Two ambitious attempts to measure the combined roles of a wide variety of “heritable traits besides IQ” indicates that those other traits do play the significant role in combination that intuition says they should.
The first was produced by an 11-person team of psychologists, mostly from England, using 6,653 sets of twins in the UK Twins Early Development Study (TEDS). The first author was Eva Krapohl.
The dependent variable was the participants’ scores on the national examination administered at the end of compulsory education in the UK at age 16. It is known as the GCSE (General Certificates of Secondary Education). The independent variables involving heritable traits were IQ and eight other domains represented by composite scores: personality (including the Big Five and grit), self-efficacy, well-being (e.g., happiness, hopefulness), parent-reported behavioral problems, and child-reported behavioral problems, plus noncognitive scales for health, school environment, and home environment.
In one sense, the Krapohl study is another example of the dominance of IQ. Unlike the studies reported in the table above, the authors did conduct an analysis that entered all nine of the domains in a single model. IQ alone explained 34 percent of the phenotypic variance while the other eight combined explained 28 percent.31
What were the independent roles of the other eight? The published tables do not answer that question directly. Twenty-eight percent divided by eight indicates that most of the other roles must have been trivially small. Collateral evidence makes it clear that the measure of self-efficacy was substantially more important than any of the others, followed by the composite measure of personality.[32] As an estimate, IQ had an independent role of about three and a half times that of self-efficacy and more than six times the independent role of personality, the school environment, or parent-reported behavior problems.33 IQ was much more important than self-efficacy in explaining phenotypic variance, several times more important than personality, the school environment, or parent-reported behavior problems. The independent roles of the home environment, well-being, health, and child-reported behavior problems were virtually zero.
The second study was conducted by a team of psychologists at the University of Texas using a sample of 811 school-age twins from the Texas Twin Project. First author was Elliot Tucker-Drob. The analysis combined several measures of cognitive ability, the Big Five personality factors, and seven character traits that have featured prominently in recent years: grit, intellectual curiosity, intellectual self-concept, mastery orientation, educational value, intelligence mindset, and test motivation.
The seven character traits all loaded on a general character factor that seems to be closely related to intellectual curiosity. The correlation of the general character factor to a combined measure of knowledge and academic achievement (Ach) was +.47. With regard to the Big Five personality factors, openness was as strongly correlated with Ach (r = +.48) but, surprisingly to me, conscientiousness was not (r = +.16).34 The general character factor contributed a lot to the measure of Ach, but not as much as the measure of fluid intelligence. The same was true of openness.35
Which Comes First? Test Scores or SES?
How does the combined influence of heritable ability and personality traits stack up against the influence of the socioeconomic circumstances into which a child is born?
Before turning to the results, I need to point out that any measure of parental SES is not only a measure of the child’s environment; it is also partially a measure of the parents’ talents (or lack of them) that produced the SES, which in turn are heritable—in the case of IQ and personality, substantially heritable. It is amazing how seldom the authors of technical articles that purport to assess the relative roles of IQ and parental SES in producing adult outcomes mention this large and inescapable confound. Having mentioned it here, I ignore it in the discussion of the following studies, because none of them provides a way to estimate the degree of confounding between IQ and parental SES. I return to the issue with some studies that do provide such estimates in chapter 12.
Performance in College
In the United States, the poster child for the indictment of tests is the relationship between parental SES and performance on college admissions tests such as the SAT: The higher the parental education and income, the higher the scores of the children. How much of this relationship is causal? How much of it is a reflection of the uncomfortable possibility that smart parents attain high SES and also produce smart children?
The exhaustive analysis of this question, presented alongside a comprehensive review of prior studies, was published in 2009 in the Psychological Bulletin by a team of psychologists at the University of Minnesota (first author was Paul Sackett). The authors presented results for a meta-analysis of College Board data, a meta-analysis of other studies using a composite measure of parental SES, and a reanalysis of major longitudinal datasets. A table summarizing the results is given in the note.[36] Boiling it down:
After controlling for the admissions test score, the correlation of parental SES and college grades dropped from +.22 to –.01 in the SAT meta-analysis, from .09 to .00 in the meta-analysis of studies with composite SES measures, and from a mean of .06 to .01 among the longitudinal studies.
After controlling for the measure of SES, the correlation between admission test score and grades was reduced only fractionally: from +.53 to +.50 in the SAT meta-analysis, from +.37 to +.36 in the meta-analysis of studies with composite SES measures, and from a mean of +.313 to +.308 among the longitudinal studies.
For practical purposes, parental SES explained nothing about the student’s college grades after adjusting for test scores.
