Monday 30 September 2013

Marriage selects for the marrying type


A current issue in UK politics is the tax status that should be accorded to the married. This is hardly a fit subject for polite company, but it means that there is a sudden interest in who gets married and who does not, and renewed discussion about whether the institution of marriage confers a benefit on its adherents, or whether it simply attracts a better sort of person in the first place. The evidence seems to favour the latter explanation, or so it would seem from a study conducted by Goodman and Greaves entitled “Cohabitation, marriage and child outcomes” produced by the Institute of Fiscal Studies, April 2010.

You may wonder what fiscal studies have to do with the institution of marriage, and whether this topic is best left to religious leaders, but governments with a penchant for taxing their long-suffering citizenry stick their fingers into the most intimate regions, transforming even the most private matters into a mesh of brigandage. There are taxes and tax benefits to marriage, to not being married, to being tenants in common, to leaving money to a spouse, to dying married, to being divorced, to having children, to not having children, and sundry other matters. You should take tax advice before so much as smiling at a stranger.

The authors have done a thorough piece of work on a familiar sample, the
Millennium Cohort Study which has graced these pages before. It is a longitudinal data set which initially sampled almost 19,000 new births across the UK between 2000 and 2002. The sample design disproportionately selected families living in areas of child poverty, in the smaller countries of the UK and in areas with high ethnic minority populations in England. As we learned before, it was a reasonably good sample until everyone started moving about, as humans usually do, and as the poorer respondents  disproportionately dropped out of the study. Tag them all at birth, make them report back forever on Twitter, and autopsy them all at death. Meanwhile, back to the study.

The authors decided to restrict their sample to those children born to couples (30% cohabiting, 70% married). It turned out that there is not much difference between the two, so one is left wondering if including single parenthood would have had an effect. A pity we won’t know this. As the authors are aware, modern relationships follow a pattern: out of roughly 10 sexual partners the last one is chosen for cohabitation; then if things go well the couple have a child and marry sometime afterwards; though some marry a little before; and then after a while if things don’t work out some couples separate and repeat the process, almost always sequentially, occasionally in parallel. 

A good point is that we have some IQ data on the verbal part of the British Ability Scale, and though at 3 year and 5 years of age we can get some good indications of ability, we lack the power obtained at age 11. There is also child development data based on parental report. As I have already revealed, there is not much evidence that marriage improves child outcomes at age 5.

“We have shown that parents who are married differ from those who are cohabiting in very substantial ways, particularly relating to their ethnicity, education and socio-economic status, and their history of relationship stability and the quality of their relationship even when the child is at a very young age. Once we take these factors into account, there are no longer any statistically significant differences in these child outcomes between children of married and cohabiting parents.” 

“much of the gap in educational and social and emotional outcomes between the children of cohabiting and married parents appears to be due to differential selection into marriage compared with cohabitation, largely on the basis of parental education and socio-economic status. These characteristics of the parents explain differences in cognitive development almost entirely, while some differences in social and emotional development remain. Much of the remaining difference in social and emotional outcomes is accounted for by differences in relationship quality between parents who are married and cohabiting when their child is born.”

The overall result of no difference (rather than the detailed findings) has been seized upon as a reason not to favour married couples with a tax break. Frankly, the tax system is so complicated that I doubt this change will make much of a difference, other than to complicate things further.

Here are the key findings drawn directly from their paper:

Just over half of mothers who are Black Caribbean are married when the child is born, compared with about 70% of mothers who are White. By contrast, almost all mothers who are Bangladeshi, Pakistani or Indian are married when their child is born.
Mothers of all religious faiths are significantly more likely to be married rather than cohabiting compared with mothers of no religion.
Both mothers and fathers in married couples are over twice as likely to have a degree as their counterparts in cohabiting couples. Married mothers are also slightly less likely to have problems reading in day-to-day life.
Fathers within married couples are twice as likely to have a professional occupation as cohabiting fathers.

Couples that are married at the time of their child’s birth are around twice as likely to be in the highest household income quintile. Married couples are much more likely to own or have a mortgage for their home.
18% of mothers in cohabiting couples first gave birth before they were 20, compared with 4.2% of married mothers, while over 30% of married mothers were over 30 at the time of their first child’s birth, compared with 21% of cohabiting mothers.
Married couples are much more likely to have lived together for a longer period of time prior to their child’s birth than cohabiting couples: over half of married couples have lived together for more than 6 years, compared with 16% of cohabiting couples. Almost 40% of cohabiting couples had lived together for less than 2 years, compared with only 8% of married couples.
Mothers in married couples are much more likely to report that their pregnancy was planned; this was the case for 75% of married mothers compared with 47% of cohabiting mothers.
There is some difference in ‘early’ relationship quality between married and cohabiting couples. For example when the child is 9 months old, 31% of married mothers report that their partner is usually sensitive and aware of their needs, compared with 24% of cohabiting mothers.
Cohabiting couples are considerably more likely to experience a period of separation of a month or longer before their child is 3 years old; this is the case for 26% of cohabiting couples and only 7% of married couples. Couples that are cohabiting at the time of their child’s birth, rather than being married, are also less likely to live together when their child is aged 3.
Mothers who are married at the time of their child’s birth and their child have slightly better health outcomes than mothers who are cohabiting. The child is slightly less likely to have a low birthweight (5.8% of children born to married couples, compared with 7.4% born to cohabiting couples) and slightly less likely to have been born prematurely (7.7% compared with 8.5%).

Mothers in cohabiting couples when their child is born are much more likely to smoke when the child is 9 months old: 41% of mothers in cohabiting couples smoke, compared with 15% of mothers in married couples. Mothers in cohabiting couples are also less likely to breastfeed their child at all, and slightly more likely to have a high score on an index of ‘mother’s malaise’ when the child is 9 months old, which is indicative of depression.

Fathers in married couples are slightly more likely to have the lowest level of involvement with their child at 9 months old, but cohabiting fathers are less likely to rate themselves as ‘good’ or ‘very good’ parents when their child is aged 3: 57% of fathers in married couples have this belief, compared with 44% of fathers in cohabiting couples. There is less difference in the percentage of mothers in each household type who believe they are ‘good’ or ‘very good’ parents.

Couples that are married at the time of the birth are more likely to have a more regular routine for their child: 46% have a regular bedtime for the child at age 3, compared with 39% of cohabiting couples. Married couples are also more likely to provide a better home learning environment at age 3 – for example, 67% read to their child daily, compared with 58% of cohabiting couples.

What are the outcomes of children born into married and cohabiting families?

By the time children are aged 3, there are already statistically significant differences in child outcomes between children born to married parents and those born to cohabiting parents. On average, children born to married parents display better social and emotional development and stronger cognitive development than children born to cohabiting parents.
The differences in children’s social and emotional development are much larger than the differences in their cognitive development at both age 3 and age 5.
The most negative outcomes for children are, on average, amongst those whose biological parents have split up, regardless of the formal marital status of the parents before they split.
The differences we observe between children born to cohabiting parents and those born to married parents are relatively small in comparison with other attainment gaps, such as the gap between children born to parents with a high and low level of education, between children born to lone parents compared with parents in any form of couple, or between parents with high and low income.

The gap in cognitive development at ages 3 and 5 between children born to cohabiting parents and those born to married parents is greatly reduced and is no longer statistically significantly negative after differences in parents’ education, occupation, income and housing tenure are controlled for. This suggests that the lower cognitive development of children born to cohabiting parents compared with children born to married parents is largely accounted for by their parents’ lower education and income, and not by their parents not being married.

The gap in social and emotional development at ages 3 and 5 between children born to cohabiting parents and those born to married parents is reduced by more than half, but remains statistically significant, once differences in parental education and socio-economic status are controlled for. This suggests that the majority of the gap in social and emotional development of children born to cohabiting parents compared with children born to married parents is largely accounted for by their parents’ lower education and income, and not by their parents not being married.

We have shown that the children of married parents do better than the children of cohabiting parents in a number of dimensions, particularly on measures of social and emotional development. But we have also shown that parents who are married differ from those who are cohabiting in very substantial ways, particularly relating to their ethnicity, education and socio-economic status, and their history of relationship stability and the quality of their relationship even when the child is at a very young age. Once we take these factors into account, there are no longer any statistically significant differences in these child outcomes between children of married and cohabiting parents.


