American Economic Association

The Inheritance of Inequality Author(s): Samuel Bowles and Herbert Gintis Reviewed work(s): Source: The Journal of Economic Perspectives, Vol. 16, No. 3 (Summer, 2002), pp. 3-30 Published by: American Economic Association Stable URL: http://www.jstor.org/stable/3216947 . Accessed: 02/02/2012 16:37
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

American Economic Association is collaborating with JSTOR to digitize, preserve and extend access to The Journal of Economic Perspectives.

http://www.jstor.org

Journal

of Economic Perspectives?Volume

16, Number 3?Summer

2002?Pages

3-30

The

Inheritance

of

Inequality

Samuel

Bowles

and

Herbert

Gintis

People

the appropriate role of markedly in their views concerning in reducing economic Self-interest and differences government inequality. in values explain part of the conflict over redistribution. But by far the most

differ

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 to take risks" tend to oppose redistributive "willingness programs. Conversely, those who think that the key to success is "money inherited from family," "parents and the "connections and knowing the right people" or being white family environment," redistribution support (Fong, 2001; Fong, Bowles and Gintis, 2002). Handing down success strikes many people as unfair even if the stakes are small, while differences in achieved playing How level is the intergenerational that underlie the intergenerational success may be unobjectionable field is considered level. even with high stakes, as long as the

playing field?1 What are the causal mechanisms of economic transmission status? Are these

1 See Bowles and Gintis (2001) for the relevant formal models and other technical aspects of this research, also available at (http://www.santafe.edu/sfi/publications/working-papers.html). Arrow, Bowles and Durlauf (1999) and Bowles, Gintis and Osborne (forthcoming) present collections of recent empirical and theoretical research. ? Samuel Bowles is Professor of Economics at the University of Siena, Siena, Italy, and Director of the Economics Program, Santa Fe Institute, Santa Fe, New Mexico. Herbert Gintis is a member of the External Faculty, Santa Fe Institute, Santa Fe, New Mexico. Both authors are Emeritus Professors of Economics, University of Massachusetts, Amherst, Massachusetts. and their websites Their e-mail addresses are ([email protected]) and ([email protected]), are (http://www-unix. oit.umass.edu/ bowles) and (http://www-

unix. oit. umass. edu/ gintis).

Journal

of Economic Perspectives

mechanisms economic better

to public policies in a way that would make the attainment of more fair? These are the questions we will try to answer. No one doubts that the children of well-off parents generally receive more and amenable success

and benefit from material, cultural and genetic inheritances. But schooling has been that in the United States, until recently, the consensus economists among success is largely won or lost in every generation. Early research on the statistical economic status after becoming between parents' and their children's relationship and adults, starting with Blau and Duncan (1967), found only a weak connection thus seemed to confirm that the United States was indeed the "land of opportuthe simple correlations between parents' and sons' income or nity." For example, in the United States reported by Becker and Tomes earnings (or their logarithms) "Aside from families vic(1986) averaged 0.15, leading the authors to conclude: timized by discrimination . . . [ajlmost all earnings advantages and disadvantages of a widely ancestors are wiped out in three generations." Becker (1988) expressed address to the American Economics held consensus when, in his presidential he concluded Association, 10): "[L]ow earnings as well as high earnings are not (p. from fathers to sons." strongly transmitted But more erational recent mobility of high levels of intergen? error: mistakes in were artifacts of two types of measurement when individuals were asked to recall the income of particularly research shows that the estimates

reporting income, their parents, and

in current income uncorrelated with transitory components Bowles and Nelson, 1974; Atkinson, income (Bowles, 1972; underlying permanent 1983; Solon, 1992, 1999; Zimmerman, 1992). The high Maynard and Trinder, in the incomes of both generations the intergenera? noise-to-signal-ratio depressed tional nomic correlation. status appear When corrected, to be substantial, times the average of the surveyed by Becker and Tomes (1986). consensus estimates of the intergenerational transmission of higher success has stimulated intergenerational many of them three the correlations for eco?

