Some anon sent me this paper, asked if there was a rebuttal somewhere. It’s a well cited economics paper, 1248 citations on Google Scholar. I wasn’t familiar with it, but reading it over, I see some things comment on.
- Bowles, S., & Gintis, H. (2002). The inheritance of inequality. Journal of economic Perspectives, 16(3), 3-30.
How level is the intergenerational playing field? What are the causal mechanisms that underlie the intergenerational transmission of economic status? Are these mechanisms amenable to public policies in a way that would make the attainment of economic success more fair? These are the questions we will try to answer.
Their abstract is lacking but their introduction summarizes what they want to do:
People differ markedly in their views concerning the appropriate role of government in reducing economic inequality. Self-interest and differences in values explain part of the conflict over redistribution. But by far the most important fault line is that people hold different beliefs about why the rich are rich and the poor are poor. Survey data show that people—rich and poor alike—who think that “getting ahead and succeeding in life” depends on “hard work” or “willingness to take risks” tend to oppose redistributive programs. Conversely, those who think that the key to success is “money inherited from family,” “parents and the family environment,” “connections and knowing the right people,” or being white, support redistribution (Fong 2001, Bowles, Fong and Gintis 2002a). Handing down success strikes many people as unfair even if the stakes are small, while differences in achieved success may be unobjectionable even with high stakes, as long as the playing field is considered level. How level is the intergenerational playing field? What are the causal mechanisms that underlie the intergenerational transmission of economic status? Are these mechanisms amenable to public policies in a way that would make the attainment of economic success more fair? These are the questions we will try to answer.
So right of the bat, it should be noted that the authors don’t fall obviously into the trap of confusing a high random measurement error for low intergenerational relationship:
No one doubts that the children of well-off parents generally receive more and better schooling and benefit from material, cultural, and genetic inheritances. But until recently, the consensus among economists has been that in the United States, success is largely won or lost in every generation. Early research on the statistical relationship between parents’ and their children’s economic status after becoming adults, starting with Blau and Duncan (1967), found only a weak connection and thus seemed to confirm that the United States was indeed the “land of opportunity.” For example, the simple correlations between parents’ and sons’ income or earnings (or their logarithms) in the United States reported by Becker and Tomes (1986) averaged 0.15, leading the authors to conclude: “Aside from families victimized by discrimination…[a]lmost all earnings advantages and disadvantages of ancestors are wiped out in three generations.” Becker (1988) expressed a widely held consensus when, in his presidential address to the American Economics Association, he concluded: “[L]ow earnings as well as high earnings are not strongly transmitted from fathers to sons.” (p. 10)
But more recent research shows that the estimates of high levels of intergenerational mobility were artifacts of two types of measurement error: mistakes in reporting income, particularly when individuals were asked to recall the income of their parents, and transitory components in current income uncorrelated with underlying permanent income (Bowles 1972, Bowles and Nelson 1974, Atkinson, Maynard and Trinder 1983, Solon 1992, Zimmerman 1992, Björklund, Jäntti, and Solon forthcoming). The high noise-to-signal-ratio in both generations’ incomes depressed the intergenerational correlation. When corrected, the intergenerational correlations for economic status appear to be substantial, many of them three times the average of the U. S. studies surveyed by Becker and Tomes (1986).
Case in point:
Estimates of the intergenerational income elasticity are presented in Solon (this issue) as well as Mulligan (1997) and Solon (2000). The mean estimates reported in Mulligan are: for consumption 0.68; for wealth 0.50; for income 0.43; for earnings (or wages) 0.34; and for years of schooling 0.29. Evidence concerning trends in the degree of income persistence across generations is mixed. Most studies indicate that persistence rises with age, is greater for sons than daughters, and is greater when multiple years of income or earnings are averaged. The importance of averaging multiple years to capture permanent aspects of economic status is dramatized in a recent study by Mazumder (forthcoming). He used a rich U.S. Social Security Administration data set to estimate an intergenerational income elasticity of 0.27 averaging son’s earnings over three years and father earnings averaged over two years. But the estimate increases to 0.47 when six years of the fathers earnings are averaged, and to 0.65 when fifteen years are averaged.
This point is also well made in this 2019 Finnish twin study which also included a meta-analysis of prior studies:
Using twenty years of earnings data on Finnish twins, we find that about 40% of the variance of women’s and little more than half of men’s lifetime labour earnings are linked to genetic factors. The contribution of the shared environment is negligible. We show that the result is robust to using alternative definitions of earnings, to adjusting for the role of education, and to measurement errors in the measure of genetic relatedness.