This is a good study, carefully done and reported. It vividly illustrates the interpretive problems we encounter in research when we “control” for variables which relate to behavioural choices. For example, anyone who still smokes, and smokes while raising a child, does so in the knowledge that they are affecting their health and that of their offspring. We can control for this in the statistical sense, by calculating what their health would be like if they did not smoke, but in real life we cannot control for their behaviour (other than giving them warnings, taxing cigarettes and so on). Life involves choices, and the effects of choices accumulate along the lifespan.

The study could have been improved by using Structured Equation Modelling, though I still welcome old fashioned regression results, despite their complexity.  As is usual in contemporary social science, there is no mention of possible genetic transmission of ability. All the significant difference which impact on child development are firmly described as being due to education and status (in theory, goods which could be distributed to people, rather than emblems of attainment).

In summary, you could argue that the conclusions should be: Forget about formal marriage, and concentrate on who you shack up with: someone who is bright (measured here by scholastic attainment and professional occupation) diligent and future oriented (measured here by socio-economic status and owing property) of good character (measured here by spouse sensitivity) dependable (measured here by not dumping you within 3 years) and someone who appears to be in good health and does not blow cigarette smoke in your face. You must not mention it, for fear of being unfashionable, but you are searching for someone who comes from a good family, and must even avoid those who were abandoned by their own parents and taken into care. In times gone by this was referred to as good breeding. Now it would be called “good genes”.

It is interesting that, in seeking to attack the tax and marriage political proposal, some commentators have overlooked the actual results of the paper, which point clearly to the advantages of living a considered life, based on thought and behavioural restraint. These are high ideals, but many people manage to attain them, most of the time at least.

Although this paper is associated with the Institute of Education, it may in time achieve the status of a revolutionary tract. On reflection, a counter-revolutionary tract.

Saturday 28 September 2013

The uses of literacy


The US National Endowment for the Arts have just published their report “How a nation engages with art”. It is very much what you would expect from such an enterprise, in that it is worthy, somewhat humourless, and slightly pretentious. A nation “engages” with art. It does not just go to the movies or watch TV. It “consumes” art rather than going to the theatre or art galleries. The study will be “cogent” to arts organizations, arts funders, and cultural economists, and will “inform their understanding” about arts audiences or help them gauge public demand for “specific arts experiences”. I detect the malign influence of a committee of public arts officials. If there is anything worse than a Cromwell prohibiting people from dancing round the maypole, it is a Committee encouraging it.

For those who truly believe that a picture is worth a thousand words, then I must report the following: the frontispiece picture is of girls doing Mexican folk dancing; two European boys doing the same dance in Mexican hats; a mixed African American and European group doing the minority activity of ballet; an iPad showing a Hispanic young man, an Asian looking guy searching for books and two young African Americans playing classical string instruments. These photos are repeated at various stages. Possibly the pictures were chosen at random, but if they are intended to tell the story of arts consumption, then they are unrepresentative. Arts consumers tend to be older, whiter women. There seems to be general agreement that this group is not cool, and is best air-brushed out of pictorial existence.

Doing arty things is considered to be Good. It “rounds” the personality. So does numeracy. I know that many families have a child who wants to write poetry and dance for a living, but why should anyone spend time indulging their exhibitionism? However, Art gets an indulgent press, regardless of quality. Perhaps there is wisdom in this: folk dancing, weaving and pottery represent assurances that the participants are not currently engaged in theft or violence. One must be grateful for small mercies, even including ballet. Here is an overview of the participation rate for various activities.




Consumption is the commercial word for enjoying oneself. Literacy comes under the “art” rubric, as far as this worthy endowment is concerned. Voluntary reading is any reading done outside prison. (In fact it refers to “reading for pleasure” rather than reading documents at work or at school).

Percent of U.S. Adults Who Read at Least One Book (Any Type), by
Selected Demographic Variables: 2012
all adults                                       55%
Male                                              45%
Female                                          64%

Hispanic                                       36%
White                                           61%
African American                       46%
Other                                           49%

Women are disproportionately the readers on whom authors rely, with white readers ahead of the other groups, particularly American Hispanics. By way of comparison, this pattern holds true of visits to art galleries and museums.

Hispanic                                  14%
White                                      24%
African American                   12%
Other                                       21%

Retirement provides more time for leisure reading, as seen below, though the oldest group seems to be tiring of the activity. Perhaps they have begun to recognise the plots.

18-24                                           52%
25-34                                           55%
35-44                                           53%

45-54                                           53%
55-64                                           57%
65-74                                           61%

75                                                 52%

As always, there is a strong monotonic between reading and education, which is also, as we know, a link with intelligence. (Incidentally, this link with intellect is found in many of the other cultural activities in the survey).

Highest Level of Educational Attainment
Grade School                  22%
Some High School          28%
High School Graduate    41%
Some College                   60%
College Graduate             73%
Graduate School              82%

What do people “voluntarily” read?

Fiction       23%
Both          58%
Nonfiction 19%

At this point the summary stops. Of course, I would have been interested in more detail about the non-fiction readers, but we may have to wait until their later and more detailed report. It would also be interesting to rank the activities by popularity, and then look at the age and ethnic discrepancies. A simple prediction would be that these would diminish for the rarer activities, since they would attract the most devoted adherents, probably the uniformly brighter ones. There is partial confirmation of this hypothesis in that electronic media art users show high participation rates for all ethnic groups. Hard core is hard core, and it does not matter how many have (proportionately) dropped out on the way.

Now, some irritations aside, this publication is not all bad. At 35,000 the survey has a healthy sample size, and since it is part of the US Census it may be possible to link it with detailed demographic data. They discuss the limitations of the data in a clear and honest way. The data is publically available, with a proper data guide. A fuller report will come out next year. I am not proposing that anyone should be asked to resign, nor to make a public apology. It is just that I would have liked an artistic publication to have shown a little more evidence of artistry. They could have chosen as illustrations some of their favourite paintings of the year, identified some up and coming artists, and displayed their numeric results in more appealing ways. It reeks too much of a corporate report in which all divisions have reached their five year plans for “a stunning plurality of art forms, genres, venues, and events and activities”. It manages to transmit this feeling even when it is reporting significant drops in participation in many cultural activities.

Above all, they could have made some true confessions, and linked their explanatory photos to their basic findings. Most of art “consumption” comes via television and the internet, and the core of literary consumption is older white women visiting galleries and reading books.

Tuesday 24 September 2013

Social mobility


It is natural to note how you compare to others. In a classroom, faced with a teacher asking a question, some children will put their hands up bursting to give the answer while others will stare at the floor. In the playground some children will run fast and others slowly. There will be notable differences in fighting, crying, laughing and making friends. Some children are popular, others have no-one in particular to play with.

Run the clock forwards to the end of schooling and those differences (which will have altered somewhat because of some late maturing skills) are revealed at the next great selection. Rather than being picked for a school team activity (fearing they will be the last to be chosen) the adolescents are being picked for the best jobs (fearing they may not get a job at all). And so on, and so on, with each career step, until the peak of achievement, whenever that turns out to be. (As a rule of thumb, by the mid-40s you will have bought your most expensive house, so that is a metric of sorts).

In this way, in the great race of life, our positions are achieved and noted by others, as we note theirs. These life steps are described in terms of class and status, and the extent and pace of changes in status provide measures of social mobility.

All measures contain artefacts. Promotion in the Army is faster during a war. Soldiers die in battle, old generals are revealed as incompetent and need to be replaced, more recruits join the Army and need to be led. Similarly, economies in transition provide great opportunities. New skills lead to new businesses and new millionaires. Old industries die back. The Internet is the best current example.

In the past century the big change was from manual labour to clerical work. Interestingly, if you count social mobility only as the move from manual to clerical, then that change happened decades ago. Astoundingly, some observers still use that specific metric to say that in the UK there was lots of social mobility in the 1960s and very little now. There is certainly far less manual labour but there is much moving about from one job to another in the white collar sphere.

Writing in Prospect, Phillip Collins argues: “A boy born into the working class is no more likely to make it into the middle class now than he was in 1900. A child who is born middle class is 15 times more likely to end up middle class than a child who is born into the working class. These odds are exactly the same as they were a century ago. The boost to social mobility is a myth and so is the stalling. The truth is that Britain is a static society in which nothing has changed.”

Of course, the last line is not right. Society is not static, and much has changed, but not the rate of social mobility seen from a class of origin perspective. As Collins rightly acknowledges at the end of his article, you cannot have social mobility without one person’s rise being matched by another person’s fall. (In fact, you can fool a lot of people a lot of the time by giving them fancy job titles. We are all executive managers now). However, for real status we have to be better than someone, and worse than someone else on a socially valued metric.