U.S. studies The economic which much similar

research. The relevant facts on empirical are now agree include the following: brothers' incomes most researchers chosen males of the same race and more similar than those of randomly

twins are much more similar the incomes of identical age differences; of well-off parents obtain than fraternal twins or non-twin brothers; the children and wealth inheritance makes an important more and higher quality schooling; of the very rich. On the basis to the wealth owned by the offspring contribution of these and other empirical transmission regularities, of economic it seems status safe to conclude that the is accounted for by a hetertrans? and cultural

the genetic of mechanisms, including of cognitive skills and noncognitive traits in demand by personality the inheritance of wealth and income-enhancing employers, group membereducation and health status enjoyed by the ships, such as race, and the superior children of higher status families. mission the transmission of economic success across generations However, of a black box. We find that the combined inheritance something remains processes

intergenerational collection ogeneous

Samuel Bowles and Herbert Gintis

and educational attainments of through superior cognitive performance operating those with well-off parents, while important, at most three-fifths of the explain transmission of economic status. Moreover, while genetic trans? intergenerational of earnings-enhancing traits appears to play a role, the genetic transmission of IQ appears to be relatively unimportant. It might be thought that the black box is an artifact of poor measurement of the intervening variables relative to the measurement of the income or earnings of mission and parents and offspring. But this does not seem to be the case. Years of schooling of school attainment, like cognitive other measures are measured performance, with relatively little error. Better measurements will of course help; but we are not likely to improve much measurement of school going on inside the on our measures The quality black box. of IQ, and recent improvements have not given us much illumination about fundamental is not that in the what's we are

problem the right variables poorly, but that we are missing some of the important measuring variables entirely. What might these be? Most economic models treat one's income as the sum of the returns to the factors individual of production trait that

strong will contribute Included are race, geographical location, height, beauty or other aspects of physical health status and personality. Thus, by contrast to the standard apappearance, we give considerable attention to income-generating characteristics that proach, are not generally considered to be factors of production. In studies of the inter? of economic transmission status, our estimates suggest that cognitive generational skills and education have been overstudied, while wealth, race and noncognitive behavioral traits have been understudied.

one brings to the market, like skills, or capital goods. But any and for which parent-offspring affects income similarity is to the intergenerational transmission of economic success.

Measuring Economic

the

Intergenerational

Transmission

of

Economic

Status

in discrete categories?by in membership for example?or income hierarchically by earnings, continuously, or wealth. The discrete approach allows a rich but difficult-to-summarize representation of the process of intergenerational of status using transition persistence 1992; this probabilities among the relevant social ranks (Erikson and Goldthorpe, status can be measured classes, ordered issue). By contrast, on the correlation these various correlations causal continuous between measures the economic based allow a simple metric of persistence, status of the two generations. Moreover, the into additive components reflecting

approaches measurement

may be decomposed mechanisms for parent-child economic similarity. Both accounting are insightful, but here we will rely primarily on the continuous of status. For reasons of data availability, we use income or earnings

as the measure of economic is status, though income (the more inclusive measure) for most applications. preferable We use subscript p to refer to parental measures, while y is an individual's

Journal

of Economic Perspectives

economic constant,

status, adjusted so that its mean, y, is constant across generations, and e is a disturbance uncorrelated with yp. Thus, = y) + ey;

f} is a

y-~y

Py(yP

that is, the deviation of the offspring's economic status from the mean is fiy times the deviation of the parent from mean economic status, plus an error term. In the of work reviewed below, earnings, wealth and other measures income, empirical success are measured by their natural logarithm unless otherwise noted. economic income elasticity, is the percentage Thus, fiy, termed the intergenerational change in parents' in offspring's with a 1 percent economic success associated change economic success. The influence of mean economic status on the economics status 1 ? fiy, is called regression to the mean, since it shows that one may of the offspring, 1 ? f}y to be closer to the mean than one's parents by the fraction expect 1989). (Goldberger, The relationship intergenerational between the intergenerational income elasticity and the

correlation

is given by

Uy where of y. If measure across generations, so ay = ay, unchanging erational income elasticity exceeds py when ay is the standard is a common deviation deviation unit-free its standard y is a natural logarithm, is of inequality. Thus, if inequality then py = fiy. However, the intergen? is rising, but is less income inequality In effect, the intergenerational is declining.

