Sunday, September 20, 2015

Children’s personal characteristics are the product of three sources: shared environment, non-shared environment, and parents’ genes.

Charles Murray has an excellent review of Robert Putnam's new book Our Kids: The American Dream in Crisis in the article The Trouble With Kids Today by Charles Murray. You have to be careful with Putnam because while his writing is always interesting, his actual analysis is often flawed (see his earlier book, Bowling Alone).

Just the other evening I came across this quote from Julius Caesar in his work, Gallic Wars.
In bello parvis momentis magni casus intercedunt.

In war events of great importance are the result of trivial causes.

Julius Caesar, Bellum Gallicum, I, 21.
I was struck by both the profundity of the truth as well as the remarkable circumstance of the age of that observation.

He captures a conundrum with which we are still wrestling - that all the odds can be stacked in your favor and yet still the roll of the die is against you and conversely you can be completely neglectful and yet circumstance ends up bestowing some benefit that was completely unearned. Yes, chance does favor the prepared mind. But not always.

You see it in correlations all the time. High IQ, good education, strong work ethic, good health and established behaviors are all supposed to predict good life outcomes and are indeed highly correlated with good life outcomes. A person with such attributes will reliably live longer, earn more, have more stable marriages, etc. ON AVERAGE. Yes, such factors are reliable predictive variables. Yet, some, among all those individuals making up the average, will fall far short despite having all the advantageous attributes. It is very difficult to determine, in advance, who will succeed and who will fail, even though on average they will all do much better than average.

For all the Marxists, critical theorists, etc. out there who want to treat human systems as scientifically manageable, this a gross shortfall. They want to pursue policy A, B, C with the expectation that it will reliably yield desirable outcomes X, Y, Z. The tragedy, for such critical theorists, is that ABC do indeed yield some percentage of XYZ but not reliably and only some percentage. Human systems are not like factory systems.

Which is where Caesar's comment comes in. You can strategically do all the preparation that is wise to do and be superbly ready for just about any eventuation. But when you switch from the strategic and prospective to the tactical and real time, trivial (and often unanticipated and unanticipatable) events do drive outcomes of great importance.

Murray describes this interplay very well.
That said, I must record my own judgment that everything Putnam recommends could be implemented full-bore—far beyond any reasonable hope—and little, alas, would change in the long term. The opportunity gap is driven by larger forces, which his policy prescriptions cannot do much about. Three reasons stand out.

First, the standard interventions are aiming at a relatively unimportant target. Children’s personal characteristics are the product of three sources: shared environment, non-shared environment, and parents’ genes. Government programs can affect only one of those three—shared environment—which, for the most important outcomes, usually has the least effect of the three.

You may not be familiar with the terms “shared” and “non-shared” environment. The shared environment includes such things as a family’s income and social status, quality of the schools, and parenting practices. The non-shared environment is the sum of random differences such as events in the womb that affect one sibling differently from another, an injury or illness after birth that affects one sibling and not the other, and peer groups that siblings don’t share. Some unknown but probably large proportion of the non-shared environment is simply statistical noise.

Aren’t the components of the shared environment the important causes of how well children do in life, as Putnam himself is convinced? For some immediate outcomes, yes; for ultimate outcomes, no. Consider the results of a comprehensive meta-analysis of more than 2,000 twin studies published in Nature Genetics in May of this year. The shared environment played a large role in the religiosity of children (explaining 44 percent and 35 percent of the variance in the two estimates presented by the study), and a substantial role in explaining problems in parent-child relationships (33 percent for both estimates).

But when it comes to the outcomes that Putnam associates with the opportunity gap, the contribution of the shared environment is modest. For “higher-level cognitive functions” (IQ), the estimates of the role of the shared environment were just 24 percent and 17 percent of the variance. For educational attainment: 27 percent and 13 percent. For conduct disorders (antisocial and aggressive behavior): 18 percent and 15 percent.

That’s not the whole story. Genes and environment interact, among other things. But my point is simple and survives the complications: the roster of standard interventions to reduce the opportunity gap is almost entirely focused on factors that have modest causal roles. Furthermore, a program lasting at most a few hours a day can influence only a small proportion of that modest causal role. The evaluation literature for interventions necessarily yields meager long-term impact even for the best-executed program because the potential effect to begin with is so small. If policy scholars are serious about having a major impact on the shared environment, they should be advocating adoption at birth and high-quality orphanages. They don’t.

Second, the opportunity gap exists alongside a substantial ability gap. Most of the graphs in “Our Kids” show the results for parents with at least a college degree versus those for parents with no more than 12 years of school and a high-school diploma. What are the IQs of those two groups? In the 1979 cohort of the National Longitudinal Study of Youth (NLSY), replicating Putnam’s assignment rules, the mean IQ of the college group was 23 points higher than that of the high school group.

In case you’re wondering, that’s not a function of race. Among non-Latino whites, the difference was 22 points. In statistical terms, those are differences of about 1.5 standard deviations. For the population as a whole, the average person in the high-school group was at the twenty-ninth IQ percentile while the average person in the college group was at the eighty-fourth percentile. Since children’s IQ is correlated with parental IQ, it is not surprising to learn that the means of the children of the high school and college groups are also separated—by about 19 points in the same NLSY cohort. Recall the modest role of the shared environment in producing that difference.

Again, my underlying point is simple. IQ has a substantial direct correlation with measures of success in life, and it is also correlated with a variety of other characteristics that promote success—perseverance, deferred gratification, good parenting, and the aspects of personality that are variously called “emotional intelligence” or “grit.” The correlations are not large, but many modest individual correlations produce large differences in life outcomes when the means of two groups are separated by as large a gap as separates both parents and children of America’s working and upper-middle classes.

Third, the gap in human capital in working-class and upper-middle-class communities has been widening over time. In 1960, just 8 percent of adults had college degrees, and many of those had pedestrian academic ability—going to college then was largely determined by socioeconomic status. In that America, an extremely large proportion of the smartest people in the country had no more than a high-school education. Data on the IQ of high school and college graduates prior to mid-century indicate that the gap between Putnam’s two groups as of 1960 was on the order of 14 points, not 23.

Since then, the sorting process has gotten much more efficient. Few high-school graduates with IQs well over 100 don’t get at least some post high-school education. It has long been recognized that the functioning of black communities took a big hit when the civil rights revolution enabled many of the most successful blacks to move out. The same thing has been happening to the country as a whole. White working-class communities have also seen an outmigration of the most able; that outmigration is continuing, and it is entrenching many of the problems in working-class communities that Putnam laments.

It’s not just that the IQ gap in working-class and upper-middle-class communities has gotten wider. The life penalties associated with low IQ have risen since 1960. If you focus on the economic changes since 1960, those with low IQ have faced a labor market in which the market value of a strong back has dropped while the value of brains has soared.

If you focus on the reforms and social programs of the 1960s, the reductions in immediate penalties for destructive behavior (e.g., doing drugs, dropping out of school, grabbing purses, having a baby without marriage) had the most effect on people who were impulsive, attracted to immediate gratification, and unable to foresee long-term consequences—qualities associated with low IQ. The effects of such changes in incentives among the smart were much smaller.

My takeaway from all this was expressed in the closing chapter of my own work on Putnam’s topic, “Coming Apart” (2012). Very briefly, I don’t think America’s civic culture will be revitalized by the kinds of programs that “Our Kids” advocates. If it is to happen, it must be through a cultural Great Awakening that leads the elites to reengage in America’s traditional civic culture; one that reverses what Robert Reich memorably labeled “the secession of the successful.” Being willing to pay higher taxes to finance more social programs is not what I have in mind.

No comments:

Post a Comment