right culture and values along with teaching them to read. I’m not sure whether this explains how things worked out w.r.t. our massive waves of immigrants, but I’ll note that we really did take in a lot of immigrants and they really did seem to assimilate pretty well.

          It’s not so obvious to me, though, why preschool will do this if 13 years of mandatory public education won’t. Is there really an extra benefit of aculturating those kids at age 3 instead of age 5? Or are we just trying to get them out of the hands of their meth-head mom and her current dirtbag boyfriend for a few hours a day, a little earlier?

          • idontknow131647093 says:

            Or are we just trying to get them out of the hands of their meth-head mom and her current dirtbag boyfriend for a few hours a day, a little earlier?

            I’d expect that if there is a good effect, it would be from this, but we should see it in basically every metric, not a few cherrypicked ones.

    • JulieK says:

      And then what? The foster-care system is already overloaded.

      • toastengineer says:

        Having been in it, it’s also pretty terrible, for a variety of reasons. You’d be taking kids out of “so terrible” homes and putting them in new “so terrible” homes, or at best orphanage level care that happens to take place in someone’s house.

    • Leonard says:

      How about instead of funding a massive entitlement so that a few kids whose lives are so bad that a few hours out of the home provides them with measurable lifelong benefits we just target spending on those few kids to get them a few hours out of the home, but mostly just leaving them with their parents, who probably kind of suck but are still probably the best guardians we can get for them?

      • DeservingPorcupine says:

        But if Scott’s hypothesis is correct, they’re clearly not the best guardians we can get for them, as putting them in a loosely-structured, not-horrible classroom with some other kids and letting them mill about for half the day apparently improves their lives.

    • Statismagician says:

      Because that’s massively more expensive (cf prisons vs. community service programs), not to mention massively more cruel and absuable, than just scrapping the ineffective academic bits of these programs and replacing them with the medical/family-support bits that work quite well?

      • Statismagician says:

        Also, and I can’t believe this didn’t occur to me before, because we already do; what do you think Child Protective Services is? These interventions are aimed at people who are below the level of awfulness where society decides you aren’t competent to raise children, but also below the level of resources where society thinks you should be able to manage jsut fine on your own.

        • There was recently a series of local news stories about a woman said to have kidnapped her own children from, as best I could tell, the local child protective service. She had visitation rights and had walked out with them.

          I eventually found one story that mentioned why her children had been taken away from her–because they were living in a car. That’s evidence of poverty, not of awful parenting.

          I’ve mentioned the Texas FLDS case before. That was one where three hundred children, ranging from infants up to adults who were claimed to be minors by authorities who refused to accept documentary evidence of age, were seized because the authorities didn’t like their religion. They were returned only after unanimous decisions by first the state appeals court and then the state Supreme Court holding that the authorities had no legal basis for holding them.

          So no, I don’t assume that the reason child protective services seizes children is because their parents are awful.

          but also below the level of resources where society thinks you should be able to manage jsut fine on your own.

          A poor American, such as the mother living in her car, is substantially richer than the average human being through most of history–quite a lot of whom managed to bring up children. Sympathy for the fact that she is much poorer than the rest of us is an argument for giving her money, not for taking her children away.

          • Statismagician says:

            Yes, I’m aware of how horrible the system often is in practice, and of how far from stated principles actual decisions can be. I’d thought the combination of saying that doing this would be cruel and abusable, and then that we currently do it based on social conventions (as opposed to e.g. outcomes research) would resolve to this; sorry, I should have spelled it out.

          • Doctor Mist says:

            Hmm. I’m usually pretty sympathetic to the argument by comparison to historical lifestyles, but something about this situation makes me reluctant.

            I think about how likely those car-living kids are to be victimized in innumerable ways (not by the mother, perhaps, but by the rest of the street-living populace, and about what it will be like for them to grow up knowing how thoroughly one-down they are.

            I can agree, at least for the sake of argument, that the mother does not deserve to have her children taken from her. But I’m far less sure that the children deserve to be left in that sorry situation because of that.

            Maybe I am too optimistic about how life under CPS would be. Crud.

          • Statismagician says:

            @ Doctor Mist – yes, you are too optimistic; examples are trivially available through Google.

          • albatross11 says:

            The tool needed to address the mom raising her kids in a car is subsidized housing. The tool available is taking away her kids till she gets a better place to live.

          • The tool available is taking away her kids till she gets a better place to live.

            The tool needed for dying of old age is to discover a way of preventing aging. The tool available is suicide while young.

            Taking away her kids makes things worse for them and worse for her. It gives her an incentive to find better housing, but she had that already.

          • albatross11 says:

            I’m not disagreeing, I’m noting that when your only tool is a hammer, lots of people get hammered on.

        • DeservingPorcupine says:

          Again, we already do it, but we’re apparently not doing it well enough if there are still kids that can be identified by this study who benefit that much merely from being outside of their crappy house.

          Note that I’m predicating this on the hypothesis being true. If all the benefits come from, say, kids getting immunizations regularly, then what I’m saying doesn’t apply. Of course, if that’s the case, that means we’re also totally unjustified in establishing universal pre-K or whatever.

    • benwave says:

      I think you might be overestimating the cost or providing universal childcare. Parents already paying for childcare themselves will start paying instead through their taxes, the only increase in money being spent is from parents not already paying for childcare, which is precisely what the policy would be trying to achieve.

      Yes, some of the cost will end up also being borne by taxpayers who aren’t parents. But that’s not an argument that the cost itself is very large, that’s a different argument.

  17. Walter says:

    I’ve always felt that pre-K (and, really, kindergarten too) is just babysitting. I feel like this kind of confirms it? That is, people don’t put their kids there so that they learn anything in particular, it is just about getting the parents some room to breathe.

    • matthewravery says:

      Scott leans hard on the “parents having time to do stuff like have a job” mechanism, but one around kids learning to socialize and have more attentive care could also work. Note that this is also consistent with the “pre-K (and really everything through HS) is just babysitting” hypothesis, it just goes a bit further.

    • TDB says:

      Elementary and even high school retains a large element of babysitting. They have to be there, but they can’t be forced to learn anything.

  18. aristides says:

    If you truly believe the Caplanian case against education, these studies provide evidence to not provide universal child care or preschool. Staying in school longer should not be a positive goal in itself, and actually lowers years participating in the workforce while making it harder to compete for jobs. Getting better jobs can easily be explained by the children spending more time in school and signalling being a better worker, without actually being a better worker. That would mean that if preschool was universalized, the gains would disappear as everyone would spend more time in school. Since they have better jobs, they would require less welfare. The only thing that might not go away upon universalization, is the decrease crime rates, but is universal preschool a cost effective way to reduce crime rates?

    • herbert herberson says:

      Are you concerned that 3-5 year olds in preschool could instead be participating in the workforce?

      • aristides says:

        No, sorry that I wasn’t clear. The proposed benefits of preschool included higher high school graduation rate, and more years spent in college. I and Caplan believe those years between 16-25 could be better spent in the workforce.

        As a personal aside, I do worry that 3-5 year olds would be better off with their parents than in preschool, and parents better off raising kids than participating in the workforce, but I do not think the data actually supports that view on a large scale. I have a personal bias as man with a wife that wants to stay at home homeschooling our children. I probably agree with Scott that my ideal proposal is something like a UBI, so that parents can use their best judgement on how to spend the money on the kids, but I disagree with him that Universal child care is an acceptable alternative in the meantime.

    • Scott Alexander says:

      Staying in school is basically the marshmallow test for teenagers. It’s not inherently useful in and of itself to sit around waiting for weirdos to give you a marshmallow, but it indicates other good qualities, and in the end you get two marshmallows.

      • Bugmaster says:

        Technically you’re right, but IMO it’s very useful for teenagers to be able to pass the GED (or whatever standardized test indicates basic literacy/numeracy). For most teenagers, staying in school is probably the easiest way to ensure one passes the GED (i.e., that one is able to read and count). Obviously, for some teenagers, homeschooling can accomplish the job, but I think such cases are in the minority.

      • aristides says:

        So my argument was that preschool only made it easier to send the signal, without improving the underlying person, while your argument is that preschool actually made them a better person, and the signal of years spent in school accurately signaled the better qualities? That would explain the decrease crime rate as well. In that case your argument makes more sense, though I would like a study that checks that the former preschoolers actually have the better qualities, not just have the signals of people that have good qualities.

        I am still not certain it is utilitarian/cost effective compared to the status quo, but I agree the preliminary evidence suggests it is. I have a personal bias against anything that furthers the signal credentialing race or keeps children in classrooms rather than playing freely, but I can at least imagine a program that kept cost low and allowed mostly play time to be effective policy. I remain sceptical that that is what will actually be passed.

  19. RC-cola-and-a-moon-pie says:

    Only thing that could have made this article even better would have been some traditional Scott Alexander discussion of the underlying evidence for these things so that the reader can assess the strength of the evidence for these alternative effects. Granted I can click through to Vox, but, without offense to Piper, Scott has earned a lot more goodwill from me in trusting assessments of scientific studies. Part of the reason I’m a little uneasy is some of the language in the paragraph (the third paragraph of the essay) discussing results. We have a “seems” in there and a reference to “subtler” effects, which makes me want particularly clear evidence of them. I assume there are no issues with the sorts of statistical problems we’ve all learned to worry about with post hoc derivation of effects from studies looking for other things, etc., but it would be nice to nail this stuff down a bit more before moving on to the lessons we draw from the studies.

