A lot of this on the surface sounds confounded and p-hacked to all heck. With maybe the girl's exposure to boys being an exception (but a lot is riding on the "it’s random enough that you can compare children who happened to be more isolated from the other main sex"). The main thing I would want to know is if they were much more careful in addressing this than most other treatments on the subject.
Yes, the claim that Dunedin school assignment is quasi-random is a big one. At least in the USA, I know that one of the biggest appeals of sex-segregated private schools to parents is precisely the belief that it'll reduce misbehavior...
I'm also very concerned to see that there was apparently an entire candidate-gene chapter (this was written in 2020, not 2010, right?). I hope this is some of the MAOA/B work which is real, as opposed to the 99% which is bogus; if not, I thought Belsky knew better.
I'm from Dunedin and went to highschool there (in the 2010s), so I guess I can speak to this a bit.
Co-ed schools were generally lower decile (=lower socio-economic backgrounds) than the single sex schools (here is data taken from wikipedia on this). The selection based on 'ease of walking to school' is still a (small) factor, but I expect this would have been a larger factor in the 70s when there was worse public transport. In some parts of NZ school zoning is a huge deal, with people buying houses specifically to get into a good zone (especially in Auckland) but this isn't that much the case in Dunedin.
Based on (~2010s) stereotypes about the schools, rates of drug use seemed pretty similar between co-ed and all-boys schools, and less in all-girls. And rates of violence were higher in all-boys schools, and less in co-ed and all-girls. But this is just my impression, and likely decades too late.
It says "In New Zealand, the choice to attend a single-sex school is not a result of a family’s desire that their child attend a religious or military institution; choice is primarily determined by which school the pupil can most easily walk to." Looks like about 20% of government-run secondary schools are currently single-sex; not sure what it was like in the 70s or so when this was done. But I could imagine that in cases where parents particularly want a particular school they still chose based on things like whether it was single-sex and not only on what was closest.
Yes, the book was published in 2020. The parts about genetics emphasize that at the time they did the research (significantly earlier), testing was a lot more labor-intensive and expensive which is why they used a method that nobody would use now. The authors' paper about the MAOA stuff came out in 2002. But if the method is also now considered to be mostly bogus, I wonder if they couldn't resist publishing research that they'd already done even if the method wasn't considered good anymore.
"In New Zealand, the choice to attend a single-sex school is not a result of a family’s desire that their child attend a religious or military institution; choice is primarily determined by which school the pupil can most easily walk to."
Or perhaps the parents' choice of where to live is influenced by proximity to the schools they prefer. Certainly in the UK, proximity to good schools is a big influence on house prices.
"In New Zealand, the choice to attend a single-sex school is not a result of a family’s desire that their child attend a religious or military institution; choice is primarily determined by which school the pupil can most easily walk to."
I mean, sure, it makes lots of sense that walking distance would be a core mediator, but what evidence do they have that walking distance is determined by chance, rather than being influenced by the family's desired school or resources?
I looked at the chapter on bullying and I found the methodology weak, given the huge inherent issues with passive observation.
It is really really hard to "control" for other factors and their efforts were quite lame. Several particular problems appear. First they correct for other factors one factor at a time. This is a failure mode when multiple factors are relevent at the same time e.g. IQ and poor parenting. Second they make no allowance for errors in measurement of factors. As one example they correct for childhood IQ to exclude IQ as a factor that may both lead to being bullied and that may be harmful independently of bullying. But they do not correct for measurement error. Any measurement of factor X will have error and thus tests based on the measurement will understate the effect of the actual factor X. In the particular case of IQ, childhood IQ is not very highly correlated to IQ in adulthood and thus is a poor proxy for IQ in adulthood. It is also poorly correlated to parental IQ and thus heavily fails to capture effects operating via shared IQ genes in the parents.
See Judea Pearl's book "Causality" for a detailed discussion of these issues and what a proper causal model looks like.