The other measure of social mobility is to classify children by their parents’ occupations, determine their social class of origin, and then look at the social class they themselves eventually achieve, their class of attainment. Interesting as this measure is, it conflates two factors: the cultural impact of parenting, and the genetic contribution of parents and ancestors. These are rarely given a chance to compete fairly in the analysis, but when surrogates are used for the latter (say the child’s intelligence at 11) they are found to be very powerful.

Daniel Nettle (2003) found that a parent’s social class accounts for only 3% of the social class mobility of their children.  The ability of the individual child accounts for 13%. One could say that ability counts for four times the social mobility conferred by class of origins.

What is usually left out of the public debate is that genetics provides a testable explanation for social mobility, based purely on appointment to jobs by merit. Regression to the mean is an observable fact about the generational transmission of intelligence. On average, children are not as bright as their bright parents, and not as dull as their dull parents. The conjunction of sperm and egg does not provide precisely the same combinations each time. Exceptional genetic combinations (being very bright or very dull) are rare, and are only partially transmitted to the next generation. Throw the genetic dice often enough, and you will be likely to have an average child. Even when exceptionally bright parents have children together, they must expect that the average ability of their children will drift down somewhat to the population mean, say at least a 20% drop, possibly more depending on how you calculate the heritability estimate.

Children vary even within a family with the same mother and father, and brought up in the same home. Perhaps the current trend to smaller families makes this variability less evident, but the range of intellects within siblings is about two thirds of that encountered in the general population. It is a big range.

To put this family finding into context, first let us look at some American data from a longitudinal cohort which show the earnings of young people as they proceed from late schooling into their working careers. (Bright kids stay at school longer, delaying earning an income, but quickly make up for it).

When these aggregated data are put into a multiple regression equation comparing parental socio-economic status and the child’s IQ as predictors of their eventual income, then the standardised Beta for Parental SES is 0.1 and for IQ o.31. In this case intelligence accounts for 3 times as much variance as social class of origin.

However, one could still argue that the relationship between class and intelligence has not been teased out. A final proof of the impact of intelligence regardless of family background would be to look at the differences between siblings within families. Given the big range of intelligence to be found in an average family, this should be the final proof. If families can really bend the rules to favour their offspring, they should be able to use their family capital and connections to get all their kids into well paid jobs.

To make the matching for background as unambiguous as possible, Murray further limited the sample to pairs of subjects who were full biological siblings and who lived in the same home with both biological parents at least through the younger sibling’s seventh year. These constraints produced a sample of 3,802 individuals who comprised 2,859 unique sibling pairs. The method used was to get sibling pairs where one sibling was in the average range, and the other was above or below the average range. Above average IQ siblings were found to have excelled over their less intelligent siblings in the following ways: they had spent more years in education; they were more likely to have obtained a university first degree; they were doing work of much higher occupational prestige; they worked two more weeks in the year; and finally, they earned higher wages: In 1993, the median earnings for the average was $22,000. Their Very Bright siblings already earned a median of $11,500 more, while their Very Dull siblings earned $9,750 less. The Brights and Dulls each fell somewhere in between.

In brief, within families, brighter siblings rise in status, and even with all the string pulling in the world, duller siblings fall. It seems that family connections don’t count for as much as expected, at least not in America. Of course, rising and falling is relative to the average wage, which has usually risen since the industrial revolution, but which has had periods of stagnation.

To move towards a conclusion, in every generation families with below average intelligence parents nonetheless add some bright children to society. At the same time, every family with above average intelligence parents add some relatively duller children to society. The less bright and usually poorer families have some children who do very well, and rise in society. The brighter and usually richer families have some children who do not do well, and fall in society. Call it “intelligence mediated mobility” or “IQ churn”.

Social mobility is always there for the taking, and for primarily genetic reasons. It happens anyway, without any governmental planning. It can be stopped by repressive regimes which put barriers in the way of talented young people “from the wrong backgrounds”. It can be distorted by regimes which pick out talentless young people because they come from some chosen “correct backgrounds”.

In contrast, selecting young people by merit will maximise the chance that the best person is appointed to the job that best serves society. That will result in social mobility, as each person moves to the best job. Be warned, however, that in terms of real status, for everyone who rises there is one who falls. If status signifiers mean anything, there must be a hierarchy. A Nobel for everyone would not be a Nobel. Social mobility is good for the risers and sad for the fallers, but the real social mobility takes place when the right person is in the right job, and the whole of society rises.

Monday 23 September 2013

Maths is a man thing*


We have talked about greater male variability before, and now there is a new paper with a new result on the vexed question of the male/female ratio at the highest levels of mathematical ability. The particular interest in mathematics is that maths is based on symbolic logic, proceeds by means of logical proofs, is classifiable by broad level of difficulty, is very hard to do well, and is thus a good test of high intelligence. Whether you can get a proper solution to a mathematical problem is a matter of proofs, not just opinions. A solution to a maths problem can be wrong, but with work you might eventually get it right. As such, it has pride of place in human thought. According to Bertrand Russell  “Mathematics, rightly viewed, possesses not only truth, but supreme beauty — a beauty cold and austere, like that of sculpture, without appeal to any part of our weaker nature, without the gorgeous trappings of painting or music, yet sublimely pure, and capable of a stern perfection such as only the greatest art can show. The true spirit of delight, the exaltation, the sense of being more than Man, which is the touchstone of the highest excellence, is to be found in mathematics as surely as poetry”.

xkcd agrees, (Fields arranged by purity so it must be true (in the non-mathematical sense of true, meaning probably correct, though subject to revision, and who knows anything anyway).

Some recent publications suggested that the male/female maths ratio had fallen from 7 to 1 down to 3 to 1. This change certainly made it into the psychology textbooks, and suggested that cultural factors and educational advances were closing the “sex gap”. The new paper looks back to 1980 and finds a male ratio advantage which is greater than 3 to 1, so we seem to be back to the previous state of affairs.

Why the differing results? First, because different samples differ. Second, because there were some selection effects in the previous paper which may have accounted for the reduction in the gap. Third, because the male/female ratio always becomes higher the higher you set the level for mathematical achievement, so some of these fluctuating results may depend on how high the bar is set.  That last result may tell you all you need to know. Being very, very good at mathematics is a man thing.

But, hold on, let’s look at the paper.

Joni M. Lakin “Sex differences in reasoning abilities: Surprising evidence that male–female ratios in the tails of the quantitative reasoning distribution have increased” Intelligence . Volume 41, Issue 4, July–August 2013, Pages 263–274.

Lakin highlighted the following points: Sex differences in quantitative reasoning help understand STEM engagement. (That is, if you don’t have the maths, then science, technology and engineering jobs may be beyond you, and this will account for a good part of the sex difference in such occupations). Prior work found secular decreases in male–female ratios at high math ability. Her study found small mean advantages: female in verbal, and male in the quantitative domain. As expected, observed greater male variability was present on almost all tests. Contrary to prior work, the male–female ratio increased for high math ability over time.

Lakin reviewed data from the Cognitive Abilities Test (CogAT; Lohman, 2011, Lohman and Hagen, 2001, Thorndike and Hagen, 1984 and Thorndike and Hagen, 1992) which measures verbal, quantitative, and nonverbal reasoning abilities for students in grades K–12 in the United States. The normative samples for each edition of the test were large and nationally representative, making the data appropriate for investigations of both means and variances. Lakin examined male–female variance ratios and mean differences across grades 3–11 and three forms of the test administered in 1984, 1992, 2000 and 2011. This study focused primarily on changes in the ability distributions over time both in the overall sample and in the proportions of males and females in the highest and lowest levels of ability.

Wai et al. (2010) had previously found a sharp decline in the ratio of males to females at the highest levels of mathematical ability (as measured by the SAT) among seventh grade students. In the early 1980s, the male–female ratio among the top 1 in 10,000 performers (0.01%) was an astounding 13.5:1 (13.5 boys for every 1 girl in the top 0.01%), but declined rapidly through the decade to remain stable at about 4:1 during the 1990s and 2000s. Wai et al. found that the ratio rapidly declined with less stringent cutoffs as well: In the top 1%, the ratio started at 1.4 in the early 1980s and declined to only 1.1 in the most recent cohort (2006–2010). For Wai et al.'s measures of verbal ability (the reading battery from the SAT), the ratio for the top 5 in 10,000 (.05%) appeared to decline modestly as well, from 1.2:1 in the early 1980s to 1:1 in the 2006–2010 cohort. For other measures of verbal and writing ability from the SAT and the ACT, the ratios (initially showing greater proportions of girls) appeared to decrease slightly (closer to parity in boys and girls) during the last 20 years. ACT-Science did not show any clear trends.