  20. sentientbeings says:

    I’ll also increase my political support for programs like these. I think these findings make universal childcare (almost) a no-brainer.

    Government programs, subsidies, and especially “universal” programs cause significant changes to incentives and selection mechanisms for their targets. Expanding availability of childcare through indirect means could adversely affect the quality of care, which could neutralize the positive long-term effects. My expectation would that it would raise the average costs involved, and those less able to afford it – presumably the people the program is intended to help most – would receive the lowest quality product, potentially lower than before. Even neglecting the direct costs, the uncertainty level seems pretty large.

    • Nicholas Weininger says:

      Yeah, and also it’s disturbingly sloppy to see a “no-brainer” conclusion from Scott without any sort of cost benefit analysis: this is somewhat like saying “this investment actually gives you some money back rather than just burning it all, therefore it is obviously a great investment.”

  21. JohnBuridan says:

    I think the inefficiency angle on universal childcare is very important. Many mothers want to be able to take care of their children. Often they raise their kids with other moms, i.e. their friends who are having babies at the same time. If we give all mothers a tax credit or simply pay mothers a certain number of dollars to defray the cost of early childhood expenditures, we can find “nudgier” ways of getting children into daycare without causing those who don’t want government daycare to get no benefit from the program.

    I often oppose “early childhood education” measures, because I think (nearly) all moms and children should be able to share the costs and benefits of the childhood intervention. I would rather incentivize communities to form their own formal and informal programs.

    • Tenacious D says:

      “Early childhood education” also adds to the occupational licensing burden on the working class. There aren’t really economies of scale for taking care of children under 5, since you need around 1 adult per 3 children, so professionalizing childcare at that age doesn’t add efficiency compared to other programs. So parents in the working class are paying as much or more as they would for community-level formal or informal programs and also face more of a hurdle to find employment in childcare–getting squeezed from both ends. Whereas the upper-middle class can hire nannies/au pairs and thus avoid the system.

      • baconbits9 says:

        A 1:3 adult to child ratio would be brutal for filling child care provider spots. The natural route to take would be to have mothers (or fathers) care for other kids as well as their own, and that tight of a ratio would prohibit many of them, 100% of people with 3 small kids, and most people with 2, from getting into this line of work officially.

        • Randy M says:

          If you have a few older kids in the mix, I think you could stretch it a bit, especially since “under 5” covers a wide range of capabilities, with the younger ages needing near constant attention when awake, to the older children probably being okay with an adult in the room but not attending to them personally most of the time.

          • baconbits9 says:

            I would say that the younger ages need near constant proximity when awake, not near constant attention.

            The issue with the 3:1 ratio is basically that a parent with a few small kids who wants to pick up child care work can’t join a certified child care provider. In one world the parent could take her kids to child care with them, and then provide care for an extra few kids. With a 3:1 ratio you are only netting out 1 kid for a mom with 2 kids, meaning that one kid has to effectively pay for the wages + rent + insurance for the 3 kids for the mom to make anything.

          • Tenacious D says:

            Yeah, I did a search to check and I guess I was too stringent on the ratio I assumed. 1:8 or 1:10 seem to be used as minimums in some jurisdictions.

      • JohnBuridan says:

        Depending on age spread of the children under 5, 1:5 could be fine….
        actually now that I think about it, children are so high variance that I am unwilling to put a rule on it.

        Sometimes one adult for three 18 month year olds is fine, other times it is absolutely impossible. It can depend on the day, the time of day, or the dispositions of the hellionslittle cherubim.

  22. idontknow131647093 says:

    Counterpoint. You weren’t wrong. And even if you were, preschool is too expensive to justify these tiny gains that have not been replicated (and probably will not be in a robust randomized study).

    Rather, because of the massive institutional bias in favor of schooling, it was inevitable that eventually one or more researchers would find a way of jiggering the stats to find some set of benefits (just as happened with Medicaid following the very negative Oregon study).

    • Kris says:

      Exactly! Let’s be good bayesians here. Even if there is no effect, there is a very high chance that someone ‘finds’ something. The fact that someone did shouldn’t count as too much evidence.

    • Scott Alexander says:

      Read the studies involved. Kelsey and I were both expecting (possibly hoping?) that would be the angle, and we both decided the evidence wasn’t consistent with that.

      • idontknow131647093 says:

        I did. Using CPC was always going to stack the deck in one direction because participation is based on your neighborhood school offering it or not. As with all non-pure-lottery allocations this means that children receiving it will be the offspring of more highly invested and likely higher intelligence parents.

      • Nornagest says:

        Some of the effects these studies find are dramatic enough that they don’t pass the giggle test for me. 48% higher college graduation? That’s in the ballpark of what you’d find by comparing the top and bottom income quartiles, which doesn’t make sense since the mechanism they’re proposing for it is that it frees up family resources. We’re in One Weird Trick territory here; either there’s something really spooky going on, or the study’s wrong.

        • HeelBearCub says:

          Mmmm, what are college graduation rates for people eligible for free pre-school? Pretty damn low.

          Youth voter rates this election are up 500% or 600% in some states. Is that unbelievable? Not at all. Because their participation was so abysmal.

          • Randy M says:

            Is that the rate of graduation of those that attend, or the rate of graduation of the entire cohort?

            Edit: It looks like the relevant stat is this:

            preschool participants had higher rates of postsecondary degree completion, including associate’s degree or higher (15.7% vs 10.7%

            That’s from a study looking at 1539 participants, so assuming roughly equal control and intervention groups, about 38-39 of 770 additional participants achieved some college degree thanks to preschool (or some other confounding variable).

            Does this sound absurd? I don’t know. Perhaps if you calculate the mother’s expected earnings from those 2-4 years, and assume that low income households are good at saving for college, perhaps that made the difference in being able to attend a college. I can recall times when my attendance at university seemed to hinge on my working class parents scrounging up a couple grand.

  23. Quixote says:

    I suspect that, one of these days, someone is going to find a major methodological problem with the way twin studies are done and the whole set of results on shared environment will go through as a crisis close in magnitude to the one about psychology of priming.

    Things that predispose me toward the above:

    1) These studies almost always give counterintuitive results. Lots of counter intuitive results are true; but most are false. The pattern of “counter intuitive result grabs headlines and thought-space” and then years later crashes and burns and it turns out the obviously true stuff was actually true is pretty common.

    2) Twin studies and genetics vs shared environment studies often tend to produce different results from other studies using other methods that study the same phenomenon from a different angle. Frequent collisions and disagreements can be a sign that something is wrong. (of course, the something could be in the other methods)

    3) This is the most important one and I probably should have listed it first. These studies on shared environment seem to always find the same magnitude of effect and the same split between genetics and shared environment no matter what specific aspect is being studied. When a method always gives the same result no matter what phenomenon you point it at, then it is a sign the results you see come from an aspect of the method rather than the phenomenon.

    • Kris says:

      1. It’s counter-intuitive relative to what you want to believe, not relative to theory. An alien who learned the theory of natural selection, or studied some animal species, would presume that important traits are highly heritable in humans as well.

      2. They only produce different results relative to studies that don’t control for genetics. Studies that do (twin studies, adoption studies, lottery studies) agree.

      3. This is just plain false. Height or IQ are more heritable than, say introversion.

      • INH5 says:

        Honest question: do lottery studies control for the fact that lottery winners are disproportionately likely to be people that buy a lot of lottery tickets?

        • Michael Watts says:

          do lottery studies control for the fact that lottery winners are disproportionately likely to be people that buy a lot of lottery tickets?

          For things like the Georgia land lottery (and its associated studies), yes. Specifically, most people entered, and it wasn’t possible to enter more than once.

          • INH5 says:

            This is the paper that studied the intergenerational outcomes of the Georgie Land Lottery. A key quote from the abstract:

            Following up two decades after the lottery, we find that winners had more children than non- winners, but did not send them to school more. Thus, the response to wealth was more on the side of “quantity” than “quality.”

            So apparently, if you give out free land to pre-industrial farmers, they’re far more likely to pop out some more kids to help them work the additional land than to sell the land and use the money to send the same number of kids that they would have had anyway to school. Very interesting, and potentially quite relevant to issues of development in economies that are still primarily agricultural. But I don’t think it’s terribly relevant to modern first world countries.

      • Quixote says:

        Hmm. Some good points.
        See below comment (a few down) on 1.
        Good point on 2, there is a lot of agreement among these and that agreement would be a point towards accepting the method.
        On 3, I was a little imprecise and you’re correct that its not “all” the same. But I do think there are more things that work out to be ~50% genetic, ~50% non shared environment, 0-X% within rounding shared environment than one would expect in advance.

    • Freddie deBoer says:

      You think it’s counterintuitive to believe that our brains, the product of our genetic endowment, are shaped by genetic factors? That seems… odd.

      • Quixote says:

        This is not a good faith response and I reported it as such. No where do I deny that brains are genetically influenced, nor would I. You are just being snippy to be snippy on the internet and are not really engaging with points.

        • AliceToBob says:

          Well, your original post was vague on what you found counterintuitive (I think you’ve fleshed it out more below). I wouldn’t judge others so harshly for attempting to fill in the gaps.

          I’m interested to know the value of X in your point 3. And for others that know about this (perhaps Kris), is the claim true?

    • Statismagician says:

      I’m worried you may be conflating ‘counter-intuitive’ with ‘philosophically inconvenient’ in your 1), could you please expand?