than py when income inequality correlation of income while coefficient p is affected by changes in the distribution the intergenerational income elasticity is not. Also, p2 measures the fraction of the variance in this generation's measure of economic success that is linearly associated with the same measure Estimates in the previous of the intergenerational this issue) and Mulligan (1997). generation. income elasticity The mean are presented in Solon

for consumption, 0.68; for wealth, 0.50; 0.29. Evidence (or wages), 0.34; and for years of schooling, concerning earnings of income is mixed. Most trends in the degree across generations persistence rises with age, is greater for sons than daughters studies indicate that persistence is greater when multiple or earnings are averaged. The years of income of economic of averaging multiple years to capture permanent aspects importance in Mazumder He used a rich U.S. Social status is dramatized (forthcoming). and

(1999, are as follows:

estimates

reported in Mulligan for income, 0.43; for

data set to estimate an intergenerational income elasticity Security Administration 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 15 years are averaged.

The Inheritance

of Inequality

elasticities of this magnitude mean that "rags to intergenerational is no more than a fantasy for most poor children? The intergenerational correlation is an average measure and may be unilluminating about the probabilities of economic success conditional on being the child of poor or rich riches" these conditional and inspecting the entire parents. Calculating probabilities transition matrix gives a more complete The results of a study by Hertz picture. decile (2002) (from appear in Figure 1 with the parents arranged by income from left to right) and with adult children poor to rich moving arranged by income likelihood children's decile along of making decile. the other axis. The from the transition of the surface height the indicated parents' indicates decile the to the

Do

the underlying correlation of incomes in the data Though intergenerational set Hertz (2002) 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 chance of trap" to catch on). A son born to the top decile has a 22.9 percent the top decile (point D) and a 40.7 percent chance of attaining the top attaining quintile. A indicates that the son of the poorest decile has a 1.3 percent chance of attaining indicates the top decile and a 3.7 percent chance of attaining the top quintile. C that children of the poorest decile have a 31.2 percent chance of occu-

the lowest quintile, pying the lowest decile and a 50.7 percent chance of occupying while B shows that the probability that a child of the richest decile ends up in the the lowest poorest decile is 2.4 percent, with a 6.8 percent chance of occupying matrix and other studies suggest that distinct trans? quintile. Hertz's transmission mission mechanisms may be at work at various points of the income distribution and Heisz, 1999; Cooper, Durlauf and Johnson, 1994; Hertz, 2001). For (Corak example, wealth bequests may play a major role at the top of the income distribu? to violence or other adverse health episodes tion, while at the bottom, vulnerability be more important. (Hertz, may Mobility patterns by race also differ dramatically 2002). Downward mobility from the top quartile to the bottom quartile is nearly five times as great for blacks as for whites. Thus, whatever it is that accounts for their blacks do not transmit it to their children as effectively as do success, successful successful top quartile whites. blacks Correspondingly, at one half the rate of whites. born to the bottom quartile attain the

Sources

of

Persistence:

Cultural,

Genetic

and

Bequest generations. influence We seek to in?

Economic uncover comes. the We do

status channels this

does

offspring the intergenerational correlation (or the by decomposing income elasticity) into additive components the conintergenerational reflecting tribution of various causal mechanisms. This will allow us to conclude, for example,

persist substantially which parental through

across incomes

Journal

of Economic Perspectives

Figure 1 Intergenerational Income Transition Probabilities

tugher

Yugbet

Notes:The height of the surface in cell (i, j) is the probability that a person whose parents' household income was in the fth decile will have household income in the jth decile as an adult. The income of the children was measured when they were aged 26 or older and was averaged over all such years for which it was observed (an average of ten years). Parents' income was averaged over all observed years in which the child lived with the parents (an average of 9.4 years). Source: From PSID data, Hertz (2002). The 10 X 10 transition matrix on which this figure is based is available at (http://www-unix.oit.umass.edu/~gintis).