Lakin draws attention to two problems with the Wai study. Those authors used the top 5% in any subject, not just in mathematics. That could have diluted the “mathematics excellence” measure such that the student were only in the top 10% of ability.  Second, and perhaps more importantly, their sample was based on students who volunteered to participate in additional testing for the opportunity to be selected for a summer enrichment program. This could have had a number of effects on the type of students who volunteer and their motivation levels relative to a random sample studied in the ordinary school context (Hedges & Nowell, 1995).

Incidentally, Lakin finds, as others have done, that white children have fallen from 80.6% of the 1984 US population, to 68.1% in 1992, to 65.0% in 2000 and currently to 55.7% in 2010. However, Lakin argues that this is unlikely to affect the sex ratios, and although this makes sense from a biological point of view, it might cause a problem from the strong social conditioning standpoint because it might be argued that “macho” Hispanic attitudes (Spanish speakers, mostly from Mexico, having risen from 6.4 to 17.6% over the same period) had affected the variability of the sexes in some way. I don’t know in what way, but if you can think of a way, let me know.

“All of the studies confirmed that, unlike the verbal domain, the differences in male–female ratios in the quantitative domain are magnified as increasingly stringent cutoffs are used. For example, on CogAT 7, the male–female ratio for the top 5% of quantitative scores was 2.02 while at the top 1% it was 2.77. Wai et al. (2010) found even more striking differences, especially in earlier years (surging up to 13.5:1 with the most stringent cutoffs in the early 1980s).7Hedges and Nowell (1995) found in the 1960s Project Talent data (the only dataset in their study with sufficient score ceilings to support such estimates) that male–female ratios were 1.3 in the top 10%, 1.5 in the top 5%, and 7.0 in the top 1%. Although the magnitude of changes in the male–female ratios varied considerably across studies, such a trend would have direct implications for understanding the low number of women observed in elite mathematics fields.”

Lakin points out that the sex ratios increase with age. Boys mature more slowly than girls. Even more importantly, the maths test have to have plenty of upper range to reveal exceptional talent. The CogAT test, whilst reasonable for normal populations does not have a sufficient range of difficult items to challenge exceptional mathematicians, who are most usually men not women. The figures in this paper might be underestimate.

That men excel in mathematics is illustrated by the results observed at Cambridge University, where all Senior Wranglers have been men, with two exceptions, Philippa Fawcett in 1890 who was not accorded her proper title because women were not allowed to graduate till 1948, and one woman thereafter in 1992. There it is in a snapshot: the awful history of women being denied their rightful place in intellectual life, and the strong likelihood that, on a level playing field the very best mathematicians are male. Lest it be thought that this is just a fenland phenomenon, since it began in 1936 no woman has ever won a Fields Medal, considered the Nobel for mathematicians.

Mind you, mathematics is hardly a normal pastime.


* Author Prof Joni Lakin comments: It’s difficult to educate the public on probability vs. categorical thinking. It’s not the way most people think or speak about gender issues, but I try to be careful in my paper about it. In other words, although overall I agree with much of what you’re saying, I would want to rewrite your claim: “Being very, very good at mathematics is DISPROPORTIONATELY, BUT NOT ALWAYS, a man thing.” People also forget that the variability hypothesis also means that being very, very bad at mathematics is also a man thing.”


Agreed. It is part of gender disputes to selectively attend to one end of the ability curve and ignore the other.

Friday 20 September 2013

Science is a lean diet

Science is not well paid. In some ways, this is a curious finding. Science has a considerable impact on society, and can transform an economy. Applied science is the backbone of technological societies. One “breakthrough” finding can leave previous technologies in the dust. A succession of “build-through” findings contribute to long term economic progress. “Build-through” is a word I have made up to describe what usually happens in scientific progress. One researcher builds upon the work of another, and eventually the body of knowledge builds up so that it pushes through a set of barriers. Cancer researchers have been looking for a breakthrough for fifty years. Meantime, they have built up a body of knowledge which has improved cancer survival rates in most cancers. Cars have been improved continuously since 1885, and science has played a large part in that process. There have been no breakthroughs. It is still the same internal combustion engine.
Science is difficult, and takes a long time to learn. A scientist probably isn’t in a position to publish independent work until they are 29 years of age. They may not get significant funding of their own (principal investigator status) until they are 40 years of age. Even then, most scientists’ earnings are likely to be very modest by the standards of similarly qualified professionals.
Why is this? In fact, many scientists working for big companies get good salaries. Some scientists become very rich. Most don’t. They work in state subsidised research labs on modest salaries (perhaps $40,000 USD) where they have reasonable job security and reasonable working conditions, but with rather few perks and no bonuses. There is some social prestige, but a constrained standard of living.
If science is as important as it seems to be, why is the pay so indifferent? Economists would say that rewards will usually match contribution if the market is working properly, and science makes a big contribution. I think that the answer is that most scientists throw away their bargaining power. They are drawn to science because they are intrinsically interested in research and are not in money. They count themselves lucky to have a job they enjoy. As the years go by they certainly notice their lack of money, but by that time it is too late to do much about it. It is rare for scientists to go on strike. Even going on a weekday protest march could mess up the lab routine, and spoil important results.
However, another interpretation is that society is not particularly interested in science, just the few bits that are immediately useful. Tax payers find science impressive in theory but dull in practice, and little of the output is actually valued. Furthermore, the individualised, cottage industry approach is wasteful, and the fact that most of the findings are made freely available cuts out a source of funding and a feedback system to guide science into socially desired research.
So, while it is natural for scientists to ask for more funding (to provide continuing jobs rather than well-paid ones) it may be equally natural for everyone else to spend their money elsewhere. Perhaps science is the freakish pursuit of a disturbed minority, and any public funding should be received with gratitude and due deference.
This note is dedicated to all lab workers who will be popping into the lab over the weekend.

Monday 16 September 2013

Religion as problem solving

Religion is a set of beliefs which people use to guide them through life. They may involve beliefs about life’s origins, about how one should live one’s life, and about what happens after death.

Beliefs are not proofs, though holders of such beliefs may imagine them to be so. They may begin as assumptions, evolve into rules for living, and embody wishful thinking and deeply held desires.

In those senses, religions are attempts to solve the problems of living. They provide a story, a rationale, a guide and a reassurance. A very human need, and a comfort in adversity.

Intelligence is negatively correlated with religious belief. Correlation is often misunderstood, because most correlations are less than unity (-0.7 between intelligence and religious belief). Real world correlations usually indicate a tendency, not a certainty. They hint at a cause, but do not prove it. Correlation is not a disproof of a causal link, and is often the first identification of a causal link, but it is merely the first step, not a proof. Correlation is not always causation, but may be a first step to identifying a cause.

The negative correlation often provokes believers, by way of disproof, to give an example of a well-known, very clever person who is also religious. Bertrand Russell was right when he made his weary observation “popular induction depends upon the emotional interest of the instances, not upon their number” (The Validity of Inference, Chp 23, Basic Writings, 1961).

Cribari-Neto and Souza have published a recent overview “Religious belief and intelligence: Worldwide evidence”


They know that intelligence positively correlates with atheism, but go on to show that intelligence impacts atheism even when they account for economic development. Religious beliefs lose strength, on average, as a country becomes richer. In most African countries the percentage of nonbelievers does not exceed 1% (Zuckerman, 2007) whereas in Sweden it reaches 64%. Kanazawa (2009) found that “each point in national IQ decreases the proportion of the population who believes in God by more than a percentage point”. Lewis, Ritchie, and Bates (2011) have shown that lower intelligence is most strongly associated with higher levels of religious fundamentalism. They also show openness negatively correlates with religious fundamentalism.

Cribari-Neto and Souza then go on to construct impact curves of intelligence on religious disbelief. This argument using beta regression models is quite complex, but to put it in plain language, once an average national IQ of 105 is achieved (global IQ is about 93) there is a strong impact on religious belief. Further increases in national IQ lead to significant drops in belief, but less strongly so. A coarse view would be that by IQ 105 the population “see through” religious explanations, and if they continue to hold on to them, it is not because of intellectual limitations.