      • Quixote says:

        I suspect you are not really engaging and are just being snippy; but I can’t remember any history of bad action from your user name, so I’ll respond as if you were trying to contribute to the conversation in good faith.

        The claim that child rearing has minimal impact on child outcomes is highly counter intuitive. To my knowledge, every major cultural tradition, from Judaism, to western European christian traditions, to the american puritans, to Chinese culture, to Hindu Indian culture, to Arabic Muslim culture, to Japanese culture all spend a decent amount of cultural effort detailing how to raise kids and talking about the importance of doing it correctly. This is not ‘philosophically inconvenient’ for me personally, this is downright odd. For a huge number of cultures to all get something this fundamental wrong would be a big deal. E.g. surprising. E.g. counter intuitive.

        • Statismagician says:

          I apologize if that came off as snippy; I really was hoping to clarify what the findings in question were. My initial thought was that you were implicitly making a strong ‘IQ/life outcomes do not significantly depend on genetics’ claim, which I think based on the above isn’t true?

          I don’t necessarily see a conflict between ‘good child-rearing (however defined) is very important for social outcomes (however defined) and development of innate potential (however defined)’ and ‘…but innate potential is also very, and plausibly more important.’ I think of, and think the literature confirms something akin to, genetics as setting the middle of a boxplot and environment as setting the tails; the various twin studies seem to me to be saying that the between-group variation is larger than the within-group variation.

          I could certainly be wrong on this; I’m not a biologist and I don’t have a perfect audit trail for all the things my brain flags as ‘read this somewhere, found convincing.’ If you’ve got a different view, or if I’m completely off-base, I’d love to see contrary data.

          • edmundgennings says:

            Quixote, that point is one of the stronger arguments I have encountered against weak biodeterminism and this is the first time I have heard it(so congratulations for coming up with it). Also given societal selective pressures it seems there would be a natural selection for societies to hold if not true at least adaptive beliefs.
            Here are my initial responses to this argument. None of which satisfactorily answer the argument but together they might.
            Parenting did matter more in the past. When one is on mathusian constraints, parental contribution does matter a lot more than in modern settings at least for survival of the child. We do know the wealth of parents did impact rates of survival of their children. Also we have massively less relevant nepotism than some other settings.
            Secondly one could imagine it being adaptive for parents to overestimate their influence so that they are more likely to do the things that do matter for keeping the kid alive. Even today, i would suspect being raised by biological parents is important for not dying as a two year old than income as a forty year old but the father is more likely to make the commitment if he thinks the long term flourishing of his child is at stake than a small change in risk of dying.
            Thirdly at least some of that is moral-religious formation which does work better than try to alter someone the stuff secular social scientists care about.
            Fourthly, I am not sure weather the premodern cultures that I am familiar with actually engaged in the costly forms of parenting that a moderate biodeterminist would view as wasteful.

        • AliceToBob says:

          To my knowledge, every major cultural tradition … all spend a decent amount of cultural effort detailing how to raise kids and talking about the importance of doing it correctly.

          If there was some consensus on how to raise kids, I’d agree that this is counterintuitive. But do they all go about it the same way? Seems unlikely.

          Without getting CW-ish, every major culture has ideas about the afterlife. They tend to differ from each other significantly in places. Do you find it counterintuitive to think some/most/all of them are wrong?

          • Quixote says:

            That’s an fair point, but I feel like there is a relevant difference between, say, views on the afterlife and food laws. The afterlife is totally unobservable. If people eat under-cooked pork, they get trichinosis and are very sick; that’s highly observable and people can notice the correlation, even if its not perfect. I would have thought child rearing is more like food law than the afterlife, but possibly my intuition on that is wrong.

          • INH5 says:

            An idea being common isn’t necessarily evidence that it is true, but it is evidence that having such an idea is useful or adaptive in a large number of environments. Even if an afterlife doesn’t exist, it’s easy to think of ways that a widespread belief in an afterlife could be beneficial. It’s harder to think of ways that strict cultural mores about child rearing could be beneficial if child rearing didn’t actually matter that much.

            As for the lack of a consensus, different child rearing practices could be explained as adaptations to different environmental conditions. And that could also play a part in beliefs about the afterlife. Cultures that lived in the desert tended to think of hell as a place of endless fire, while the Norse thought of the underworld as very cold. Buddhism, originating in a tropical region that is within spitting distance of some very cold mountains, has both hells that are very hot and hells that are very cold.

          • albatross11 says:

            My take on the lack of much environmental effect in adoptions studies has always been that, as a society, we’ve done a pretty good job of getting almost everyone the sources of positive environmental effects (sufficient nutrition including micronutrients, schooling) and blocking most of the negative environmental effects (vaccines, good sewers, getting rid of lead paint and lead in gasoline, laws against abuse and neglect). So what we’re seeing is evidence that we’ve done such a good job there that most of the remaining effects we’re seeing are either random stuff outside the parents’ control, or they’re genetic.

            [ETA] That’s consistent with a world where we have an intuition that childhood environment is very important (it was until the last couple generations in first world countries), and also consistent with what I understand of the Flynn effect (rising raw IQ scores).

          • INH5 says:

            @albatross11: I don’t think that’s very consistent with the fact that, at least in the United States, the Flynn Effect has continued at a steady rate of ~0.3 points per year until at least 2014. Everything that you mention was already solidly in place in America by 1994, but something changed over the next 20 years to further increase average IQ scores by ~6 points.

          • albatross11 says:

            Damn, another beautiful theory beaten up by a nasty gang of facts.

          • Faza (TCM) says:

            The continuation of the Flynn effect post 1994 can be explained if we assume that the development of individual IQ is a satisficing, rather than maximising, process.

            The distinction is relevant, because if we are satisficing, we will generally only see the minimum level necessary and attainable. If the attainable level is lower than the necessary, IQ will cap out at the attainable level. If it is greater, it will cap out on the necessary level (again, generally speaking).

            The environmental considerations pointed out by albatross11 address the attainable level. However, it is possible that the necessary level has continued to increase, while still being below attainable level. This would lead to the manifestation of the Flynn effect, tracking necessary level, without telling us anything about maximum IQ attainable.

            In practice, is there anything on hand to explain why necessary IQ would have to increase in the US between 1994 and 2014? The first thing that comes to my mind is the proliferation of computers/moblie devices/internet. Given that even basic computer literacy requires a somewhat different mental approach than interaction with humans or simple machines and that the internet tends to expand one’s intellectual environment considerably – especially if the starting point is none-too-high – it isn’t terribly surprising that the man on the street has gotten a bit smarter over the past 20 years.

            It bears emphasis that “passive consumption” of ICT/internet are likely sufficient for the effect to manifest – no advanced geekery required. It would be enough for people to simply have to deal with somewhat more abstract problems than previously and to be exposed to a peer group (by which I mean: people who share similar interests) that is on average a bit smarter than the people in their immediate real-world surroundings. None of this is a major stretch of the imagination.

            (Aside: The social effect is likely to have an upwards direction, because dealing with people less intelligent than you tends to be dissatisfying, even if you aren’t an intellectual ace yourself.)

            Also worth noting is that any “complication” of everyday life over the long-term would lead to the manifestation of the Flynn effect – if the satisficing hypothesis is correct. As long as there were no biological constraints at play people would continue to get “smarter”, simply because an increasingly complex environment during growing up would stimulate greater intellectual development. The kind of variation we see in the current population would still manifest, because of differing personal circumstances (both biological and environmental) – it’s just the baseline that would shift.

            Conversely (and with a bow to Popper), we would expect to see an “anti-Flynn” effect if the intellectual requirements for a “normal life” became lower over a long period.

          • JulieK says:

            Given that even basic computer literacy requires a somewhat different mental approach than interaction with humans or simple machines and that the internet tends to expand one’s intellectual environment considerably – especially if the starting point is none-too-high – it isn’t terribly surprising that the man on the street has gotten a bit smarter over the past 20 years.

            Or at least, has gotten better at certain kinds of tasks, while perhaps getting worse at “interaction with humans or simple machines.”

          • albatross11 says:

            If this model is true, it seems like we should also see higher IQs from people who are raised in high-IQ environments. Thus, adopting an average kid into a family full of very smart people living in a place where the kids are mostly very smart should raise their IQ. Is there any data that shows this effect?

          • Igon Value says:


            Your link doesn’t mention the Flynn effect (it’s just the URL to the Vox article from Scott’s post). Wikipedia says that the Flynn Effect may have plateaued in Norway, Denmark, etc., sometimes in the 1990s and even reversed since then (possibly due to immigration). Is it different in the US?

            Anyway, albatross11 may have been wrong on the specific factors (nutrition, etc.) but he may still be right that society has become more meritocratic for other reasons, and that most people reach their potential IQ (or close to it). This would make intelligence seem more inheritable than it was in the past (or that it is in other places).

          • INH5 says:

            I must have copied the wrong link by mistake. Here’s what I meant to post:


        • quanta413 says:

          The claim that child rearing has minimal impact on child outcomes is highly counter intuitive.

          It’s not clear to me that you have the correct claim in mind. The claim should be that child rearing within the parameters of a typical Western nation doesn’t seem to add much to the variance we see in outcomes.

          It may be that child rearing has large effects, but everyone does roughly the same thing. So it ends up not contributing much to differences between people within population.