that a certain genetic

fraction

inheritance

of the intergenerational correlation of IQ or by the fact that the children fact about correlation coefficients

for by the of rich parents are also that this can be done.

is accounted

wealthy. It is a remarkable

we use does not require that we introduce variables in any Moreover, the technique particular order. Suppose that parents' income (measured by its logarithm, yp) and education (s) affect offspring income (also measured by its logarithm, y). offspring this intergenerational Like any correlation correlation coefficient, ry can be as the sum of the normalized coefficients of measures of expressed regression education regression (Py y) and offspring (pys) in a multiple each multiplied the correlation between and the regressor predicting y, by yp itself is just 1). A normalized (which, of course, for parental income regression is the change in the dependent coefficient units, variable, in standard deviation associated with a one standard deviation change in the independent variable. The parental income income direct effect of parental income is the normalized coefficient of parental regression from this regression. The education of of this decomposition component

Samuel Bowles and Herbert Gintis

Figure 2 Representing a Correlation as the Sum of Direct and Indirect Effects

the intergenerational correlation which gives breakdown,

is called

an indirect effect.2 Figure

2 illustrates

this

yyp intergenerational correlation

Pypy

'ypsHys effect + indirect effect.

= direct

As long as the multiple regression is coefficients are unbiased, the decomposition valid whatever the relationship the variables. Specifically, it does not require among This decomposition that the regressors be uncorrelated. allows us to be more precise parental number about effect our "black box" claim in the introduction. is a substantial fraction We mean that the direct in a of the intergenerational correlation this comparison 1972; Bowles and Nelson, (Bowles, effects. Note and that other

allowing 1974; Atkinson, 1999). Maynard and Trinder, 1983; Mulligan, Our strategy is to estimate the size of these direct and indirect decomposition uses correlations between parental in the example?thought These correlations with of course.

of studies

the

incomes

variables?schooling generating process. causal relationships, the correlations

to be causally related to the incomeincome need not reflect parental

But the above decomposition can be repeated for in parental income and the causes of offspring income, some cases permitting causal interpretations. For example, a study of the role of wealth in the transmission and offspring process could ask why parental income wealth are correlated. Is it bequests or the cultural and inter vivos transfers between of savings behaviors that account for this correlation? Or do we simply not know why parent and offspring wealth is correlated and as a result should avoid giving the data a causal interpretation? Similarly, parent-offspring similarity in human capital may be due to genetic or cultural inheritance of whatever it takes to and to acquire skills and behaviors that are rewarded in the persist in schooling for labor market. Unlike the models of parental and child behavior accounting and presented in this issue by Grawe and Becker persistence by pioneered 2 This transmission

decomposition can be found in Blalock (1964) and is described in the Appendix to this paper. Goldberger (1991) describes the standard regression model with normalized (mean zero, unit standard deviation) variables on which it is based.

10

Journal

of Economic Perspectives

not giving an adequate causal account Mulligan, our approach is more diagnostic, of the transmission but indicating where to look to find the causes. The process, next sections of this paper will explore such decompositions.

The

Role

of

Genetic

Inheritance channels

of

Cognitive

Skill

One of the transmission of its prima facie plausibility, discussions to it in popular

deserves

but also because

of the subject. scores on cognitive tests is cognitive skill. The similarity of parents' and offsprings' of IQ between and offspring well documented. Correlations parents range from of average parental and 0.42 to 0.72, where the higher figure refers to measures et al., 2000). The and McGue, 1981; Plomin IQ (Bouchard average offspring to earnings both directly and via schooling in a variety of studies that estimate determinants of earnings using IQ (and related) test scores. The direct effect of IQ on from multiple regression studies that typically use the loga? earnings is estimated rithm of earnings as a dependent variable and estimate the regression coefficients contribution attainment of cognitive has also been functioning established of explanatory on a cognitive test, variables, including performance other measures) of schooling, a measure of parental economic years (and perhaps effect of IQ social status, work experience, race and sex. The indirect and/or is to higher levels of educational attainment through its contribution operating to predict IQ (along with other variables) using measures of childhood the level of schooling obtained. We have located 65 estimates of the normalized coefficient of a test regression score in an earnings equation in 24 different studies of U.S. data over a period of three decades. Osborne deviation in Bowles, Gintis and Our meta-analysis of these studies is presented The mean of these estimates is 0.15, indicating that a standard (2002). variables change in the cognitive score, holding constant the remaining estimated of a variety