I tweeted the findings, but let’s repost here for ease of reference:
Compare the length of measurement vs. heritability. We know that measurement error is a big problem in behavioral genetics studies (and crime studies), who routinely underestimate A and C components and overestimate E (‘everything else’). Anyway, so back to the paper:
Though the underlying intergenerational correlation of incomes in the data set Hertz used is a modest 0.42, the differences in the likely life trajectories of the children of the poor and the rich are substantial. The “twin peaks” represent those stuck in poverty and affluence (though we do not expect the term “affluence trap” to catch on). A son born to the top decile has a 22.9 percent chance of attaining the top decile (point D) and a 40.7 percent chance of attaining the top quintile. A indicates that the son of the poorest decile has a 1.3 percent chance of attaining the top decile, and a 3.7 percent chance of attaining the top quintile. C indicates that children of the poorest decile have a 31.2 percent chance of occupying the lowest decile, and a 50.7 percent chance of occupying the lowest quintile, while B shows that the probability that a child of the richest decile ends up in the poorest decile is 2.4 percent, and a 6.8 percent chance of occupying the lowest quintile. Hertz’ transmission matrix and other studies (Corak and Heisz 1999, Cooper, Durlauf and Johnson 1994, Hertz 2001) suggest that distinct transmission mechanisms may be at work at various points of the income distribution. For example wealth bequests may play a major role at the top of the income distribution, while at the bottom vulnerability to violence or other adverse health episodes may be more important. Mobility patterns by race also differ dramatically (Hertz 2002). Downward mobility from the top quartile to the bottom quartile is nearly five times as great for blacks as for whites. Thus successful blacks do not transmit to their children whatever it is that accounts for their success as effectively as do successful whites. Correspondingly, blacks born to the bottom quartile attain the top quartile at one half the rate of whites.
These findings have replicated well. For instance, this Swedish study using register data show this two-channels situation for wealth (“At the extreme top (top 0.1%) income transmission is remarkable with an intergenerational elasticity of approximately 0.9.”!).
What they don’t note here is that this differential regression towards the mean is a prediction of the hereditarian model of black-white gap. This finding goes all the way back to the 1970s. Jensen writes in the 1973 book Educability and Group Differences:
The data shown in Scarr-Salapatek’s Table 3 (p. 1288), how ever, make this interpretation highly questionable. These data allow comparison of the mean scores on the combined aptitude tests for Negro children whose parents’ level of education and income are both above the median (of the Negro and white samples combined) with the mean scores of white children whose parents’ education and income are both below the common median. The lower-status white children still score higher than the upper-status Negro children on both the verbal and the non-verbal tests. Although non-verbal tests are generally considered to be less culture-biased than verbal tests, it is the non-verbal tests which in fact show the greater discrepancy in this comparison, with the lower-status whites scoring higher than the upper-status Negroes. But in this comparison it is the upper-status Negro group that has the higher heritability (i.e., greater genetic variance) on both the verbal and non-verbal tests. Thus, the lower heritability which Scarr-Salapatek hypothesizes as being consistent with Negroes’ generally poorer performance because of environmental deprivation applies in this particular comparison to the lower-status white group. Yet the lower-status white group out-performs the upper- status Negro group, which has the highest heritability of any of the subgroups in this study (see Table 9, p. 1292).
This finding is more difficult to reconcile with a strictly environmental explanation of the mean racial difference in test scores than with a genetic interpretation which invokes the well established phenomenon of regression toward the population mean. In another article Scarr-Salapatek (1971b) clearly explicated this relevant genetic prediction, as follows:
Regression effects can be predicted to differ for blacks and whites if the two races indeed have genetically different population means. If the population mean for blacks is 15 IQ points lower than that of whites, then the offspring of high-IQ black parents should show greater regression (toward a lower population mean) than the off spring of whites of equally high IQ. Similarly, the offspring of low-IQ black parents should show less regression than those of white parents of equally low IQ. (Scarr-Salapatek, 1971b, p. 1226)
In other words, on the average, an offspring genetically is closer to its population mean than are its parents, and by a fairly precise amount. Accordingly, it would be predicted that upper-status Negro children should, on the average, regress downward toward the Negro population mean IQ of about 85, while lower-status white children would regress upward toward the white population mean of about 100. In the downward and upward regression, the two groups’ means could cross each other, the lower-status whites thereby being slightly above the upper-status Negroes. Scarr-Salapatek’s data (Table 3) are quite consistent with this prediction. Her finding is not a fluke; the same phenomenon has been found in other large-scale studies (see Chapter 4, pp. 117-19).