          Give me a population where half the population whacks their children in the noggin or half the population is iodine deficient, and the amount of variance due to shared environmental effects will skyrocket. Populations with those sorts of problems aren’t the populations studies typically deal with.

          • Wency says:


            It seems human beings evolved to place a very heavy emphasis on child-rearing, which also made it into our cultures, because in the ancestral environment (and even in many environments in the world today), child-rearing has an immense effect on life outcomes, mainly by protecting children from utter deprivation.

            Today, in the West, it’s pretty easy to protect your children from utter deprivation. But this situation is historically novel, and we still have not adjusted for it, so we find ever more outlets for our instinct to parent.

            One other element of good parenting that was valuable historically but may not show up as better life outcomes for your children: imbuing them with loyalty to you, so that they might be more inclined to protect and care for you as you age instead of hating you. This notion may be transmitted more culturally than genetically. OTOH, it’s possible that loyal older children might contribute more to the household and therefore help their mother produce more children towards the end of her reproductive life.

            Also children from happy, well-parented families may have been more loyal to each other and therefore more likely to assist in one another’s reproductive success (though this would obviously show up as better life outcomes).

        • JulieK says:

          To my knowledge, every major cultural tradition… all spend a decent amount of cultural effort detailing how to raise kids and talking about the importance of doing it correctly.

          My impression is that a parent’s efforts in child-rearing get diluted and (nearly) drowned out by the second-hand influence on your child of the parenting their peers experienced. In other words, rather than having a big influence on your own child, you might end up having a small influence on all the children in the neighborhood.

          For example, whether the majority of families in your neighborhood have 1 parent or 2 parents has more of an impact on your child than whether your own home has 1 or 2 parents.

          If so, that would be an even stronger reason for a culture to stress the importance of good parenting, if it’s true that parenting has an impact, and the benefits are not primarily received by one’s own children.

          • I’ve seen the claim that in primitive societies parents do little child rearing, that once a child is weaned, it is mostly interacting with other children rather than with its parents.

      • rlms says:

        The Wilson effect is extremely counter-intuitive.

        • albatross11 says:

          [Wilson effect: IQ becomes more heritable as you get older.]

        • Salem says:

          That’s not counter-intuitive at all (or at least, people who find it counter-intuitive have strange intuitions).

          The score at the end of the match more closely tracks underlying team quality than the score at half-time, and for pretty much the same reasons.

          • rlms says:

            I don’t think that’s right. On a meta level, if the explanation was that simple then people wouldn’t feel the need to come up with explanatory theories based on genes having different effects at different times, and gene-environment interaction.

            On a concrete level: in your game-I-assumed-was-football-but-on-rereading-realised-wasn’t-specified, the situation is that a team has an inherent long-term goal-scoring ability of say 100 and experiences random factors throughout the match that mean their goal-scoring ability might be 90 after 10 minutes when only one thing has changed it and that thing has been negative, but in the long run the random factors will average out to have zero effect. If for a particular match this is not the case (say there is a maniac on a nearby rooftop who occasionally shoots members of the team in the leg, causing random factors to have a negative effect) then the statement about the score tracking inherent ability more accurately as time goes by will not be true.

            But if the team is a person with an IQ, and the random factors are the effect of the environment on their IQ there is no reason to expect that they will have an average effect of zero! Say someone has a genetic IQ of 110 which we average with an “environmental IQ” (total effect of environmental factors) to get their measured IQ. Your analogy assumes that the environmental IQ must also be 110. People have come up with theories about why this must be true (or more precisely why people with genetic IQ 110 tend to have environmental IQ 105 rather than 100, since heritability approaches an asymptote that is less than 1, but these theories are not a priori obvious.

          • Igon Value says:

            My guess is that the Wilson Effect is due to our inability to properly measure the IQ of kids. There is too much variation as a function of small age differences..

          • quanta413 says:

            I agree with rlms. It’s not like a sports team. We don’t measure adult IQs by taking an average over past scores or achievements.

            But I think the idea that the most plausible explanation for increasing heritability is a gene x environment effect is a mistake.

            At least one study finds that the heritability of height increases with age from about .5 when little to about .8 by age 18 too This is a similar pattern to what we see with IQ. Time shifted somewhat but similar trend in direction and total change in heritability estimate. That’s probably because humans have a long adolescence and don’t finish developing for a long time, not because genetically tall people seek out environments that make the appropriate height or something. It’s probably the same with IQ.

          • rlms says:

            My understanding (from skimming a couple of papers) is that a lot of traits increase in heritability with age, but not all (and not even all mental traits). Also, some studies have shown that the heritability of IQ starts decreasing again once you reach 60.

    • idontknow131647093 says:

      I think your thought process is a bit unclear to me. Why do you find twin studies give counterintuitive results? I usually find the results fairly intuitive.

      I’ve read around elsewhere in the thread that you think these kinds of studies underrate child rearing. What I would counter with is that this is because in the vast majority of circumstances, children that are currently alive in a developed country are receiving child rearing that is close to a “good enough” threshold, such that the differences in child rearing are not actually that large from the POV of a human child. It is not like we end up comparing 2 sets of triplets (one identical triple and one not) where one kid is raised by wolves, one by hunter gatherers in the north pacific, and one by rich suburbanites.

  24. David Conroy says:

    > The thing most of the early studies were looking for – academic ability – is one of the only things it doesn’t affect.

    It wasn’t designed to have affect academic ability, according to the guy who ran the “Follow Up” studies, which were so-designed. And it was proven quite beyond doubt that academic ability could be successfully targeted. It’s one of the most surprisingly under-acknowledged resultsin all of the social sciences.

    Siegfried Engelmann: A certain percentage of the students had to come from Head Start. Because Head Start was an obvious failure and they were concerned. It had no instructional component, and it was modeled after the middle-class preschool. While the middle-class preschool is probably okay for middle-class kids, the kids that we worked with were far behind in terms of language skills and…

    David Boulton: So it was more concerned with creating parental freedom than it was in actually helping the children get ready for school.

    Siegfried Engelmann: Right, yes. Anyhow, that made it a poor model for disadvantaged students. But fundamentally, Project Follow Through was designed to bail out Head Start. It was a horse race, the idea [of the APT reports] was to declare a winner or winners, those who produced the best results in K-3, to show that Head Start was not a total disaster.

    David Boulton: How could it have done that unless it was also using a control group of kids that weren’t in Head Start to show the advantages of Head Start?

    Siegfried Engelmann: Well, they had that. They had a vast number of comparison groups. For each school that was involved, there was a comparison school. They weren’t perfect, because the comparison schools tended to have higher socioeconomic ratings. They were not as disadvantaged. But, in addition to that, the data from all of the individual comparison schools were pooled. Then there was a certain non-disadvantaged mix as part of the formulated average school. So you had your non-disadvantaged population, and also (I can’t remember the exact requirement) I think over 60 percent of the kids had to have gone through Head Start. But they had data on the Head Start kids and the non-Head Start kids. It was a very elaborate study. It cost, I don’t know, hundreds of millions.

    David Boulton: So ‘Project Follow Through’ was a prototype – a model that would later be followed in many ways by the National Institute of Child Health and Human Development.

    Siegfried Engelmann: Right, right. The APT findings were suppressed largely for political reasons. In 1976 when Follow Through was being evaluated, Gene Glass, head of the Ford Foundation at the time, appealed to the National Institutes of Health with an incredible statement. He said something to the effect that, “The use of quantitative data is inappropriate and what we need is case studies. We need to document various aspects of the program so that informed consumers can make intelligent decisions.”

    And of course, it was total baloney. Wes Becker responded, with what I thought was an extremely succinct response, “As the problem with the disadvantaged is identified by data and scores; certainly the solutions to the problems would have to be manifested with data and scores.”

    David Boulton: Certainly it all has to correlate somehow…

    Siegfried Engelmann: [laughs] Yeah. They wanted to identify the problem qualitatively, and then solve it with methods that didn’t generate any data. Becker also pointed out that if we’re going to use case studies, how do we know we’re using typical case studies unless we use some kind of intelligent sampling processes?

    David Boulton: Yeah, and some common system of attributes that would allow you to scale through the data.

    Siegfried Engelmann: Right. So, the net result was that the results of Follow Through were suppressed. The report that came out on Follow Through was that the project was a failure, which implies that all of the models were failures. And then they just rode off into the sunset with some kind of blazing saddles and that was that.

    • Freddie deBoer says:

      The largest, highest-quality studies (and modern studies, to boot) have all found little to no academic advantage to Head Start and similar programs. The studies that have found strong effects are older, smaller, and lower quality.

      • David Conroy says:

        The study I’m talking about is the largest highest quality study, And it does show a major academic effect, as do many many other studies on Engelmans instructional programs. Also, that headstart didn’t work is not in question. It didn’t teach academic skills and so, unsurprisingly, academic skills were not learned. Engelmans method tested in “Follow through did work astonishingly well though.

  25. ThomasStearns says:

    “Preschool is a shared environmental effect; your parents have a big effect on whether or not you go to preschool”.

    I don’t see how preschool is a shared environmental effect. Not to leave my own family, my brother and I went to the same preschool, but we went for different amounts of time = not a shared effect. I’m sure a large share of people went to different preschools than their siblings did, and it’s seems pretty ridiculous to assume preschool homogeneity.

    This is unsettling to me because the other day I was asked to explain what I meant by “shared/nonshared” environment in conversation, and maybe got it wrong.

    • Kris says:

      Yeah, but it’s probably highly correlated. If preschool has an effect, and there is significant variation in preschool years across families, it should show up in the shared env effect.