special attention not only because attention given of the extraordinary of This is the genetic inheritance

the natural logarithm of earnings by about onechanges deviation. the mean value of the normalized By contrast, in the same equation the coefficient of years of schooling predicting regression in these studies is 0.22, suggesting a somewhat natural log of earnings larger (including schooling), seventh of a standard effect of schooling. We checked to see if these results were dependent independent on the weight of overrepresented authors, the type of cognitive test used, at what the test was taken and other differences age among the studies and found no effects. An estimate of the causal impact of childhood IQ on years of is 0.53 (Winship and Korenman, (also normalized) 1999). A rough schooling estimate of the direct and indirect effect of IQ on earnings, call it b, is then b = = 0.266. 0.15 + (0.53) (0.22) significant Do these indirect inheritance causal direct and two facts?parent-child similarity in IQ and an important role for IQ in generating a major role for genetic earnings?imply of intergenerational economic of cognitive ability in the transmission

The Inheritance

of Inequality

11

status? One way to formulate this question is to ask how similar would parental and Also, offspring IQ be if the sole source of the similarity were genetic transmission. how similar would the incomes of parents and offspring be if there were no other transmission and channel? and some genetics (the details a few terms?phenotype, are in the Appendix and in Bowles For this we need

and the genotype, heritability A person's IQ?meaning, to many economists. a test score?is a phenotypic trait, while the genes influencing IQ are the person's is the relationship the two. Suppose between that, for a genotypic IQ. Heritability in genotype a standard deviation difference is associated with given environment, a fraction h of a standard deviation difference in IQ. Then h2 is the heritability of Gintis, 2001) genetic correlation?unfamiliar 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, Otto and Chris1997; Feldman, of statistical tiansen, 2000; Plomin, 1999). The genetic correlation is the degree association of parents and children, which is 0.5 if the parents' between genotypes are uncorrelated But couples tend to be more ("random genotypes mating"). similar in IQ than would occur by random mate choice ("assortative mating"), and this similarity is associated with an unknown m of their genotypes. The correlation effect is to raise the genetic correlation of decomposition Using the method of parent and offspring to (1 + m) / 2. in the previous section, the introduced to genetic IQ that is attributable offspring

and y between parental of IQalone is the heritability of IQtimes the genetic correlation. Thus, we have y = h2(\ + m)/ 2. The correlation between parent and offspring income that is attributable to genetic inheritance times the of IQ is this correlation inheritance normalized analogous correlation effect of IQ on the income of parents, times the (direct and indirect) effect for the offspring, or yb2. Another way to see this is to note that the between parental income and offspring IQ that we would observe were

correlation

the genetic inheritance of IQ the only channel at work is yb, and this times the effect of offspring IQ on earnings, which is b, gives the same result. of genetic above, we see that the contribution Using the values estimated of IQ to the intergenerational inheritance transmission of income is

(h2(l + 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 if only ballpark estimates) and the genetic inheritance of IQ were the reasonable, mechanism then the for intergenerational income transmission, only accounting be 0.01, or roughly 2 percent the observed correlation. Note the conclusion that the contribution of genetic intergenerational 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 intergenerational correlation would