Chapter 4 has:
But if we match a number of Negro and white children for IQ 7 and then look at the IQs of their full siblings with whom they were reared, we find something quite different: the Negro siblings average some 7 to 10 points lower than the white siblings. Also, the higher we go on the IQ scale for selecting the Negro and white children to be matched, the greater is the absolute amount of regression shown by the IQs of the siblings .8 For example, if we match Negro and white children with IQs of 120, the Negro siblings will average close to 100 , the white siblings close to 110 . The siblings of both groups have regressed approximately halfway to their respective population means and not to the mean of the combined populations. The same thing is found, of course, if we match children from the lower end of the IQ scale. Negro and white children matched for, say, IQ 70 will have siblings whose average IQs are about 78 for the Negroes and 85 for the whites. In each case the amount of regression is consistent with the genetic prediction. The regression line, we find, shows no significant departure from linearity throughout the range from IQ 50 to 150. This very regular phenomenon seems difficult to reconcile with any strictly environmental theory of the causation of individual differences in IQ that has yet been proposed. If Negro and white children are matched for IQs of, say, 120, it must be presumed that both sets of children had environments that were good enough to stimulate or permit IQs this high to develop. Since there is no reason to believe that the environments of these children’s siblings differ on the average markedly from their own, why should one group of siblings come out much lower in IQ than the other? Genetically identical twins who have been reared from infancy in different families do not differ in IQ by nearly so much as siblings reared together in the same family. It can be claimed that though the white and Negro children are matched for IQ 120, they actually have different environments, with the Negro child, on the average, having the less intellectually stimulating environment. Therefore, it could be argued he actually has a higher genetic potential for intelligence than the environmentally more favored white child with the same IQ. But if this were the case, why should not the Negro child’s siblings also have somewhat superior genetic potential? They have the same parents, and their degree of genetic resemblance, indicated by the theoretical genetic correlation among siblings, is presumably the same for Negroes and whites .9
Similar regression would be expected between parents and children but there are no adequate cross-racial studies of this for IQ. A rigorous study would require that the Negro and white parents be matched not only for education, occupational status, and income, but also for IQ. A genetic hypothesis would predict rather precisely the amount that the offspring of Negro and white parents matched for these variables would differ in IQ. The only existing evidence relevant to this hypothesis is the finding, in a number of studies which attempted to match Negroes and whites for socioeconomic status, that the upper-status Negro children average 2 to 4 IQ points below the /ow-status white children (Shuey, 1966, p. 520; Scarr-Salapatek, 1971a; Wilson, 1967), even though it is most likely that the upper-status Negro parents were of higher IQ than the low-status white parents. The regression-to-the-mean phenomenon could account for the cross over of the average IQs of the children from the two racial groups.
We replicated this pattern in our 2019 paper Filling in the Gaps: The Association between Intelligence and Both Color and Parent-Reported Ancestry in the National Longitudinal Survey of Youth 1997 section 4.5.
Authors discuss intelligence and earnings (or income):
We have located 65 estimates of the normalized regression coefficient of a test score in an earnings equation in 24 different studies of U.S. data over a period of three decades. Our meta-analysis of these studies is presented in Bowles, Gintis, and Osborne (2002a). The mean of these estimates is 0.15, indicating that a standard deviation change in the cognitive score, holding constant the remaining variables (including schooling), changes the natural logarithm of earnings by about one-seventh of a standard deviation. By contrast, the mean value of the normalized regression coefficient of years of schooling in the same equation predicting the natural log of earnings in these studies is 0.22, suggesting a somewhat larger independent effect of schooling. We checked to see if these results were dependent on the weight of overrepresented authors, the type of cognitive test used, at what age the test was taken and other differences among the studies and found no significant effects. An estimate of the causal impact of childhood IQ on years of schooling (also normalized) is 0.53 (Winship and Korenman 1999). A rough estimate of the direct and indirect effect of IQ on earnings, call it b, is then b = 0.15 + (0.53)(0.22) = 0.266.
Here they have apparently forgotten their own lesson about measurement error. In fact, as we know, when we do large scale studies with more decent measurement, this correlation is quite a bit larger than .27. NLSY79 shows a correlation of .37 in the overall sample, and by sex, 0.48 and 0.30 for men and women, respectively.