  26. benjdenny says:


    Does this apparent inconsistency change if we think of the children’s improvement not as a function of preschool, but as a function of having the kind of parent who would take advantage of preschool? If I imagine a world where there’s a gold ingot in everyone’s front yard that you just have to go get, I’d imagine the children from families that declined to get the gold ingot would have much worse outcomes than those that did, even if a week later we decided to give everyone who didn’t get the ingot the equivalent cash value anyway.

    Given that the preschool thing almost universally indicates that your parent is the job-getting kind and the non-preschoolers in the same income brackets would be more likely to be something different than that, I’d expect some sort of outcome difference even if preschool didn’t do anything. If we could tag the parents with children currently in preschool in a low income bracket, then change the universe so preschool had never existed but otherwise nothing had changed, I’d expect the tagged parent’s children to do better as a group than the non-tagged, even though nobody was going to preschool now – they are very probably fundamentally better people as a group.

    I know there’s a whole argument to be had about whether any of the bullshit I just said is true or not, but my question is: Taking as a given that it is, would it alleviate your confusion any? I’m not being snarky, I legitimately don’t know.

    • Statismagician says:

      Sort-of. Private preschools are either insanely expensive (and SES tracks really well with overall family success) or provided by churches/prestigious workplaces like universities/major hospitals (membership in/employment by tracks really well with overall family success all else being equal); I suspect there’s a parental-engagement mediating effect within SES groups, but this would presumably be, and in fact is, masked by the between-group variance if you try to match Head Start participants against people who attend St. Fancy Guy’s Academy for Gifted Toddlers or whatever. Public preschool programs are generally free or very low-cost, and often recruit heavily among the target population, so this hypothetical parental-engagement effect would be much smaller. Also parental engagement is a direct target outcome of lots of these programs, further muddying things.

      • benjdenny says:

        I’m not sure that parental engagement is the right metric for what I’m talking about, though. I’m thinking about it this way:

        The lower a household income is, the more vital free/low cost childcare becomes. If you have a population of, say, single mothers making >25k, their options look something like A. Use the childcare and work B. Don’t do that and live off of others, often very poorly and C. Outliers(found buried gold, grandma pays just enough for survival but demands mom stays home, ect). In that case, the people who go “fuck it, I’m not using that service, I’m just gonna stay home” are probably going to by-and-large skew to be shittier people than people who went “I’ll drop the kids off at daycare(thank god for daycare) and then go work a job”.

        At that point I don’t think parental engagement is the only metric – If I take a “fuck it” person and a “I’ll work” person, even if their kids are getting equal amounts of engagement one person’s kids are getting it from a shittier person. I’m not sure if I’m thinking about this right, and I can’t imagine how you’d control for it, but I’m also not sure how they could possibly draw a “single mothers with no or low income who have access to public daycare but don’t use it” and not have a much larger proportion of people who have just given up on life than other groups.

        • Statismagician says:

          That all makes sense. I wasn’t using ‘parental engagement’ sufficiently rigorously; I ought to have said something like ‘conscientiousness+engagement in child’s life+effort-towards-improvement.’ I think given how low the barriers to entry are for public options (people are actively trying to find you to give you this free program intended to solve lots of problems for you and your child), the degree of dysfunction you need to refuse it is sufficiently-higher than the base dysfunction rate to be less important than it would otherwise be. This is not necessarily true everywhere or for all programs; I’m basing it off of anecdotes and personal experiences from the research side.

          So, basically, I agree in principle, I just don’t think it’s a significant methodological concern in the target population. I’m not wedded to this, and I’d like to see it studied in both low- and high-SES populations just to make sure, though.

          I’m not sure what to think about the ‘proportion of people who have just given up on life;’ my first thought is that there may a higher relative effect in high-SES groups, but a higher absolute effect in low-SES – i.e. an upper-middle-class kid with a dysfunctional single parent will be worse off than an upper-middle-class kid with a highly functional single parent to a higher degree than the equivalent very poor kid; even though very obviously the poor kid is much worse-off in absolute terms the difference compared to a functionally-parented very poor kid is probably smaller.

          • benjdenny says:

            I think we agree mostly. I’d be especially interested if there was ever research Re: your last paragraph there. I agree that the absolute difference between rich, good parent and rich, bad parent is pretty huge, probably bigger than poor/good-poor/bad, but I wonder if that’s enough to push them into “unsuccessful” territory as defined by the paper pretty often. I wouldn’t gut-feeling expect a rich kid to require as much welfare or be stuck in as bad of jobs as either kind of poor parent’s kid no matter how bad their parents are, but I wouldn’t be surprised to find that I’m dead wrong on that either.

          • Statismagician says:

            I’ll see if I can find anything in the literature. Might be able to do some back-of-the-envelope investigation with public data, too.

  27. JulieK says:

    The supply of available daycare isn’t keeping up with demand.

    Children younger than 3 are the fastest-growing age group in the District [of Columbia], with a 26 percent increase in the number of infants and toddlers between 2010 and 2014. The explosion of young families is putting a new strain on an overtaxed child-care industry.

    Many new parents describe a maddening search for child care, encountering wait lists 200 names deep or two years long and child-care administrators who do not return repeated phone calls, overwhelmed by requests from prospective clients who start knocking sometimes before they are even pregnant.

    After all of that, when a coveted spot becomes available, many parents seize on it, even if they don’t need it — yet.

    The city’s universal preschool program, a free full-day program for 3- and 4-year-olds, represents a light at the end of a tunnel for many parents in the District. But child-care advocates say the ramping up of universal preschool actually exacerbated the child-care shortage.

    Many private centers once served children from infancy to prekindergarten, relying on larger preschool classes to underwrite more expensive infant programs. But they could not afford to stay open once the preschool private market diminished.

    • Nicholas Weininger says:

      Isn’t this partly because DC has absurd credential standards for child care workers?

      • kominek says:

        most everywhere in the US makes it awful to run any sort of child care operation.

        if all this day care stuff is so useful, the first thing to do is drastically relax those requirements across the country.

      • JohnBuridan says:

        There was a good article in Jacobin about child-care credentials and home-based care.

        The article argues for strongly credentialed, centralized, universal childcare as the safest and best childcare. They argue further that home-based and unofficial child-care services are dangerous and neglectful and lack necessary oversight.

  28. Kris says:

    I’m all in favor of updating on new evidence, but your change of heart from ‘basically useless’ to ‘no-brainer’ is puzzling. No intermediate states?

    Especially given the whole area should be a big question mark. Yes, these findings contradict twin studies, adoption studies, lottery studies. Not to mention, the studies finding positive effects seem to be of much lower quality than the ones finding no effect (longtitudinal vs random).

    And this whole business of claiming let’s do X because of Y, finding Y is false, then going to ‘but there is Z’ really stinks. Sure, one could rationally arrive at such a conclusion. But it seems more like searching for reasons to adopt the conclusion we already want. Especially since the conclusion is very popular.

    Even if there is an effect, it doesn’t immediately justify government intervention. How big are the effects? What are the opportunity costs? How do we handle public choice concerns? Is it legitimate to steal money for this cause? etc.

    So I’m really curious. How did you go all the way to ‘no-brainer’ instead of “I am totally confused by this field. Let’s study it more. Let’s definitely NOT commit any significant tax dollars to this new universal program before we understand it better.”

  29. miguelmadeira says:

    I thought that the main point of the pre-school is “to have a place where to put the children during the workday” was more or less consensual; or not? Being in the second place “It is good for him/her to interact with other children and learn that he/she is not the center of the world”. In the real world, I never see anyone saying that preschool is good because “raises IQ” or to “learn how to learn” or anything like that.

    [And never see anybody really making a distinction between “pre-school” and “kindergarten”; and with “daycare” the difference is because “daycare” – well, “atividades de tempos livres”/ATL, who I think is equivalent to American “daycare” – is for children who already are in the formal school, and after school, go to the ATL]

    • Statismagician says:

      You’re right about daycare vs. *-school, and about what preschool/kindergarten is actually for, but a lot of American preschools and essentially all preschool-type social programs try to justify themselves as being enriching foundations for future academic success.

      • JohnBuridan says:

        You don’t deny that kindergarten can be a more or less enriching environment though, do you?

        • Statismagician says:

          Not at all, I’m just not very surprised that the difference turns out not to matter much in the grand scheme of things, all else being equal (obviously kindergarten is more valuable if the alternative is ‘child placed in bare concrete cell with no human interaction’ vs. ‘child plays with siblings, relatives, other toddlers in a safe environment’).

  30. Douglas Knight says:

    The lead effects are huge.

    No, they aren’t. Do you have a source that gives actual numbers?

    • Statismagician says:

      Would you like to try again?

      The overwhelming scientific consensus is, and has been for decades, that lead is in fact very bad for you, and especially so for young children, as you know perfectly well. If you’ve got contrary data, we’d all love to see it, because lead prevention programs are really expensive.

      • j1000000 says:

        Have people done studies on Flint?

        • HeelBearCub says:

          Flint is a blip, in terms of lead levels that were actually attained.

          Here is Kevin Drum. Note the chart where we can see that lead levels started to rise and then dropped again as action was taken They did not rise to anywhere near the levels they were 20 years before when 50% of children had levels above 5 m/d, and nearly 10% had levels above 10 md/d. The peak of the water effect was ~6% above 5 m/d.