Despite these findings, some people argue that the correlation may be influenced by cultural effects taking place at the family level, namely that some households are religious and others are not, and that somehow this explains away or reduces any conclusions which can be drawn on the basis of the correlation.

So, it is with particular interest that one turns to the issue of religious belief within families. All siblings are subjected to the same parental beliefs as part of their upbringing. Will there be difference between siblings in terms of religious belief as a consequence of differences in their intelligence?

Yoav Ganzach and Chemi Gotlibovski “Intelligence and religiosity: Within families and over time”

They studied the 1997 cohort of the National Longitudinal Survey of Youth (NLSY97). The NLSY97 is a probability sample of 8984 Americans (with over sampling of Afro-Americans, Hispanics and economically disadvantaged whites) born between 1980 and 1984. About 35% were Catholic, 26% Baptists, 29% other Protestants, and the rest from small denominations and religions. The participants came from 6819 households, 1862 of them included more than one participant. As a result 3192 of the participants came from households that included two participants and 835 came from households that included 3 or more participants (as 96% of the same household participants were siblings, they use the term “siblings” rather than the “same household members”). They find that differences in religiosity between siblings are determined by intelligence.

They say: “The current results provide strong empirical evidence for a causal link between intelligence and religiosity. The cross-sectional analysis suggests that intelligence influenced cross-sectional differences in religiosity, and that this effect cannot be explained by background correlates of intelligence. The longitudinal analysis, which focused on changes on religiosity (and therefore did not examine the effect of intelligence on levels of religiosity) suggests that intelligence drove changes in religiosity in that the more intelligent were primarily those that became less religious.”

“Our analysis also provided some insight into the process by which intelligence affected changes in religiosity over time. It suggests that when growing up, the more intelligent became less religious because of two, perhaps even three, reasons. First, they obtained more education, which in turn negatively affected their religiosity. Second, controlling for changes in education, they were more influenced by the processes of growing up than the less intelligent. And third, education tended to have a stronger effect on their religiosity than on the religiosity of the less intelligent, although the evidence for this effect is weak.”

In summary, the relationship between intelligence and the absence of religious belief looks pretty strong.
Bertrand Russell thought it “odd that modern men, who are aware of what science has done in the way of bringing new knowledge and altering the conditions of social life, should still be willing to accept the authority of texts embodying the outlook of very ancient and very ignorant pastoral or agricultural tribes.”
Of course, we should not consider it odd, merely that it is part of human frailty to pray for good things and hope for the best, including even the life everlasting. Fingers crossed may be silly, but according to Pascal’s Wager a little silliness does no harm if it is low cost, and might conceivably prevent a great harm.

May all your gods and your comforts go with you.

Saturday 14 September 2013

Intelligence, personality and the world’s largest meta-analysis


You may recall that I had already spoken about intelligence and personality in July  arguing that there was a case for bringing these two measures together, at least in the sense of looking at the correlation between the common factors in each. Intelligence is linked with a good personality, in the sense of being altruistic, agreeable, relaxed, conscientious, sociable, and open-minded, with high levels of well-being and self-esteem. I concluded “We certainly need further studies… (to assess) the possibility of unification across individual differences in cognitive ability and personality under the banner of life history theory.

Unknown to me, for the past 4 years Kevin Stanek and his research team at the University of Minnesota have been collecting studies that examine the relations between personality and cognitive ability for what appears to be the world's largest meta-analysis. The goal of their research is to provide a detailed examination of the relationships between many cognitive ability factors (e.g., perceptual speed, verbal ability, etc) and personality traits (e.g., the Big Five factors and their facets, compound traits, etc). 

The University of Minnesota produced the Minnesota Multiphasic Personality Inventory in 1939, and millions have taken the test, as I did when studying undergraduate psychology. They obviously have some cultural capital in personality studies. It is good to build on strengths.

Kevin is looking for datasets with personality and cognitive ability data. All you need to do to contribute is provide the means, standard deviations, and inter-correlations for the personality and cognitive ability variables in your data set as well as the mean and standard deviation for age, % male/female in the sample, and % in various ethnic/racial groups (if collected). Only test-assessed (i.e., not self-estimated) cognitive ability measures and self-report, rating scale measures of personality (Likert type scales, not forced-choice comparisons) for non-pathological samples of participants age 12 or older are being included in the meta-analysis. No individual-level data needs to be provided, so the confidentiality of all of your participants' data will be maintained. All contributing researchers will be acknowledged. If you would like to contribute or have any questions, please e-mail Kevin at

There are two things I like about this proposal. It will look at the correlations between intelligence and personality, and it will be based on a very large collaborative study.

Can you dig out any data you may have in your filing cabinet, or pass on this message to anyone who might have such data?

Liberty, equality, fraternity and…….


One should not get too hung up about words, particularly those related to French concepts.

In English language books neither equality or fraternity are used very much. Liberty is far more popular. Perhaps Tom Paine managed to have some influence after all, if only by transmitting one of the three concepts the French Revolution launched upon the world. After his death the word begins to fall out of English writing. It is a calamitous decline, from 0.016% down to a current 0.0025%, a 6 fold decline. Liberty has gone out the window, has been ground down to the ground.

Alarmed, I sought to console myself with democracy but this proved a disappointment. It is only a little more popular than equality and fraternity. It seems to be a World War hangover. It comes up a bit after the first round of slaughter in 1920, and rises a bit higher after the second round in 1943 to a peak of 0.0058% and declines somewhat thereafter. It was a slogan which hasn’t really caught on.

This counts as a finding, at least in the social science sense of that term. I have tracked a slogan and found a story with a moral. We have lost our liberty. Cue the credits, and a lachrymose swelling of violins.

Before jumping to conclusions, it is best to consider trivial alternatives.

Freedom is a word I rarely use
Without thinkin', mm hmm,

Taking the great thoughts of Donovan as my cue, freedom comes to the rescue. More popular than democracy but aping its general form, it surpasses liberty in 1906. These may be matters of emphasis and nuances of meaning, but freedom seems to have gained the edge by virtue of suggesting the unrestrained exercise of rights, whereas liberty is perhaps more political.

Finally, I cast about for a concept to act as a control variable, something which would put these important political and social ideas into context. My first, and rather cynical choice fits the bill perfectly. By 1800 it was a bit more popular than liberty and added to its popularity till the 1930s, from whence it had a period of decline, but is now almost as popular as two centuries ago. Political fads may come and go, but every scribbler knows what really matters, because no-one but a blockhead ever wrote, other than for money.


Friday 13 September 2013

DON’T give me a child until he is SEVEN


It is the start of the school year, so The Daily Telegraph, London, has done a cheery article about a group of “senior figures” calling for schooling to start later than the current 5 years of age. Presumably, the editors felt that a story knocking early education would enrage and discourage their readership, who had probably paid over the odds to get their offspring into education as early as possible. Indeed, most of their 5 year old darlings are probably battle-hardened mini executives, who have done toddler group, kindergarten, pre-school, violin, ballet, and can already text on their iPhones.

One of these senior figures is a Cambridge educationalist, Dr David Whitebread, who has actually taught in primary schools for 12 years, and has published a great deal on childhood play and cognition. I suspected he was providing the evidence base for the group’s proposals. I have looked at the most relevant of his papers below.

My initial thought on reading the story was that the actual age of starting schooling was not a tremendously important matter, and sometime between 5 and 7 was probably fine. Finland does well with a late start to schooling. Some countries use lots of additional tutoring at various ages, which muddies the picture. Furthermore, children mature at different rates, and have markedly different learning speeds. They have different interests and personalities as well. I also understand that for some hard pressed parents early schooling gives a very welcome child minding service, of better quality than they can afford themselves. For those children an early start at publicly funded schools probably assists them, because the alternative facilities are often poor. Other children are eager to learn their letters and their numbers at 4 or 5. People vary.

Considered calmly, this seems part of a larger question: is it worth sending children to school at any age? There is a significant home-schooling movement (that should be home-educating, but anyway) and those children would be an interesting comparison group. Another instructive group are those with awkward birthdates, which place them just out of their age cohort by one year: a lagged comparison. Finally, though slightly more weakly, one can compare school starting ages across different countries. This is slightly weaker because there is always some racial, historical, cultural or other difference which may invalidate the comparisons. Nonetheless, a good comparison of home-schooling, delayed cohorts and cross national comparisons should provide some useful data. Since the group is proposing a major policy change, one would expect their supportive evidence to be based on very large sample sizes, preferably at a national level. So, intrigued by their bold suggestion, I tried to find out more.