12

Journal

of Economic Perspectives

and the heritability of IQ 100 percent without appreciably the qualitative conclusions. The estimate results from the fact that IQ is changing of economic success. just not an important enough determinant of genetic inheritance of IQ to parent-offspring Might the small contribution perfectly correlated error in the cognitive measures? similarity of incomes be the result of measurement There are two issues here. First, what is the reliability of the test: whatever the test measures, does it measure well? Second, what is the validity of the test: does the test measure the right thing? The concern that the tests are a very noisy measure is In fact, the tests are among the more reliable variables used in standard misplaced. where reliability is measured between tests earnings equations, by the correlation and retests, between odd and even numbered items on the tests, and by more methods. For the commonly used Armed Forces Qualification Test sophisticated (AFQT), for example?a of cognitive a measure successive between days. that the tests measure the wrong thing, is weightier and concern, less easy to address with any certainty. Could it be that cognitive skills not measured on existing test instruments are both highly heritable and have a major impact on a more substantial fraction of the transmis? earnings, thereby possibly explaining The second measures that are substantially process? The search for general cognitive of success in adult roles began with Edward uncorrelated with IQ and predictive Thorndike's Some alternative test instru? (1919) paper on "social intelligence." such as Robert Sternberg and collaborators' ments, "practical intelligence" predict et al., 1995; Williams and economic success in particular occupations (Sternberg fame and fortune that would have 1995). But despite the substantial Sternberg, three genera? accrued to success in this area, the quest that Thorndike launched tions ago has yielded no robust alternative to IQ, let alone one that is highly measured of economically but as yet unThus, the possible existence important but should at this heritable general cognitive skills cannot be excluded, stage be treated as speculation. to think that available estimates overstate the imporIndeed, we are inclined of earnings, since in many respects tance of general cognitive skill as a determinant heritable. sion skills?the test used to predict vocational success that is often used as correlation between two test scores taken on

days by the same person is likely to be higher than the correlation the same person's reported years of schooling or income on two successive

in either case results from taking a test is like doing a job. Successful performance a combination of ability and motivation, the disposition to follow instrucincluding to work ethic and other traits likely to contribute tions, persistence, independently one's earnings. This is the reason we eschew the common label of a test score as term "cognitive performance." skill," but rather use the more descriptive a leading student of cognitive testing, writes: "Low problem Eysenck (1994, p. 9), in an IQ test is a measure of performance; may influence solving personality rather than abstract intellect, with measurable effects on the IQ. An performance test lasts for up to 1 hour or more, and considerations of fatigue, vigilance, IQ "cognitive arousal, etc. may very well play a part." Thus, some of the explanatory power of the

Samuel Bowles and Herbert Gintis

13

earnings does not reflect cognitive measure in predicting to the successful other individual attributes contributing

cognitive

skill, but rather of tasks. performance

Genetic

and

Environmental

Inheritance

of IQ explains little of the intergenerational the genetic inheritance Although transmission this says nothing of other about the possible process, importance traits. Indeed, the remarkable transmitted income similarity of identical genetically twins compared to fraternal twins suggests that genetic effects may be important. We will use the similarity well as the environmental of income as the genetic heritability of intergenerational transmission. component But two words of caution our are in order. First, as we will demonstrate, estimates are quite sensitive to variations in unobserved it is Second, parameters. sometimes then that if the heritability of a trait is substantial, mistakenly supposed the trait cannot be affected of twins to estimate

view is dramatized

much by changing the environment. The fallacy of this the case of stature. The heritability of height estimated from by 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 and 11 inches?a bit taller than Norwegian and U.S. 8 inches a whopping taller than the Hadza huntermilitary gatherers in southern Africa (Floud, Wachter and Gregory, 1990). But the fact that recruits in 1761 were shorter than today's Hadza shows that even quite Norwegian heritable traits are sensitive to environments. from a What can be concluded average servicemen 5 feet and the Sudan

finding that a small fraction of the variance of a trait is due to environmental is that policies to alter the trait through variance environments will changed that differ from the environments on which the environments require nonstandard estimates Mandela are based. the case of South Africa, where roughly two-thirds in 1993 (the year before Nelson Consider

became president), of the intergenerational transmis? sion of earnings was attributable to the fact that fathers and sons are of the same race, and race is a strong predictor of earnings (Hertz, 2001). That is, adding race to an equation sons' earnings reduces the estimated effect of fathers' predicting earnings heritable Because the traits designated by over two-thirds. by "race" are highly and interracial parenting role of we thus find a substantial uncommon, in the intergenerational inheritance transmission of economic status. Yet, clear in the case of South Africa under apartheid traits derived that the economic

genetic it is especially

of the genetic inheritance of physical from environmen? importance tal influences. of skin color and other racial What made the genetic inheritance markers central to the transmission process were matters of public policy, not human nature, including the very definition of races, racial patterns in marriage and the discrimination suffered of the Thus, the determination by nonwhites. in a transmission genetic component process says little by itself about which public policy can or should level a playing field. the extent to