          • j1000000 says:

            Still — doesn’t that form a natural long-term experiment, to measure the outcomes of that 6% against other children who were not exposed? Are there papers of this sort, where they measure the results of new lead-free projects vs. others nearby that were exposed to lead paint?

            I’m sorry, I’m not good at looking for studies. I just only read this Kevin Drum article where he points to Nevin saying that, with a 23 year lag, lead explains 90% of the difference in crime. That seems such a large effect to be implausible to me.

          • The Nybbler says:

            If small effects are probably noise and large effects are implausible, you’ve got a fully general argument for rejecting everything.

        • Statismagician says:

          Also, the long-term developmental sequelae would be just now starting to show up measurably for the worst-exposed children, I believe.

      • Douglas Knight says:

        Does that linked page contain any effect sizes? A first glance does not reveal a single one.

        Worth the expense of remediation is a very different question from “huge” for the context:

        All I can think of is that maybe shared environment can matter, but is so small in the grand scheme of things that it’s below the threshold where zoomed-out studies of everything can detect it. That would help reconcile the two literature bases. But it doesn’t seem right. The lead effects are huge.

        • Statismagician says:

          That’s a fair question, mostly they give blood lead levels which are associated with various conditions but don’t specify the rates except for “for every 10 µg/dL increase in BLLs, children’s IQ was found to be lowered by 4 to 7 points.” Let me find some papers for you – are you more interested in relative risks for various things by lead-poisoning status, or a dose-response curve for some particular condition/s?

          • Douglas Knight says:

            4-7 IQ points per 10 µg/dL would explain a negligible proportion of the variance today. It would be about 10% of the variance in 1975. Since that is about all of the observed shared environmental effect, that would be huge. But is 4-7 the consensus? Most papers I find put it at 1-2, 1% of the 1975 variance, though some put it at 10-20, all or more of the variance.

          • Statismagician says:

            According to a lot of the things I’m finding, there really isn’t a lot of modern research on lead effect sizes (as opposed to lead levels or lead poisoning prevalence), apparently because the consensus that it’s really bad all the time at any level has been so strong for so long and, related, because we’ve been so successful at lowering population/environmental lead levels. Given what we now know about the quality of 1970s-1990s research, I think this means it’s at least worth replicating some of the older claims. I’ll continue looking into it.

          • SamChevre says:

            I think it’s important to note that “IQ loss” is a terrible proxy for “harms of lead exposure”; all the anecdata I have points out that the real harm is diminished impulse control/shorter time horizon–which is only very weakly correlated with IQ.

      • Doug says:

        Everything cited at the CDC regarding IQ is simply an association study. Lead may be associated with diminished IQ, but we have no idea the magnitude of its causative impact. Considering that lead exposure is highly correlated with SES, and hence parental IQ, it’s extremely likely that there are major confounders in any population wide lead association.

        > This paper supports those findings with evidence of uncontrolled confounding by parental education, intelligence or household management from several papers. It suggests that inappropriate statistical tests and aggregation of data representing different exposure routes partly explain why confounding has been overlooked. Inadequate correction of confounding has contributed to incorrect conclusions regarding causality at low levels of lead.

  31. notpeerreviewed says:

    I think 15%-20% is the normal finding, not an outlier. Read the fine print in the big ole’ metastudy:

    In the abstract, they say “the data are inconsistent with substantial influences from shared environment”, but the word “substantial” is doing a lot of work. On page 704, they say the average shared environment effect across all studies in the past 50 years is 17.4% – unless I’m reading the s^2 statistic incorrectly.

  32. Douglas Knight says:

    I am extremely skeptical of studies with an intermediate cause that don’t measure that intermediate cause. If the evidence is just that the intervention causes an effect 20 years later, but causes no observable effect in the intervening time, I do not believe it.

    It is very difficult to do 20 year studies, but if you have a hypothesis about a new intermediate effect, you could do a 5 year study to see if it has an effect that lasts that long. I’ve been hearing for a decade that Head Start raises conscientiousness. Has anyone done a 5 year study of this? I haven’t heard of it, but maybe that’s just because such studies don’t get publicity because no one wants to convince people like me. Piper does cite a regression discontinuity paper for the 5 year health effects of Head Start. That’s a good start! Does it replicate today?

  33. mdb says:

    Effects 20 years after treatment? That is pretty weak to claim you were wrong about preschool. how many confounds are there? Would parents that worked be better roll models? That is just one of the thousands of confounds. I could see switching to undecided, but wrong is a stretch. One thing about the early small studies versus the large programs like head start, is the quality of instruction/employees. I never looked, but I would bet many of the small studies had grad students and post docs providing the instruction and care. Head start is as much an employment program as it is a preschool. The quality of care and instruction, I believe would not be comparable and that is why the results do not replicate.

    I am agnostic about head start and many of the progressive programs, I am sure in total they have an impact, but many are small and cumulative. That is why, I still think results driven social spending would be better use of scarce resources. They should always be looking for a better way, and throwing billions to expand, what is at best a marginal program, is not a smart use of funds.

  34. Statismagician says:

    This very much tracks with some research I’ve been involved in, and illustrates how unbelievably important it is to be clear on what an interventions outcomes vs. outputs are supposed to be in social science research. I’m moderately confident that further investigation will show that it’s the improved-nutrition and family-functionality components of these programs which are important, which isn’t as new a finding as you might think.

    I had to read a giant stack of papers on home-visiting programs several years ago and the averaged conclusion was that none of the ‘how-ready-is-this-toddler-for-kindergarten’ metrics showed any meaningful effect, but that families were better at providing a safe, stimulating environment and were more likely to be involved in school activities, and that there were hints of improved abstract-problem-solving abilities compared to other people in similar SES groups; these are exactly the sorts of things I’d expect to have outsized long-term effects, and in fact did as early as the 1960s (well, fine, 80s-90s for the lifetime outcomes) with the Perry Preschool Project. Head Start, and basically all other ECE-type programs, are different combinations of perinatal medical care, connection to social services, life-skills training for parents, and home safety/stability interventions stapled onto very similar-looking preschool programs; I’m not at all surprised that these have meaningful effects even though there’s no academic effect. Indeed, although I haven’t looked into it deeply, I also wouldn’t be surprised if the correlation of educational attainment with positive life outcomes generally was some sort of abstract meta-proxy for not having needed this kind of thing in the first place.

  35. notpeerreviewed says:

    “…15-20% from everything doesn’t seem consistent with measurable effects from preschool alone…”
    Can you elaborate on that? Are you saying that, based on the size of the preschool effects, you would expect to see shared environment effects larger than 15-20%?

  36. baconbits9 says:

    My only caveat in agreeing with this perspective is that Chetty finds the same effect (no academic gains, but large life-outcome gains years later) from children having good rather than bad elementary school teachers.

    Straight forward question- did the studies cited make sure that the effect wasn’t from districts getting more resources pulling in better than average teachers?

  37. Deiseach says:

    This suggests that preschool is beneficial not because of the curriculum or because of “teaching young brains how to learn” or anything like that, but for purely social reasons.

    Preschool education is slightly more complicated and complex than “kids will get good grades on tests/will do better academically when they start proper school”, and even though the socialisation effects are very much intended, the curriculum does merely “let’s just get a bunch of kids in a room and let them socialise”.

    Here’s links to the Irish curriculum for preschools called Aistear (there’s a related framework for early childhood practitioners called Síolta, if you want to trawl through it and have fun with all sixteen Quality Standards that have to be met – we have to incorporate this where I’m working and we do get inspected on it).

    Purpose of Aistear
    Aistear is the curriculum framework for children from birth to six years in Ireland. It provides information for adults to help them plan for and provide enjoyable and challenging learning experiences, so that all children can grow and develop as competent and confident learners within loving relationships with others. Aistear describes the types of learning (dispositions, values and attitudes, skills, knowledge, and understanding) that are important for children in their early years, and offers ideas and suggestions as to how this learning might be nurtured. The Framework also provides guidelines on supporting children’s
    learning through partnerships with parents, interactions, play, and assessment.
    In supporting children’s early learning and
    development Aistear
    ■ identifies what and how children should learn, and describes the types of experiences that can support this
    ■ makes connections in children’s learning throughout the early childhood years and as they move from one setting to another
    ■ supports parents as their children’s primary educators during early childhood, and promotes effective partnerships between parents and practitioners
    ■ complements and extends existing curriculums and materials
    ■ informs practice across a range of settings, disciplines and professions, and encourages interdisciplinary work.

    It has suggestions for what parents and childminders/creche workers/nursery schools/preschools can do for babies on up. Here’s the link to the Principles and Themes and here’s the Guidelines – sample text:

    What do I assess and when?
    In assessing, the adult looks for evidence of children’s progress across Aistear’s themes:
    ■ dispositions: for example curiosity, concentration, resilience, and perseverance
    ■ skills: for example walking, cutting, writing, and problem-solving
    ■ attitudes and values: for example respect for themselves and others, care for the environment, and positive attitudes to learning and to life
    ■ knowledge and understanding: for example classifying objects using colour and size, learning ‘rules’ for interacting with others, finding out about people in their community, and understanding that words have meaning.

    The adult focuses on what children do, make and say. For example, he/she might observe babies watching each other and initiating communication through a hand-touch or a screech, or toddlers working collaboratively to move and build a mound of stones, or young children investigating sounds as they create instruments from objects of different materials, shapes and sizes. The adult uses the aims and learning goals in Aistear’s themes to interpret and build on these experiences. Collecting information over time is especially important with children from birth to six years as their learning and development does not follow neat patterns or happen at the same rate for each child.