How strong is their evidence for this proposal?

Dr David Whitebread confirms that he is preparing a chapter on this issue, and this may well cover a lot of relevant evidence, but in the mean time his most relevant publication  is “School Readiness; a critical review of perspectives and evidence.” David Whitebread & Sue Bingham, University of Cambridge, which was prepared for the Association for the Professional Development of Early Years Educators OCCASIONAL PAPER NO: 2

The paper makes a number of statements about child development and learning, mostly based on Vigotsky’s theory of social constructionism, in which the ‘curriculum’ centres upon the interests, experience and choices of young children, which enables an holistic pedagogical approach. To my reading there are only two reference which speak directly to the age of starting schooling.

Suggate, S. (2007) Research into early reading instruction and luke effects in
the development of reading. Journal for Waldorf/R. Steiner Education, 11
(2), p.17

This paper suggests that the age of beginning to learn to read no longer has a demonstrable effect at age 14. Although the sample sizes are relatively small, and it refers to the benefits of one approach, the Waldorf/Steiner educational method in the main, this seems plausible. The representativeness of parents who chose this approach is questionable, but the result is not surprising.

Schweinhart, L. J., Montie, J., Xiang, Z., Barnett, W. S., Belfield, C. R., &
Nores, M. (2005) Lifetime effects: The HighScope Perry Preschool study
through age 40. Monographs of the HighScope Educational Research
Foundation, 14. Ypsilanti, MI: HighScope Press

The Perry Preschool study reported spectacular results in the 1970s, and made me believe that the American black/white achievement gap had been sorted out by intense early childhood interventions. The sample was 123 poor African Americans. It is only fair to say that this study is something of a battleground, with all the usual debates about what the results really prove. I will take the generous line that it proves that very early intervention can have positive effects.

The Whitebread paper does not mention the Abecedarian study, which is usually considered to have been a better controlled study. This showed a massive early effect of intense early childhood intervention (dating from when the mother became pregnant till the child was 5) and no additional benefits from extending the compensatory education till age 8.  The results faded quickly, but a very useful 4 IQ points on average have remained, and life outcomes are a bit better in the experimental group.

Both these studies are about the benefits of getting Black children into schooling long before they are 5, with no benefits after 5. These intervention studies certainly don’t suggest any advantages to waiting till 6 or 7.

Frankly, if this paper is to be the intellectual underpinning of a very significant proposed change in educational practice, it looks a bit thin at the moment. It might be absolutely right to delay school till 7, but this paper does not succeed in making that case, and if the upcoming chapter is going to provide the proof, then it would have been better to wait for it to be published.

Wednesday 11 September 2013

Heritability estimates and unexplained variance


Nobody owns unexplained variance. However, standard heritability estimates quite rightly assign the variance which cannot be explained on a genetic basis to “the environment”. For example, in twin studies genetics researchers argue that the higher similarity of monozygotic twins compared with dizygotic twins is down to the greater genetic similarity of the former. Yes, all twins have family environments in common, but the actual observed differences are due to the degree of genetic similarity, and from that the heritability estimate is derived. The rest is “environmental” with little specification as to what that includes, though some distinctions are drawn between family home environments and all other environments. Errors and shortcomings of the genetic model are conceded to the environment.

Now consider whether one can get down to more detail. For example, consider whether one could study environmental conditions so as to specify how much of the overall “environmentality” is due to specific family and school factors, and how much is not attributable to any measureable environmental factor but must be considered an error term. This drilling down to real causes has proved difficult. It is easy to come up with a list of putative causes of “environmentality”: perhaps poorer mothers do not have the time to spend with their babies, or simply do not realise they need to talk to their babies before the babies can talk back to them, or do not have books in the house, or do not read their children night-time stories, or any stories at any time, or just sit their child in front of a TV, or let them play video-games and do not take them to museums. All these are plausible to some degree.

Fostering out children to more intellectually demanding and supportive nurseries or adoptive parents can give estimates of how much can be achieved by enriched environments. There are supportive studies regarding early childhood tutoring (Abecedarian project) early schooling, extra tutoring, and adoption. How much variance can be attributed to each of these enriched environments?  The Abecedarian project boosted IQ by 4 points. Adoption may have boosted intelligence by a like amount. How much of the environmental variance is accounted for by these measures? I do not know for sure, because many of these measures are plagued by covariance confounders. The percentage of environmentality accounted for by specific procedures seems to be rather small. This makes sense, in that boosting child development requires many things to be done, but which are they?

However, we do not usually argue that that such slim pickings prove that the environment does not have much of an effect. We merely assume that something about the environment is effective, but we have not pinned it down yet. It is somewhere out there, like lead in paint, pesticides in food, and perhaps “negative attitudes” (shouting at the child)  in child rearing.

Turning to heritability estimates, it is not yet possible to show which individual genes, or which groups of many genes, bring about the heritability effect. We know that heritability accounts for variance from twin and other relatedness studies. That is not in doubt. What is in doubt is whether the individual components can be named and numbered.

However, some people regard the low rate of variance accounted for by the deeper gene hunting studies (say 1 to 3% of the variance in replications so far) as proof that heritability is only 1 to 3%. Not so. Heritability estimates indicate the total effect. Gene hunts are trying to ascertain the names and numbers of the individual genetic variants. That will take time. However, genetics have one key card: humans start being one cell big. That cell is the starting point. All other cells derive from it. Good to know the moment of creation, the Big Bang of life.

It will also take time to find the individual components of upbringing that make a difference to child outcomes. Environmental factor hunts need to be more specific. Exactly how should one bring up a child so that they do well at school and in life? One starting point is the Home Observation for Measurement of the Environment which has a correlation of about 0.2 with Wechsler intelligence for children (Molfese, DiLalla Bunce, 1997, Table 2) For a recent (2004) paper, look at

I think that the list is pretty comprehensive as regards what you can usefully observe in a home visit. There is no way that this list can be seen as being independent of the genetic quality of the parent. A good score is not an independent measure of the environment. It is a measure of how good a parent is at looking after and caring for a child (most generally, their own child). However, the detailed list is useful in specifying good and poor quality parenting in adoptive and compensatory tutoring environments.

Although any variance heritability estimates can’t account for is counted as a win for the environment, there must be an error term somewhere. By drilling down from the observed general effect to the specific causes, gene hunters are accounting for some of the variance, and that amount may increase as gene analysis becomes cheaper and more powerful.

How well are the environmental factor hunters doing? That is, how well are they doing if we require them to produce pure measures? For example, not talking to your child can be a genetic effect as well as an environmental effect. Similarly, not having many books in the house can be due to the parents having neither the capacity nor the interest to read books. The lack of books in the home can be the emblem of a lack of intellectual curiosity, rather than the cause of it.

There is an error term, and while the “genetics” explanation can be shown to be requiring improvement, so can the “environment” explanation if it is held to the same standards.

No-one owns the unexplained variance. 

Tuesday 10 September 2013

Childhood adversity and lifespan: Authors reply


In my post on the above paper I made a number of criticisms, including that the paper should have been entitled “Adverse parenting and premature mortality”. My argument was that the family-related adversities were not truly external, and that genetic effects should have been considered. Read the whole thing here:

The corresponding author Michelle Kelly-Irving has kindly replied on behalf of all the authors, and I think it only fair to give this top billing as a separate item, so that subsequent comments can be appended as readers wish.


I don’t agree with you about the ACE measure being a proxy for parenting (or bad parenting). The measure aims to capture events that are likely to be stressful to a child, and that may have a permanent impact on their subsequent physiological reactions to stressful conditions. I agree that parenting is certainly implicated. However, should a child be separated from either parent through death or divorce is certainly not down to inadequate parenting. Similarly, a child who ends up in care is likely to be going through a stressful phase of their lives, but I’m not sure we can make the assumption about the parenting they received beforehand. It would make more sense to actually try to measure elements of parenting, which is certainly worth doing. There may be some variables in the NCDS that allow for this.


At age 11 any measure of IQ is hugely confounded with socioeconomic and psychosocial circumstances. I am not clear as to what IQ would be measuring here – or ever, in fact.