I would not base any serious decison on the findings in the book. They are as likely to be anti-knowledge as real knowledge IMHO.
I am coming around to the view that any study whose methods are not prepublished should be assumed to be p-hacked.
Cross-posted from Otherwise, where I'm writing these days.
Other book note posts on books about children: The Anthropology of Childhood, Don't Shoot the Dog
The book: The Origins of You: How Childhood Shapes Later Life by Jay Belsky, Avshalom Caspi, Terrie E Moffitt, Richie Poulton
[edit: see comments for useful caveats; I probably wasn't skeptical enough of the book while reading.]
Main idea
The “dual-risk hypothesis” appears in a lot of their research: “an underlying or latent 'vulnerability' that puts one at risk of developing a problem, but according to this hypothesis, the problem only arises when a particular stress is experienced or encountered.”
A lot of the findings in the book have some element of this – some risk factor (situational or genetic) is associated with worse outcomes only when it’s combined with some other risk factor (usually related to parenting: insecure attachment, or harsh or inattentive parenting.)
Personally my takeaway from this book is that warm, attentive parenting, especially in the first three years, is protective against a number of bad outcomes. And it’s under my control in a way that some other things are not.
About the book
I’m not covering most of the findings, just the ones I happened to take notes on.
The authors talk a lot about methodology. At first I loved this – instead of just saying “here’s a finding” they break down a bunch of research decisions they had to make. You definitely come away with a picture of why it’s hard to do good social science research.
But I was less interested in this as the book went on – each chapter was largely about the research process, and didn’t get to findings until 2/3 through each chapter. After a while I just started skipping to the findings in each section.
People have strong opinions about Belsky, the main author, and he might need to be taken with a grain of salt. He says people don’t want to listen to him because he’s taken a politically incorrect stance that daycare has disadvantages. Others say he’s overstating the conclusiveness of his research because of his own ideology.
They follow three major observational studies:
The Dunedin study is an exciting source because it’s continued tracing the same people for almost 50 years now. It also started with a representative population (all the three-year-olds in town), rather than people whose parents had the free time to enroll them in a study, etc.
Caveats
Associations aren’t destiny. They do a fine job of explaining this, for example “Even though maltreatment increases the risk—a statement of probability, not certainty—of later criminality by about 50 percent, it remains the case that most maltreated children do not become delinquents or criminal adults.”
A lot of the measures seem like they might have been gameable by the families: when you observe parents interacting with children, won’t they be on their best behavior? When you ask parents about spanking or teenagers about their drinking, will they answer honestly? I’m sure there’s some social desirability bias here where people say and do what they think will look good. But there’s enough variation that they’re clearly not all doing this to the full extent they might, so there’s still something to study. Parents in the 1970s were often pretty frank about physical punishment. “One child, when asked if he had ever started a fight and tried to hurt someone… responded, “Only Catholics—do they count?””
The authors pay a lot of attention to class as a confounder, which is great as this is often not controlled for. But weirdly they never mention ethnicity. The population of Dunedin, New Zealand is of 11% Maori descent. I can imagine there might be significant difference in childrearing practices, etc.
They found relationships between several variables: quality of care, quantity of care (number of hours per week), attachment style as measured by the Strange Situation experiment, and maternal sensitivity to the child (as measured by observing the mother and child interacting).
They find that both quality of care and quantity of care matter. There are a bunch of effects they look at, concluding that “there remained evidence of both good news and bad news when it came to the effects of day care, with most of the good news reflecting the developmental benefits of good-quality care for cognitive development and most of the bad news reflecting developmental risks to social and behavioral functioning of early, extensive, and continuous care.”
BUT both positive and negative effects are small in magnitude. Type and quantity of daycare doesn’t make that much difference, as far as they can tell. “The kind of family that a child grew up in proved to be much more developmentally significant than his or her day-care experience.”