    Children also have developmental milestones, which health professionals check at certain times during early childhood. In addition, diagnostic assessments play an important role in helping to identify children with special educational needs. Although most practitioners do not carry out diagnostic assessments, they often notice early signs of potential difficulties and can bring their concerns to parents and help them get in touch with relevant professionals.

    I imagine American standards are somewhat similar, though I have no familiarity with the No Child Left Behind Head Start programme.

  38. vpaul says:

    Predicted blog post five years from now:
    Preschool: I was wrong about being wrong

    The studies cited in the Vox piece go the opposite direction of daycare studies. It’s strange that preschool (essentially daycare at a slightly later age) would cause strong positive long term effects for children, while daycare (preschool at a younger age) would cause moderate to small negative long term effects for children (the conclusion from daycare research, granted study quality is low).

    In fairness I believe daycare studies showed that for poorer children the long term effects were less negative.

    I think it’s plausible that preschool could provide positive effects for poorer and disadvantaged children. But I would predict some negative effects too, possibly for different populations, in line with daycare research and research that shows academic achievement following preschool actually decreases. I expect subsequent studies will tamp down on the optimism.

    • Statismagician says:

      Important caveat: daycare probably really is just child-warehousing, and shouldn’t have significant effects beyond the parent being able to work more easily (and people for whom this is a deal-breaker might very well be otherwise disadvantaged, masking any positive effect). Early-childhood education programs are often this, plus medical care, plus better nutritional, plus connection to other social services, plus parental life/parenting skills training, so for some programs at least it’s an apples to oranges comparison, with the results expected if the academic bits are useless but the others aren’t. I haven’t read the daycare studies you have in mind, to be fair, so this might be wrong.

      • deciusbrutus says:

        Daycare is also social exposure to other children and a variety of adults, while being raised alone at home is not. Being raised at home with a number of siblings two standard deviations above average also means being raised by parents that are far from normal.

        • Being raised at home with a number of siblings two standard deviations above average also means being raised by parents that are far from normal.

          I can’t tell if you think that’s a bad thing, a good thing, or merely a problem for studies to account for.

          • Statismagician says:

            I think it’s probably any of the three, in ways which can’t be usefully predicted from demographics.

            EDIT: My opinion, not an interpolation of @deciusbrutus’s thought.

          • albatross11 says:

            So you’d think that being raised around extraordinary siblings should have some long-term effect, but I think it’s hard to find such an effect in adoptions studies, where you should be able to find it.

          • I think it’s a good thing, but that’s from a very small sample size.

      • vpaul says:

        You seem to be describing the ideal early childhood program, I imagine many programs (especially large scale ones) fall short of that ideal (I haven’t read all the studies).

        Many daycare studies seem low quality, to the best of my knowledge randomization is very difficult or impossible, and I think it’s impossible to control for the differences between a family that sends their child to daycare and one that doesn’t (even if all demographic / economic factors are identical, a family choosing not to send a child to daycare is un-quantifiably different). One point in favor of the results is that the most researchers would be biased towards finding no negative impact, so results which show negative impacts are somewhat stronger in my mind.'s_elevated_cortisol_levels_at_daycare_A_review_and_meta-analysis (it seems the cortisol effects are larger under 36 months, which is consistent with different effects between daycare and early childhood education). (an article but good overview).
        (I think this was one of the best observational studies, it found negative effects even for relatively short amount of time spent in daycare).

  39. AnthonyC says:

    Thanks Scott. Posts like this are confirmation that my reasons for coming to SSC are sound.

    Since it’s election day, I’ll add: I long for a world where leaders could get bonus points for doing the very thing this post does. I don’t expect to get it, but I would *love* to see what a society like that could achieve.

  40. baconbits9 says:

    I’m reading through the links and I have some questions. One of the studies says

    Design, Setting, and Participants This matched-group, alternative intervention study assessed 1539 low-income minority children born in 1979 or 1980 who grew up in high-poverty neighborhoods in Chicago, Illinois. The comparison group included 550 children primarily from randomly selected schools participating in the usual early intervention. A total of 989 children who entered preschool in 1983 or 1984 and completed kindergarten in 1986 were included in the Chicago Longitudinal Study and were followed up for 27 to 30 years after the end of a multicomponent intervention. A total of 1398 participants (90.8%) in the original sample had educational attainment records at 35 years of age. The study was performed from January 1, 2002, through May 31, 2015.

    My first reading of this is that 2089 children were selected for the study and 989 children finished the study, is this correct?

    • Nornagest says:

      It’s awkwardly written, but 989 + 550 = 1539, so by my reading 1539 kids were studied, of which 550 were slotted into the control group and the remaining 989 into the intervention group. Of the 1539 students studied between these two groups, 1398 achieved some kind of educational credential by age 35. This paragraph doesn’t say how many did per group.

      • Douglas Knight says:

        It’s not that 90% had a credential, but that for 90% they were followed into adulthood, so that it could be determine whether they had a credential. The 10% rate lost to follow-up was the same in both arms.

  41. Anthony says:

    <voice=infomercial>Will It Replicate?</voice>

    Ms Piper’s conclusion is that the beneficial effects are primarily due to preschool being reliable day care are stronger and more focused than Scott’s conclusions above. Assuming that she’s right, the question still remains if the benefits are from preschool, or from having parents who would take advantage of preschool.

  42. srconstantin says:

    It seems from the Vox article that the healthcare effects might dominate.

    Head Start is packaged with healthcare. There’s an obvious mechanism by which getting medical treatment might lead to better life outcomes.

    Medical treatment is only partly a shared environmental effect. Yes, whether you get it or not depends on your parents; but whether you *need* it or not depends a lot on genetics. For healthy kids, different amounts of access to medical care in different families won’t matter that much. For sick kids, it will. But more closely related siblings (i.e. identical twins) given different access to health care might still have more similar life outcomes than nonidentical twins given different access to health care, since the fraternal twins are also more likely to have different risks of disease. So access to medical care, treated as a shared-environment effect, will look weaker than it is, since the *importance* of medical care is partly genetic.

    I think this argument could also apply to lead and other physical environmental effects. Suppose environmental toxins are worse for people who already have a genetic “vulnerability” of some kind. Then twin studies will underestimate the effects of environmental toxins, since identical twins are more likely to share a vulnerability than fraternal twins.

  43. Doug says:

    I tried to make a serious effort-post closely examining the research cited by Vox. However it seems to have been eaten by the spam filter. Extremely frustrating. But the full comment can be found on Reddit here:

    My conclusion was that Scott is being extremely overconfident in the strength of the evidence. Nothing in the piece should cause one to significantly alter priors, let alone flip their opinion. All of the cited studies have some sort of glaring issue. I go into much more detail in the above post.

  44. The Element of Surprise says:

    Chetty finds the same effect (no academic gains, but large life-outcome gains years later) from children having good rather than bad elementary school teachers.


    strengthens rather than weakens the Caplanian case against education, since the studies find that the educational parts of preschool are not useful, and better teachers and curricula do not affect the benefits.

    These two seem directly contradictory to me, or am I misreading something?

    Maybe this is an effect similar to the one described in Too Good to be True, where the thing / method / outcome that has been intensively looked at with high quality research shows little effect, but the things that one would not look for a-priori, that only get published if there is actually a positive finding, show a larger effect.

  45. TDB says:

    I wonder how different the results would be if they studied play groups that don’t try to educate. I suspect that Kids learn a lot from social interaction during play with other kids.

    Do we advocate compulsory pre-K? 😉

  46. Dave92F1 says:

    Just to point out the obvious, correlation is not causation.

    How do we know that parents who bother to send their kids to preschool just do a better job of parenting in general – with the good effects coming from the better parenting, not the preschool?

  47. deciusbrutus says:

    Would it make sense for all child-rearing to be performed by specialists in a creche, rather than continuing to require that children be raised by random untrained unqualified people?

    • Hoopyfreud says:

      I’ve had similar thoughts and concluded that a policy of making sure that no child experiences parental love is horrifing.

    • albatross11 says:

      We have some examples of such schemes, in orphanages and such. My very limited understanding is that these are often pretty awful.

    • John Schilling says:

      Creating high-quality adults by this process will be a very very expensive and complex proposition. The abstract altruistic desire to see atomic individuals achieve fulfillment in their lives, will only go so far in accomplishing all this. So I’m guessing most of the people who pay for this, and most of the people who actually do this – including most of the people who do the detail work of writing and enforcing the rules of how this can be done – will be interested in creating high-quality adults for some other, more specifically targeted, purpose. Insofar as the task is delegated to a bureaucracy, that purpose will be the enforcement of the Iron Law.

      I once noted that it is a Very Good Thing that humans find it difficult to create children that they do not love. But you might be able to pull it off with this approach, if you try. Please don’t try.

      • pontifex says:

        If only Mustapha Mond were on the ballot for state controller, this would have a chance. I always thought the department of human resources wasn’t doing enough.

        Seriously though, we might end up with some kind of system like this for space colonization. Exposure to radiation like you get in the depths of space makes humans unsuitable for repdroduction. You might end up with a worker drones + queens setup.

  48. Michael Arc says:

    It seems to me like there’s a systematic tendency for the left to give bad reasons for doing the right thing. I think I understand why as well. Happy to discuss in meatspace next time I’m in the Bay, best guess is March.