14

Journal

of Economic Perspectives

of heritability use data on pairs of individuals with varying degrees of For example, and fraternal twins are and environments. identical genes their upbringing, but fraternal twins are to similar environments during exposed less closely related genetically than identical twins. Under quite strong simplifying Estimates shared in the Appendix) one can exploit the variation in genetic assumptions (explained and environmental similarities among pairs of relatives to estimate heritability of a economic trait such as income, or other standard variables. years of schooling to use this method. The model underlying Taubman (1976) was the first economist affect human calculations assumes that genes and environment following which produces earnings, as the equation below indicates, but the effects of capital, to income are unaffected wealth and other contributions by genes and environ? the ment subsequently. and environments have additive First, genes assumptions. be correlated, and environment but the direct effect of "good effects?genes may of the quality of the is independent coefficient) genes" on earnings (its regression Here are the environment and conversely. = h(genes) Thus, an individual's earnings can be written effects. and will be introduced

earnings Second,

4- jS(environment)

4- idiosyncratic

are uncorrelated with (for the fraternals) the good-looking twin does (for example, de? not get more loving attention). Third, the environments affecting individual are as similar for members of fraternal pairs of twins as for the identical velopment within-pair genetic differences environmental differences within-pair twins pairs. ("random Fourth, mating"). of earnings and fraternal in best data and Solon the earnings Given these of the two parents are uncorrelated that the we show in the Appendix assumptions, is twice the difference between the earnings correla? genotypes twins. As the difference available?the Swedish between Twin these two correla? Registry studied by U.S. Twinsburg data set give an estimate of assumptions

(h2) heritability tions of identical tions is 0.2

sets

Bjorklund, Jantti studied by Ashenfelter

(forthcoming) and Krueger (1994)?these

and a smaller

h2 equal to 0.4. of genes for of random mating, the correlation Because, due to the assumption of fraternal the fraternal correlation twins' earnings twins is 0.5, the implied of twins' because of genetic factors is h2/2. The fact that the observed correlation this estimate is explained exceeds by the fact that twins share similar earnings about the degree of environments. Thus, once we know h2, we can use information effects of these environments to estimate how large the environmental similarity would the observed earnings correlations. have to be to generate are random mating and common environments The assumptions concerning of unrealistic and can be relaxed. First, we need an estimate of my, the correlation The relevant measure is the earnings potential (the parents' earnings genotypes. the degree of assortation, because correlation of actual earnings would understate many women do not work full time). The degree likely to be considerably larger than on genotype of assortation for the simple is on phenotype reason that the

The Inheritance

of Inequality

15

is the phenotype, not the genotype (which is unobservable), as we will see. and the two are not very closely related for the case of earnings, that the genotype for potential earnings of parents is half as similar as are Assuming the actual incomes of brothers, the correlation would be about 0.2. Second, note that because it was assumed that the environments experienced the two identical twins are not, on the average, more similar to the environments by of the two fraternal twins, the fact that within-twin-pair are less earnings differences for the identical twins must be explained entirely by their genetic similarity. But if the alike, identical twins for example) more similar environments (because they look experience than the fraternals, the estimate will overstate the degree of twins share more similar environments than fraternal

basis of the assortation

heritability. It is likely that identical twins and other

and Nichols, Otto and Chris1976; Feldman, siblings (Loehlin Rice and Reich, 1979; Rao et al., 1982). Estimates of the tiansen, 2000; Cloninger, of identical twins are more similar than those of extent to which the environments fraternal effects twins are quite imprecise, and we can do no better than to indicate the of using plausible alternative Just how sensitive the estimates assumptions. in the in the assumptions are to reasonable variations differences concerning of twins' environments can be estimated correlations by assuming some degree of statistical

association of genes and environment, but not iden? with the correlated than the tical genes of the fraternal twins giving them less correlated environments identical twins. Table environment 1 presents estimates based on various of this genesmagnitudes effect. As the assumed correlation between genes and environment of the environments of the identical twins rises, and some of the earnings similarity of the identical twins, the of heritability falls.3 We take the third numerical column of

the correlation increases, because this then explains

estimate resulting Table 1 as the most reasonable sions follow.