There’s some interesting dual-risk stuff here:
“Children who spent more time in child care (in a center, a family day-care home, or by a nanny), indeed those who averaged just ten or more hours per week across their first fifteen months of life (risk factor 1), were more likely than other children to develop an insecure attachment to their mother, as seen in the Strange Situation at fifteen months, if—and only if—they also experienced insensitive mothering.” “when infants spent more time in nonmaternal care . . . mothers were less sensitive to their infants when observed interacting with them . . . Conversely, less time spent in child care predicted more sensitive mothering”, both of these across ranges from 6 to 36 months.
“More time spent in child care and low-quality care each (separately) amplified the risk of insecurity associated with having a mother who proved to be rather insensitive in the way she interacted with the child.”
They measured self-control based on things like restlessness and low frustration tolerance in observations at age 3 and 5, and reports from parents and teachers of things like fighting, distractibility, and impulsivity.
Even after controlling for childhood socioeconomic status and intelligence, “adults with limited self-control in childhood grew up to have more cardiovascular, respiratory, dental, and sexual health problems, as well as more inflammation.” At age 32, they had more problems with drugs and alcohol, more criminal convictions, less savings, and more money and credit problems. There was a dose response: “a little more self-control resulted in slightly better outcomes, having moderately more self-control predicted moderately better outcomes, and having still more self-control forecast still better outcomes.” This was still true after removing the children with an ADHD diagnosis.
The authors say that basically all five-year-olds are impulsive and aggressive to some degree. Most children grow out of this, but some don’t so much. “Children living in high-SES [socioeconomic status] and middle-SES neighborhoods experienced a normal trajectory of improvement in behavior — reductions in antisocial behavior from ages five to twelve years. Thus, in contrast to all other children, boys growing up in “deprived” neighborhoods experienced no improvement whatsoever in delinquent and aggressive behavior during this time. As twelve-year-olds, they were still behaving badly, as though they were five-year-olds.”
BUT when you control for maternal behavior the neighborhood effect disappears:
“It was because mothers in more disadvantaged neighborhoods proved less warm and supportive in their parenting and monitored their children less that their children proved to be more antisocial in their behavior. In fact, when these parenting processes were taken into account statistically (that is, controlled for), the previously detected effect of neighborhood disadvantage on children’s delinquent and aggressive behavior completely disappeared! It was only because of the adverse effects of neighborhood on parenting, then, that neighborhood disadvantage predicted and presumably influenced children’s antisocial behavior. . . . This means, of course, that when parents in disadvantaged families embedded in deprived communities provide greater warmth and monitoring than would otherwise be expected given the makeup of a neighborhood, their children are less likely to become antisocial.”
The warmth finding seems like it could be just a result of flawed methodology – this is from the twin study, where they ask mothers to speak for 5 minutes about each of their twin children at age 10 and score based on her vocal tone, facial expressions, and expressions of empathy and sympathy. If one of your children is already a delinquent, it’s not surprising you might seem less warm when talking about them.
But the monitoring thing seems valid. They asked “whether the mother knows the friends the child hangs out with, knows where her child goes in his spare time, whether the child needs permission to leave home, and whether the mother knows what the child does while the child is outside the house.” I expect some confounding here, but it probably understates the finding if anything: if your kid gets in trouble a lot at age 10, I’d expect that to cause you do more monitoring of their behavior rather than less.
These researchers tested the theory of slow and fast reproductive strategies. The theory is that girls in safer, more resource-rich environments might reach puberty later – a “strategy” (biological, not consciously planned) which delays childbearing. Girls in higher-stress environments might reach puberty earlier (which increases your chances of passing on genes in a harsh environment). Their findings supported this:
“First, girls growing up without fathers sexually matured earlier than girls growing up in intact, two-parent families. Second, girls growing up in high-conflict families matured earlier than their counterparts exposed to minimal family conflict.”