    • acymetric says:

      Not to pry, but is there a reason you don’t want to discuss it here (after bringing it up)? I don’t disagree, at least at face value, but I would definitely be interested in your explanation.

  49. jesduff says:

    I have a possible explanation for why we would find shared environmental effects in the context of crime but not in other areas. Because criminal activity is scorned by the vast majority of society, committing any small crime requires a decision about group identification. Once a person has committed a couple of crimes, to make sense of how their own behaviour relates to the rest of society, they’d need to identify with a criminal identity that permits such behaviour. If that element of human psychology weren’t at play, they might commit a couple of crimes purely because they were egged on by a sibling etc but then it would be extremely unlikely that it would predict delinquency later in life considering how different they would be as a person.

    Contrast this to things like preschool where nobody forms an identity about the fact that they’re in preschool because so is every other kid they know; such an identity would serve no purpose, and this means no behavioural traits would be needlessly proliferated into adult life.

  50. edmundgennings says:

    Technical quibble with point 3. There is a decent amount of measurement error in measuring both intelligence and other life outcomes. Thus given error in variables bias the correlation between how someone does on one iq test and their income one year will be considerably lower than their intelligence and their lifetime economic achievement including not just nominal salary one year but all economic benefits.

  51. DragonMilk says:

    For parents who raised their kids:

    How many kids, and how often would you have playdates with other kids? Do you think their level of socialization was sufficient?

    I personally don’t want to subject my kids to the disease bag of preschool where some exhausted adult tries to pass the time with a bunch of screaming kids not feeling well, but all my friends have told me that they have no doubt my kids could be smart, but may need others to help socialize them, haha.

    • Igon Value says:

      I have the same problem. I don’t want to send my kids off to school –let alone pre-school– and yet I get the sense that it is important that they interact with other kids, or that it is important that they get used to the classroom environment, etc.

    • Mark V Anderson says:

      This is a really hard question to answer. My two kids are now in their twenties. I was a bit concerned that they did not have many friends to play with when they were little, because not many kids in the neighborhood, and I never got the memo that I was supposed to purposefully find other kids to bring them together, and not count on the neighborhood. Of course both kids also went to daycare and of course school for many hours of the day, so they probably got lots of socialization there. What it comes down to is I have no idea if they got enough. I think overall at least some daycare is a good thing. My kids liked their time in daycare. But every kid is different, and every kid gets lots of socialization in school (unless they are home schooled). I wouldn’t worry overmuch about it, because kids are resilient.

    • brmic says:

      They enjoyed preschool tremendously to the point of having to be cajoled into going home. Providing opportunites for socialization was an argument initially, but wouldn’t have been sufficient to keep them there against their will. (FWIW the other argument was ‘diversity’ in the sense that my habits and language patterns and the toys at home are limited in range and I believe it does them good to be exposed to the patterns of other people.)

    • Randy M says:

      My three daughters are being home-schooled. How much interaction with others they get varies with the life changes their friends families go through, various co-ops and play groups forming and dissolving, schedules of activities and so forth. But in general each week each child has one informal group for a few hours, one to two classes with groups (gymnastics, kung fu, etc.), one to two classes with an an individual instructor (piano, horseback riding), one church service, one to two trips to Grandma’s possibly with cousins, and a couple hours playing with neighbor kids.

  52. graehl says:

    Have you considered the appropriate multiple-hypothesis-testing discount for this “oh, it turns out this intervention actually had this *other* modest-sized benefit instead of the one we were intending”?

  53. slovakmum says:

    Preschool can “normalize” children enough so that they are not considered retarded. No kidding. In Slovakia, there is a significant Gypsy minority living in very poor conditions and an absurd percentage of their kids attent special schools for mentally disabled children. Subsequently, they do not have access to some form if higher education. The human rights agencies press for mare careful way of evaluating mental ability, because several proven cases of errors are known. Also, they want to enable access to higher education even for those labeled as mentally disadvantaged. Preschool must be helpful as well, because the errors in diagnosis are often caused by the lack of stimuli the children received previously.

  54. b_jonas says:

    Isn’t this actually the traditional European wisdom?

    We have kindergartens and creches for keeping children before school age, so that their mother (or some other relative) can work on weekdays (except possibly during the summer break). In theory, it is compulsory in Hungary for children to go to kindergarten for at least one year, in the same sense that 10 years of education is compulsory for at least three decades now. In practice, most parents want to send their children to as many years of creche and kindergarten as they can afford, but of them can’t, because there isn’t enough capacity in the system of kindergartens and creches. This is just like how most parents do want their children to have 12 years of school (the full length of the primary education system), and most parents would want their children to go to college after that, but many can’t afford it. (I don’t see much of this directly, because of class separation and because I live in Budapest, but at least once I was shocked by a billboard advertising a “kindergarten, with garden”, because I naively assumed that all kindergartens would have a garden.) There is always still a significant minority of very poor parents who want their children to finish school as early as possible, sometimes earlier than the legally compulsory 10 years of school, so that the children can start doing work that earns money for the family, but that obviously doesn’t affect kindergarten. The tradition says that the kindergarten system was started by Brunszvik Teréz in 1828, so there are various statues of Brunszvik Teréz and institutions named after her.

    But my childhood memories and the old stories I read agree on that children go to kindergarten to play, ideally both indoors and outdoors, together with other children of similar age (under the care of a kindergarten pedagogue, who are almost all female by the way). The children only start learning in an organized way when they go to school, starting in a September between their 6th or 8th birthday depending on their abilities, and leave kindergerten the spring immediately before that. In particular, the meat of school is considered the classes when teachers teach and the children of the class sit at their desks, generally in approximately six 45-minute intervals within a day, and the recesses between the classes and lunch break and possibly extra childcare after the classes are considered necessary additions, not the main point of the school. On the other hand, kindergarten traditionally does not generally have classes at all. It has times allocated for indoor play, outdoor play, meals, washing your teeth, and mid-day nap time (with the children in beds and possibly sleeping, but at least being quiet and not distracting other children who are sleeping), but not time for classes.

    It was only in the last 10 or 15 years when the popular image of kindergertens started to change, and in the non-poor regions of Hungary, many kindergertens started to emphasize education. Today we find a lot of kindergartens advertising as “képességfejlesztő óvoda”, offering learning activities apparently modeled after the American pre-school system, such as lessons where children learn songs and words in English.

    Update: apparently entognatha says the same in their comment.

  55. ana53294 says:

    In Spain, at least in bilingual provinces, preschool does actually teach something academic – the local language. It is also very helpful for immigrants.

    My guess is, in poor non-English speaking communities, preschool will have an effect on trades for the first few years – until kids learn English.

  56. mtl1882 says:

    I haven’t had time to read through all the responses, but these debates always bring one thing to mind for me. No, I don’t believe people are born a blank state, and yes, I believe genetics plays a big role, though of course the results are heavily shaped by life experiences.

    But if you torture yourself by reading about “feral” children, or those completely neglected from birth, or starting shortly after birth, by mentally ill parents, one thing is clear. No matter what your genes indicate in the way of intelligence, if someone doesn’t spend time developing your cognitive abilities early in life, you end up the same: utterly destroyed. Intellect and personality has to be developed by human contact. And if the situation isn’t remedied within the first few years, it never is. The capacity for language, or at the very least grammar, is simply gone.

    I do not know the subtle implications of this, such as how much minor damage could be caused very early on by a moderate lack of engagement, but it certainly seems worth exploring. I think there is a good possibility that encouraging the mental engagement of toddlers through preschool could make a considerable difference. It would catch the ones who are being somewhat neglected but not in such a way that they’d be removed from the home, such as frequently being left to watch TV for hours on end with no parent talking to them. I am not particularly concerned with targeting low-income families who I think are often envisioned as the beneficiaries. I think this sort of thing goes on at all income levels, and some parents are just simply clueless.

    I assume this is the same issue that causes attachment disorders and headbanging among toddlers in orphanages. It seems hard for some people to imagine that something that occurs that early in life could be so traumatic, and I think people often make assumptions about genetic causes, but if you read about “feral” children (for lack of a better word), it’s very clear that the presence or absence of some sort of interaction from infancy is pretty much definitive. We can make a difference at that age, even if we can’t quite pinpoint it. That much is clear.

  57. echidna says:

    Could the effect of lead be particularly important during pregnancy? Unless you have a surrogate mother, you cannot differentiate between genetic effects and foetal environment effects in humans.

  58. John K says:

    I’m not surprised that there are other, non academic, benefits displayed. I’m American and my wife is Italian. We would spend part of our time in Italy and I was interested to see that the Italian state “asilo” preschool system has almost universal enrollment for preschool-age children. I was also surprised to see that, unlike the preschools we experienced in Boston, there was much less focus on academics. In fact, the teachers would laugh when my wife would tell them about how there was a constant hope of teaching kids to read as preschoolers. Their feeling was that preschool is for teaching kids how to behave in group, sit still, eat properly, and generally get ready to learn, and there was plenty of time for academics later.
    At 8, my son moved to an Italian public school full-time. He had done well in US public and Catholic schools, but the complaint in Italy was that he needed to learn focus. After a year, he is now doing very well, and I can see that their approach had really set the stage for later rigor – he works far harder and has higher expectations than we had experienced in the US.
    (I imagine that the unremarkable PISA scores in Italy are because of the high percentage of kids who pursue vocational tracks in high school)