The point that a prior variable linked to both ACE and premature mortality that we do not know about, or cannot take into account, is of course possible. That variable could be a genetic one. Genetics analyses using GWAS are not coming up with solid explanation for heritable traits. It is most likely a much more complex process of heritability than a mendelian-style genetic one, if there is something going on heritability-wise.

David Landes, economic historian


In January this year I wrote about the wealth of nations, including a paragraph on David Landes, an economic historian whose work I greatly admired. Today I have been passed on the sad news that he is dead, and I wanted to draw attention to some of his conclusions.

The brief mention I made of him in my original post is shown below.

Smith’s account (on economics) has not been bettered, but it has been extended by many economists. One particular economic historian with a great respect for geography, David Landes dared to update the text (1998) in a scholarly masterpiece which he entitled “The Wealth and Poverty of Nations”. His conclusion, stated with some reluctance, was that the difference between rich and poor countries was based on culture. He went no further than that, and certainly made no mention of differences in intelligence or personality.  However, his key observations were that successful societies innovated, and taught their children how to manage those innovations. This is intellectual problem solving in all but name.

The full post, which gives all the context, is in the link below.

David Landes was at pains to be a good communicator. He wrote short words in short sentences. He was too much of a scholar to boil down his key points on one slide, so I did that for him, on a single page I had up on my office noticeboard. It confused many of my psychiatric colleagues. It was drawn from page 217 of “The Wealth and Poverty of Nations” Little Brown and Company, 1998. ISBN 0-316-90867-3

Ideal case for material progress and general enrichment:

1 Master new technology. Build, manage and operate instruments of production.

2 Teach this knowledge to the young, by formal education or apprenticeships.

3 Chose people for jobs by competence and relative merit; promote and demote on the basis of performance.

4 Afford opportunity to individual or collective enterprise; encourage initiative, competition and emulation.

5 Allow people to enjoy and employ the fruits of their labour and enterprise.


Gender equality, no discrimination on the basis of irrelevant criteria, preference for scientific rationality over magic and superstition.

Political and social institutions which favour progress and enrichment

a) Secure rights of private property, the better to encourage saving and investment.

b) Secure rights of personal liberty.

c) Enforce rights of contract, explicit and implicit.

d) Provide stable government, of laws rather than men.

e) Provide responsive government, that will hear complaints and make redress.

f) Provide honest government: no rents to favour or position.

g) Provide moderate, efficient, ungreedy government, holding taxes down, giving back surpluses and avoiding privilege.

What interested me, from a psychological point of view, was that Landes was reluctant to talk about culture making a difference between one set of peoples and another. He was somewhat apologetic, saying that culture had sulphurous connotations. Odd. One should look for explanations wherever one can find them, and evaluate their contribution energetically. Culture is a collection of habits, both good and bad, which have developed over time in response to problems, and contain solutions of varying levels of appropriateness and success. People compare cultures, and pick up the ideas and products which they find useful. The wheel took a long time to invent. Some cultures never achieved it. Those that did gained an advantage. Some cultures are better than others in prospering in competition with each other.

Later work has supported the idea that culture makes a difference, and that people make cultures. For example: “Cognitive Capitalism: The Effect of Cognitive Ability on Wealth, as Mediated Through Scientific Achievement and Economic Freedom
Heiner Rindermann and James Thompson DOI: 10.1177/0956797611407207


I picked this figure rather than the more complicated and complete Fig 4 simply for purposes of exposition. I think that any group of people who had an eminent scientist in the past two thousand years can claim to have a culture, and it is good to see the eminent-scientists measure of culture making a significant contribution to current attainments in science, technology, engineering and mathematics, and thereby a contribution to Gross Domestic Product.

Finally, on a wider but certainly currently pressing issue, at a time when our global economy is in one of its recurrent swirls of turmoil, and economists are propounding confident schemes of improvement, it is salutary to recall that Landes ended his magnum opus thus: “The one lesson that emerges is the need to keep trying. No miracles. No perfection. No millennium. No apocalypse. We must cultivate a skeptical faith, avoid dogma, listen and watch well, try to clarify and define ends, the better to choose means.”

Sunday 8 September 2013

A modest proposal for neuroscience research


In “Three mild suggestions for Psychology” 13 August I made three suggestions which I summarise below.

1 Agree upon some basic measures. It would be alarming in any other discipline if, after a century of enquiry, we still had no agreement about what psychological measures we should apply as a matter of course.

2 Agree to pay attention to previous research, and relevant research in related fields. Psychology has an alarming tendency to ignore previous work, particularly when terminology changes, which is frequently. It often also ignores problems with measuring techniques, as if it were optional to attend to these matters.

3 Agree to collaborate, and move from sole practitioners in cottage industries to more systematic large scale research projects. Academic advancement is based on “making a name for one’s self” which encourages apprentice piece publications, repetition of papers each dealing with sub-sections of a data set, and anything which brings attention to a person.  Imagine if the smallest publishable sample size was 500 persons: might that drive up the representativeness and reproducibility of results?

Earlier, on 18th April, I had made some gentle criticisms in “Not all neuroscience is rubbish, just 92% of it” which concluded:  “Next time you see a pretty MRI picture of the brain, look at the sample size, the sample representativeness, the protocol and the statistical assumptions before believing a single pixel of it.“

By the end of August both prayers had been answered. The proposal below is smart, feasible, and has great potential. Can we publicise it and help implement it?

Cogn Affect Behav Neurosci. 2013 Aug 28. [Epub ahead of print]

How to produce personality neuroscience research with high statistical power and low additional cost.

Mar RA, Spreng RN, Deyoung CG.

Department of Psychology, York University, 4700 Keele St. W., Toronto, ON, M5T2X5, Canada,


Personality neuroscience involves examining relations between cognitive or behavioral variability and neural variables like brain structure and function. Such studies have uncovered a number of fascinating associations but require large samples, which are expensive to collect. Here, we propose a system that capitalizes on neuroimaging data commonly collected for separate purposes and combines it with new behavioral data to test novel hypotheses. Specifically, we suggest that groups of researchers compile a database of structural (i.e., anatomical) and resting-state functional scans produced for other task-based investigations and pair these data with contact information for the participants who contributed the data. This contact information can then be used to collect additional cognitive, behavioral, or individual-difference data that are then reassociated with the neuroimaging data for analysis. This would allow for novel hypotheses regarding brain-behavior relations to be tested on the basis of large sample sizes (with adequate statistical power) for low additional cost. This idea can be implemented at small scales at single institutions, among a group of collaborating researchers, or perhaps even within a single lab. It can also be implemented at a large scale across institutions, although doing so would entail a number of additional complications.

[PubMed - as supplied by publisher]
In conclusion: Let no-one doubt the power of this blog to shape the course of science.

Are girls too normal? Sex differences in intelligence


Here is a snapshot from Ian Deary’s lecture  “Ten Quite Interesting Things About Intelligence Test Scores” (Teaching Intelligence which aims to show the difference in the distribution of boy’s and girl’s intelligence scores. The point of interest is that even when boys and girls have exactly the same levels of intellect, differences in the standard deviations have considerable impact at the extremes of intelligence. Girls are more normal, boys are more extreme, so there are more boys at extremes, and the more extreme the extremes, the more boys. This finding at age 11 may go some way to explaining the sex ratios observed in later life occupations, or perhaps only partially so.


If, like me, you find the graphic a little difficult to understand, look at the new version below.



Starting from the left, you can see that those children who have an average IQ of 60 (very low) are 58% boys and 42% girls. At average IQ 100 there are 48% boys and 52% girls. At IQ 140 there are 58% boys and 42% girls. I have also increased the size of the Y axis, so that you can see a 40 point range of percentages from 30% to 70% rather than the 20 point range in the original, which made things look too dramatic. Both figures deliberately use zero suppression (don’t show the full 0 to 100 percentage range) to exaggerate the visual effect, which is legitimate, because otherwise the effect would be very difficult to see. If you just stick one IQ distribution over the other it is hard to see the difference.

In both figures you must remember that there are many more children in the middle ranges (perhaps we should show very wide bars here), and very few at the extremes (perhaps we should show very thin bars). Girls are normal, boys somewhat less so. Very bright children are more likely to be boys in the ratio 58:42.

This shows that girls predominate over boys at the centre of the intelligence range, but there are fewer at the extremes, which is where boys are slightly more likely to be found.  Assuming that the boy/girl difference is sustained in adulthood, then if men and women are promoted to jobs entirely on merit, and work equal hours, there will be 58 men to 42 women in the elite occupations.