“Parents who were characterized as harsh spanked their children for doing something wrong, and they expected them to obey without asking questions and to be quiet and respectful when adults were around. Harsh parents regarded respect for authority as the most important thing for the child to learn. They also believed praise spoiled the child, so they provided few hugs and kisses. Thus, what we discovered was that the more parents regarded and treated their four-and-a-half-year-old harshly, the earlier their daughters had their first period. This was the case regardless of the age at which mothers themselves had their first period.” The difference here is not huge, we’re talking like 9 months difference on average.
Body fat also affects age of puberty – malnourished and athletic girls start menstruation later. But the difference in environment persists even when you control for weight.
“By age fifteen, girls who matured earlier had engaged in more sexual risk taking than other girls. The early-maturing girls who experienced harsh parenting not only had their first period before other girls but by age fifteen were more sexually active.” The difference persisted: “Early-maturing girls had more sex partners than other girls, into their thirties.”
But once again they wondered if attachment mediated this:
“The already established accelerating effect of a harsh rearing environment at age four and a half years on early pubertal development did not apply in the case of girls who were securely attached to their mothers at age 15 months; it only did so when daughters had established, as infants, insecure attachments to their mothers. These results are graphically displayed in Figure 7.2. Thus, a secure attachment operated as a resilience factor, preventing the acceleration of pubertal development when it would otherwise have been expected.”
Boys as a risk factor to girls
“It didn’t take a lot of insight to suspect that boys—perhaps especially older, “bad” boys—played a big role in leading early-maturing girls into temptation. Consider the sort of boy who would be attracted to an early-maturing girl, whose figure is voluptuous but who is still cognitively and emotionally a child.”
They use a cool methodology – in Dunedin the local school might be single-sex or might be co-ed, and it’s random enough that you can compare children who happened to be more isolated from the other main sex. New Zealand law didn’t allow asking teenagers questions about sexual activity, but they looked at other norm violations.
“At age thirteen early-maturing girls in mixed-sex schools were much more likely to break rules—that is, engage in norm violations such as stealing money, going to R-rated films, and getting drunk—than those in single-sex schools, and this difference in type of school did not emerge in the case of girls who were maturing on time or later than other girls. . . .
In other words, when early-maturing girls—the very kind we had previously found to be more likely to engage in sexual risk taking—found themselves in a boyless school environment, they were not led into temptation.”
From the chapter on “Child Maltreatment, Genotype, and Violent Male Behavior”
[Edit: looking at this history of this research, the authors published it in 2002 and some attempts to replicate it have done so, while others have failed.]
They thought environmental factors might explain why studies of genes and psychiatry often failed to replicate. “We reasoned that this inconsistency across research studies might occur if the participants in these investigations differed in the social or environmental causes of their mental disorder.”
At the time genetic studies were much more difficult and expensive, so they chose one gene to study, the monoamine oxidase A gene or MAOA. “Genetic deficiencies in MAOA activity—that is, low levels of the enzyme resulting from a particular variant or version of this gene—had already been linked with aggression in mice and humans by other investigators.”
So they looked at violence in men with different MAOA variants and different levels of mistreatment in childhood.
Environmental factors they considered mistreatment:
About 25% of the boys in the Dunedin study had one of these risk factors (“probably mistreatment), and 10% had two or more (“severe mistreatment”). When the boys were grown up, the researchers assessed their level of violence using
“While antisocial behavior is greater the more certain it is that the study member was subject to maltreatment as a child (that is, severe>probable>none), this dose-response relation proved especially pronounced in the case of individuals carrying the low-activity rather than high-activity MAOA gene variant, just as predicted by the diathesis-stress / dual-risk thinking that informed our work.” Another dual risk situation – the genetic risk was only relevant when the boys were also mistreated in childhood: “If study members carried the “risk” gene— the low-activity MAOA variant— but were not subject to child maltreatment, they were no more likely than those carrying the low- risk genetic variant to engage in violent or otherwise antisocial behavior.”