Educational Attainment, Income, and Occupation
Quantitative explorations of the comparative roles of IQ and parental SES in producing economic success began with a pair of books in the 1970s for which sociologist Christopher Jencks was the principal investigator and first author: Inequality, published in 1972, and Who Gets Ahead, published in 1979.37 The results of subsequent studies have not been consistent in their estimates of the magnitudes of the independent effects of IQ and parental SES, but they have all found such effects.
The following table shows six databases with good pedigrees. Three of them—the Aberdeen Birth Cohort, the National Child Development Study, and Project Talent—were part of the preceding table on IQ and personality factors. The other three are a sample drawn from the Scottish Mental Survey of 1932 (first author was Ian Deary) who were followed up at ages 38–41, and the 1979 and 1997 cohorts of the Longitudinal Survey of Youth project mentioned in the introduction to Part III.38 They show results over a broad period, from the 1970s up to the present. The members of the earliest cohort reached age 40 in 1972; the oldest members of the most recent cohort are just turning 40 in 2020.
COMPARATIVE ROLES OF IQ AND CHILDHOOD SES IN ADULT OUTCOMES
Scottish Mental Survey of 1932
Birth year: 1932
Age at follow-up: 38–41
Educational attainment
Childhood IQ: +0.20
Childhood SES: +0.33
Adult social class
Childhood IQ: +0.53
Childhood SES: +0.40
Aberdeen Birth Cohort of 1936
Birth year: 1936
Age at follow-up: 64
Adult reading test
Childhood IQ: +0.65
Childhood SES: +0.10
Adult SES
Childhood IQ: +0.52
Childhood SES: +0.34
Nat’l Child Development Study (UK)
Birth year: 1958
Age at follow-up: 50
Educational attainment
Childhood IQ: +0.40
Childhood SES: +0.30
Occupational prestige
Childhood IQ: +0.28
Childhood SES: +0.16
Project Talent (USA)
Birth year: 1933–36
Age at follow-up: 28
Educational attainment
Childhood IQ: +0.41
Childhood SES: +0.25
Adult income
Childhood IQ: +0.08
Childhood SES: +0.09
Occupational prestige
Childhood IQ: +0.39
Childhood SES: +0.18
NLSY, 1979 Cohort (USA)
Birth year: 1957–64
Age at follow-up: 42–49
Educational attainment
Childhood IQ: +0.52
Childhood SES: +0.16
Adult income
Childhood IQ: +0.30
Childhood SES: +0.08
NLSY, 1997 Cohort (USA)
Birth year: 1980–84
Age at follow-up: 31–35
Educational attainment
Childhood IQ: +0.46
Childhood SES: +0.22
Adult income
Childhood IQ: +0.21
Childhood SES: +0.13
Sources: Deary, Taylor, Hart et al. (2005): Fig. 3; Staff, Hogan, and Whalley (2017): Table 3; Cheng and Furnham (2012): Fig. 2; Damian, Su, Shanahan et al. (2014): Tables 3, 5, and 6; and author’s analysis, NLSY79 and NLSY97.[39]
All of the metrics represent some form of standardized regression coefficient, but once again there are too many differences across the studies to compare them. Rather, focus on the relative sizes of the coefficients for childhood IQ and childhood SES. The generalizations to be drawn are that all but two of the coefficients for childhood IQ are larger than the coefficients for childhood SES and that in all but one case IQ has a greater effect on educational attainment than it does on adult income or occupation.[40]
Not much should be made of specific magnitudes with a sample of six, but, to get a sense of the table, the median ratio of the effects of childhood IQ to those of childhood SES was 1.75. And of course, the putative effects of childhood SES were inescapably partly genetic, as I noted earlier.
Recapitulation
Cognitive ability and personality strengths, both substantially heritable, are important to achieving success in life—educationally, in earnings, and in professional achievement. To that extent, class structure is, in the words of Proposition #9, importantly based on differences in abilities that have a substantial genetic component. The relationship between these abilities and success does not amount to genetic determinism for individuals. It is, however, strong enough to decisively shape the social structure of modern Western societies. Of the many abilities involved, the general mental factor g is usually dominant. Proposition #9 is something that we haven’t had to argue about for decades, despite all the furor that has accompanied the idea that success is affected by genes. It’s demonstrably true.
But it is understandable why that truth still raises hackles for so many people. It smacks of self-satisfaction with the way things are and indifference toward those who were unlucky in the genetic lottery. It is also understandable that scholars and policymakers alike have put a great deal of effort into attempts to use education and other measures to help the unlucky. How well have they succeeded? What are the constraints on helping? Those extremely sensitive issues are up next.
13
Constraints and Potentials
Proposition #10: Outside interventions are inherently constrained in the effects they can have on personality, abilities, and social behavior.