Of course, this finding is also germane to the genetic versus social conditioning debate. It is easy to argue that restrictions in society affect all women, and that these barriers push mean abilities downwards. However, it turns out that there were no sex differences in average intelligence even in 1932, and not thereafter, despite many fears to the contrary. Girls and boys have the same average intelligence, but not the same standard deviation. (Whether men and women have the same average intelligence we can debate later).

However, it is less easy to give a social conditioning account of why girls’ standard deviations would be more narrow, because it would involve proposing a force capable of simultaneously stopping girls from being very bright and stopping them from being very dull, and being able to do so in such a way that it showed up in a test taken by children working on their own without teacher or class input. How would one simultaneously hold back a bright girl from giving a bright answer and stop a dull girl from giving a dull answer? Furthermore, how would one do this so subtly as to narrow the girls’ distribution of abilities only very slightly?

It is also difficult to explain how social conditioning could affect the standard deviation of brain volume. The sample size of 89 normal subjects is too small, but total brain volume as measured by MRI reveals suggestive results. The 42 males have a total brain volume ml of 1354 ± 111 and 47 females a total brain volume ml of 1215 ± 105.  Reite et al. BMC Psychiatry 2010 10:79   doi:10.1186/1471-244X-10-79. By the way, I am not making a brain/intelligence association here (though there is a significant positive correlation) just a link with the common findings of sexual dimorphism.

Darwin regarded greater male variability as a fact, though of unknown cause. Greater variability suggests, but does not prove, a genetic cause. However, since the effect is only visible at the extremes it is important to ensure representativeness of the the study samples. The most representative sample is an entire population to make sure that neither dull nor bright children have been lost to the testing procedure. Wendy Johnson, Andrew Carothers and Ian Deary have the full population samples from Scotland in 1932 and 1946.  “Sex Differences in Variability in General Intelligence: A New Look at the Old Question.” Perspectives on Psychological Science 2008 3: 518 . I have used their data in this posting. They also have a good discussion about whether this finding is sufficient to explain observed difference in men and women’s occupations and achievements.

The male to female ratio in intellectual demanding occupations is higher than the ratio observed at age 11. Part of that is down to the fact that measurements at age 21 may be more representative of adult levels. There is some suggestion that girls develop faster than boys, and the relative advantage reverses in favour of men as they reach maturity.  Girls have an early advantage in height, which then reverses in later development. Part of the occupational difference in later life is that many men are much more likely to put in long hours, and to value their careers and their status above other interests. Matched for time invested, the differences are smaller, and sometimes favour women. Importantly, the sexes have different interests. We shall return to all that later.

Of course, the great Dr Heim knew all about this 40 years ago, based on undergraduate intelligence tests. The title of the relevant chapter may be judged too inflammatory for current sensibilities, but it was said in jest.

Saturday 7 September 2013

Childhood adversity and lifespan


Here is a study which deserves attention, based on an excellent sample which is dear to me and to many social scientists: the 1958 National Child Development Study (NCDS) which included all live births during 1 week in 1958 (n = 18,555) in Great Britain. If any dataset deserves a monument, this one does. For my generation of psychologists it provided a steady flow of publications on child development, so it is natural to regard it as one of the family, a familiar companion. Their 1972 “From Birth to Seven” had a great impact, serving as a benchmark for the times. Call it the thinking person’s “Seven Up” (a film which tracked a few children).

Now a new team have looked at this old warhorse again, and compared each child’s  documented early childhood “adverse experiences” with their lifespan, finding that real adversity has shortened lifespans. More specifically, a total of 3.2% of all the sample were dead before 50 and for the boys, where the adversity effect was strongest, the low rate of death doubled for those who had two or more adverse events. (The great advantage of a contemporaneous record of these experiences is that we do not have to rely on faulty and possibly biased recall in adulthood). The finding makes sense, from a trauma point of view. The child has been subjected to a dreadful event, the trauma has embedded itself during a sensitive period in development, and as a consequence children’s resources have been diminished,  and it all comes out in the end, by causing a somewhat premature end. An unpleasant finding, but a salutary one, since it shows how vulnerable children are to external events.

The effect is only slightly diminished by controlling for the effects of social class and education. In summary: an impeccable sample, distinguished authors, appropriate statistics, and a clear exposition of the effects in copious tables. It is natural that UCL should choose to mention it in their weekly newsletter, which is where I found out about it. As someone who worked for a long time in trauma, the paper is particularly welcome. Case proven.

Before turning away, have a look at the list of adverse events.

1. Child in care: child has ever been in public/voluntary care services or foster care at age 7, 11 or 16.

2. Physical neglect: child appears undernourished/dirty aged 7 or 11, information collected from the response from child’s teacher to the Bristol Social Adjustment Guide.

Household dysfunction, as described by Felitti et al. [16], is a dimension of adversity consisting of four categories each contributing to the score:

3. Offenders: The child lived in a household where a family member (person living in the same household as the child) was in prison or on probation (age 11 years) or is in contact with probation service at 7 or 11 years; the child has ever been to prison or been on probation at 16 years.

4. Parental separation: The child has been separated from their father or mother due to death, divorce, or separation at 7, 11 or 16 years.

5. Mental illness: Household has contact with mental health services at 7 or 11 years; Family member has mental illness at 7 and 11 or 16 years.

6. Alcohol abuse: Family member has alcohol abuse problem at 7 years.

About 5% to 7% of the total sample had experienced these events, which put them in what we might call the troubled minority.

Now comes the question: which of these events are truly external? To prompt you, in case you find the question hard: what sort of parents end up with their children in care; physically neglect them; get put in prison; have mental illness; abuse alcohol; and cannot stay together when their children are young and need them? Might they conceivably be parents who are not passing on the very best genes to their offspring? Far from being external, I see these events as arising from bad parenting, and one very vivid possibility is that the offspring have been subjected to a double burden: bad experiences, and genes which make them vulnerable. Even if the authors find it hard to countenance the notion of the heritability of mental illness (see previous post below on “Bad Blood”), one expects some comment.

So, how do the authors deal with this possibility? They do not mention it.

Here is what they say about “prior confounders”: “To examine the relationship between ACE and premature death, prior confounding variables potentially associated with both ACE and mortality were adjusted for in the multivariate models. Among the variables available at baseline, collected from the cohort member’s mothers via a questionnaire at birth, we identified those most likely to be social or biological confounding factors. Household and parental characteristics were included: mother’s age at birth, overcrowding (people-per-room), mother’s partner’s social class (manual/non-manual), and if this was unavailable the mother’s father’s social class was used, mother’s education level (left school before/after minimum leaving age), and maternal smoking during pregnancy (no smoking/sometimes/often/heavy).”

As you will see, there are many things about the parents which they do not know, presumably because they were not asked in the original survey. Social class and education are certainly a help, but they are rather broad categories, and hardly a proxy for biological markers.

What else could these authors have done? Well, they could have looked at intelligence measures, since lower intelligence is associated with shorter lifespans. The children’s intelligence was measured at 11. It does not cover all the possible biological markers, but is associated with it. Finding no association at all would strengthen the trauma explanation.

Gale et al. included intelligence in a study on this sample’s smoking rates, and found “women with lower IQ in childhood were more likely as adults to smoke during pregnancy and to be a smoker currently. Structural equation modelling showed that the effects of childhood IQ on smoking behaviour were indirect, as they were statistically mediated by educational attainment and age at first birth. There was some effect of educational attainment and age at first birth on smoking behaviour over and above the effect of intelligence.”

If the data are there, why not use them? The authors of the current paper did take measures of education, social class and health at age 23 and they say “in our model, these adult variables at the age of 23 are a proxy for behavioural patterns in early adulthood and controlling for them serves as a first step to understanding possible mechanisms.” Controlling for those did reduce the hazard ratios somewhat.

What could they have done? I assume that in this study the age at death of the parents is known. If the pattern of premature death is as strong in the parents as in their children (for the sample as a whole, not just the adverse event group) that would be germane, and would weaken the trauma-leads-to-death argument.

In summary, I think that a few more checks would potentially strengthen this study. The effect they have found seems noteworthy, but they could at least have mentioned a genetic explanation. In terms of exploring alternative hypotheses,there are so much data in this goldmine of child development that it would be a pity not to complete the picture.

There is one other thing the authors could have done. They could have changed their title to: “Adverse parenting and premature mortality”

Good sample, though.