The authors continue to estimation of heritability of IQ (actually, intelligence):
For this we need some genetics (the details are in the Appendix and in Bowles and Gintis (2001)), and a few terms—phenotype, genotype, heritability and the genetic correlation—unfamiliar to many economists. A person’s IQ—meaning, a test score—is a phenotypic trait, while the genes influencing IQ are the person’s genotypic IQ. Heritability is the relationship between the two. Suppose that, for a given environment, a standard deviation difference in genotype is associated with a fraction h of a standard deviation difference in IQ. Then h2 is the heritability of IQ. Estimates of h2 are based on the degree of similarity of IQ among twins, siblings, cousins and others with differing degrees of genetic relatedness. The value cannot be higher than 1, and most recent estimates are substantially lower, possibly more like a half or less (Devlin, Daniels and Roeder 1997, Feldman, Otto and Christiansen 2000, Plomin 1999). The genetic correlation is the degree of statistical association between genotypes of parents and children, which is 0.5 if the parents’ genotypes are uncorrelated (random mating). But couples tend to be more similar in IQ than would occur by random mate choice (assortative mating) and this similarity is associated with an unknown correlation m of their genotypes. The effect is to raise the genetic correlation of parent and offspring to (1 + m)/2.
Their mistake here is to confuse childhood heritability with adult heritability. This point hadn’t at the time been hammered home sufficiently, and their cited study here Devlin et al, is a meta-analyzed that disregarded the age interaction (Wilson effect, as we call it now), and produced a spuriously low value of 48% and narrow/additive h² of 34%. A more sensible broad value would be about 80% for adults, and perhaps 50% additive. The best study on this topic is really this extended family study from 2012, which found: additive genetic factors (44%), genetic dominance (27%), phenotypic assortment (11%) and non-shared environmental factors (18%). I am assuming their phenotypic assortment is a genetic effect, so it goes under genetics too (assortative mating increases genetic variance), giving 44+27=71% without it, and 82% with it. This was not based on a latent variable approach, so values need to be moved upwards a bit for random measurement error. Another family study from 2008 estimated additive h² of 67%, which was based on a latent variable, so no effect of random measurement error (but construct error is present because they only used Raven’s).
So their central calculation is using quite the low ball numbers:
Using the values estimated above, we see that the contribution of genetic inheritance of IQ to the intergenerational transmission of income is (h2(1+m)/2)(0.266)2 = .035(1 + m)h2. If the heritability of IQ were 0.5 and the degree of assortation, m, were 0.2 (both reasonable, if only ball park estimates) and the genetic inheritance of IQ were the only mechanism accounting for intergenerational income transmission, then the intergenerational correlation would be 0.01, or roughly two percent the observed intergenerational correlation. Note the conclusion that the contribution of genetic inheritance of IQ is negligible is not the result of any assumptions concerning assortative mating or the heritability of IQ: the IQ genotype of parents could be perfectly correlated and the heritability of IQ 100 per cent without appreciably changing the qualitative conclusions. The estimate results from the fact that IQ is just not an important enough determinant of economic success.
I leave it to the reader to play with the numbers by plugging in some other values, but obviously, the effect will be higher than what they found since all their estimates are too low.
Authors move on to making the usual deductivist fallacy regarding changeability/modifiability and heritability:
But two words of caution are in order. First, as we will demonstrate, our estimates are quite sensitive to variations in unobserved parameters. Second, it is sometimes mistakenly supposed that if the heritability of a trait is substantial, then the trait cannot be affected much by changing the environment. The fallacy of this view is dramatized by the case of stature. The heritability of height estimated from U.S. twin samples is substantial (about 0.90, Plomin et al., 2000). Moreover there are significant height differences among the peoples of the world: Dinka men in the Sudan average 5 feet and 11 inches—a bit taller than Norwegian and U.S. military servicemen and a whopping 8 inches taller than the Hadza hunter-gatherers in Southern Africa (Floud, Wachter and Gregory 1990). But the fact that Norwegian recruits in 1761 were shorter than today’s Hadza shows that even quite heritable traits are sensitive to environments. What can be concluded from a finding that a small fraction of the variance of a trait is due to environmental variance is that policies to alter the trait through changed environments will require non-standard environments that differ from the environmental variance on which the estimates are based.
(Read Sesardić 2005 on this stuff.)
In this case, they mix up between group variation too, which obviously will include some actual genetic differences, and temporal differences, which include much less genetic difference, though still some. Taller men have more children, and this is more so than shorter women have more, so we have been selecting for height for a long time, and I am not sure which fraction of the secular increase in height (since 1761!!) this explains, but maybe 30%?
Overall, this is a good paper, interesting and important question, the right kind of methods (mediation analysis combined with heritability), just ultimately wrongish conclusions due to their failure to take into account the age interaction for heritability and the considerable measurement error problems. This is definitely a study that should be redone using newer values by an economics-intelligence-behavioral genetics combined team.