Outside interventions can change people profoundly. An inspiring teacher can fire the imagination of a bored adolescent and change the trajectory of that student’s life. Friends who intervene in an alcoholic’s life can change its trajectory. It happens all the time. But Proposition #10 is not about individual successes. Rather, Proposition #10 is about whether people can be changed by design in large numbers. That’s a more ambitious objective, and it is made more ambitious yet by the material you have read in the preceding three chapters. Those findings lend themselves to another syllogism:
1. If the shared environment explains little of the variance in cognitive repertoires, and
2. If the only environmental factors that can be affected by outside interventions are part of the shared environment,
3. Then outside interventions are inherently constrained in the effects they can have on cognitive repertoires.
In other words, it is not within our power to do much to change personalities or abilities or social behaviors by design on a large scale. This chapter explores some ways in which the conclusion of the syllogism might be challenged:
“The first premise is wrong for some important outcomes.”
“The first premise is wrong for the early stages of life.”
“The first premise is wrong when it comes to changing self-concept.”
“The second premise is wrong because some aspects of the nonshared environment can be affected by outside interventions.”
“But you’re ignoring epigenetics!”
“The First Premise Is Wrong for Some Important Outcomes”
In the Polderman meta-analysis discussed in chapter 11, there were exceptions to the generalization that the role of the shared environment is trivially small. The shared environment explained 36 percent of the variance for basic personal interactions, 24 percent for problems related to upbringing, 26 percent for disorders due to multiple drug use, and 25 percent for educational attainment. If an outside intervention could remediate a high proportion of the shared environmental influences that cause problems in these areas, they would be meaningful accomplishments.
The Comparatively Minor Role of Outside Interventions
The question is whether that’s a reasonable expectation. The things that fall into the shared and nonshared environment include all the effects of the family, school, and neighborhood exerted during a young person’s waking hours. The usual kinds of outside intervention—counseling, tutoring, mentoring, after-school activities, job training—amount to a few hours per week. Even if the quality of those interventions is excellent (a difficult thing to achieve in itself), we’re talking about a tiny fraction of a youth’s waking hours. A teenager with a drug counselor might really like and listen to his counselor. But he walks out of his meeting with the counselor into the same shared home, school, and neighborhood environments that explain 26 percent of the variance in drug use. If the shared environment explained 90 percent of the variance, then maybe a few hours a week of an outside intervention might make some measurable difference. If the shared environment explains just 26 percent of the variance, the outside intervention has to be big—boarding school, for example, moving the family out of the neighborhood, or adoption into a new family. Ordinary levels of outside intervention are too small relative to the competing influences.
A similar logic applies to outside interventions that attempt to modify the behavior of parents. Consider the outcome with the largest role for the shared environment (36 percent), basic personal interactions. That number says that parents can legitimately think they’ve had a major role in raising sociable children. But if a 10-year-old is exhibiting problems with basic personal interactions, a program that tries to change the parents’ parenting practices has arrived far too late. Even if the program is successful, the child is not going to turn on a dime just because his parents are behaving somewhat more positively and effectually than they did for the preceding 10 years. If outcomes that are primarily influenced by parenting are the problem, then solutions will have to begin as close to birth as possible.
I will get to the empirical track record of outside interventions later. Now, I’m drawing your attention to underlying realities that are too seldom made explicit. Outside interventions of normal magnitude and intensity make sense for an extremely limited set of problems that are analogous to health problems that can be cured by antibiotics or surgery.
Problems that fit those criteria are rare, but they exist. Here’s an example: School systems in large urban areas are notorious for tolerating chaotic classrooms in a handful of schools in the most impoverished part of town. There’s no excuse for it. Children who are eager to learn are pre
vented from doing so, with lifelong consequences, and yet an outside intervention can completely cure that problem in a day: Install strict rules of in-class conduct, and promptly and without exception eject disruptive students from the classroom. Teachers will be able to teach and the remaining students will be able to learn.
The difficulty, of course, is what to do with the students who have been ejected. Solving their problems is a matter of changing their personalities, abilities, or social behavior, or a combination of all three, and that’s what the first premise of the syllogism says we don’t know how to do.
The underlying point of my example is why the solution works: We know how to help people who already want to do something and are artificially prevented from doing it. My solution doesn’t have to change the students who are already trying to learn. It just needs to provide them with an environment in which they are enabled to do what they already want to do. We are not constrained from helping with outside interventions. We’re just constrained in whom we can help with what kinds of problems.
If instead we try to change people who aren’t ready and able to change given the opportunity, we’re back to a situation in which powerful competing forces, acting through both genes and the nonshared environment, overwhelm the magnitude of the intervention that seeks to produce change. Expecting to see a major impact from outside interventions is usually unrealistic.
Human Diversity Page 29