set of estimates. On this basis, two striking concluof earnings the First, the heritability Second, appears substantial. effects are also large. The normalized coefficient of environmental regression = 0.38, which on earnings is fie environment with the normal? may be compared ized regression of years of schooling in an earnings coefficient for a measure equation, attainment exhaustive. of ]8^ and correlation estimates) What is the intergenerational correlation of earnings implied by our estimate in addition to h and (3e, we require the h? To answer this question, of parents' with genes (which is already implied earnings by our and the correlation of parents' earnings with environment. The first from our earlier captures which is 0.22. Thus, while educational meta-analysis, it is far from of the relevant environments, aspects important

3 The Swedish Twin Registry data set assembled by Bjorklund, Jantti and Solon (forthcoming) has data not just on twins, but on many pairs with varying degrees of relatedness (half-siblings, for example) and may allow more robust estimates using the methods developed by Cloninger, Rice and Reich (1979), Rao et al. (1982) and Feldman, Otto and Christiansen (2000).

16

Journal

of Economic Perspectives

Table 1 Estimating the Heritability of Earnings

Assumed Correlation of Genes and Environment Heritability of Earnings (h2) Normalized Regression Coefficient: Genes on Earnings (h) Environment on Earnings (]3) Correlation of Environments: Fraternal Twins Identical Twins Notes:The association of genes with environment is represented by the normalized regression coefficient of genes on environment. This table assumes that parental earnings-determining genes are correlated 0.2, and the correlation of fraternal twins' environment is 0.7. We use the correlations of income for identical twins of 0.56 and of fraternal twins of 0.36, taken from the U.S. Twinsburg Study, and assume that these are also the correlations of earnings.

column the

is simply h times 2 gives our estimates. The genetic contribution between and or parental earnings offspring genotype, The environmental contribution, h2(l + m)/2. similarly, is (3e times a correlation of parents' and environment to yield a total 0.74) selected earnings (namely correlation

in Table

of 0.4. earnings correlation that genetic inheritance may account for almost one-third of the correlation is somewhat in light of our negative intergenerational unexpected, the inheritance of IQ. The surprising of both findings concerning importance intergenerational The estimate environment related contribution and genes point to a puzzle. to IQ and if the environmental If the genetic contribution contribution is much is not strongly larger than the

of years of schooling, what are the mechanisms for per? accounting sistence of income over the generations? We shall return to this puzzle, but will turn to data other than twins studies first to show that the same puzzle arises.

Human

Capital attainment schooling and perhaps other is persistent across generations and has clear traits that are rewarded in labor markets, an

Because links to skills account

of intergenerational status based on human capital has facie plausibility. The data already introduced of allow a calculation strong prima the portion of the intergenerational income correlation for by the fact accounted that offspring of high-income (measured in years). This parents get more schooling is the correlation of parent income and offspring schooling (about 0.45) multiplied of schooling in an earnings equation (0.22 by the normalized regression coefficient from our meta-analysis), or 0.10. This correlation is substantial, particularly in the light of the fact that it is restricted of IQ (because to the effects independently our estimate of years of schooling operating in of 0.22 is from earnings functions

of the transmission

Samuel Bowles and Herbert Gintis

17

Table 2 Contribution Transmission Income 0.20 0.09 0.12 0.41 of Environmental, Genetic and Wealth Effects to Intergenerational

Earnings Environmental Genetic Wealth Intergenerational correlation 0.28 0.12 0.40

Notes:The income column and the estimated contribution of wealth are discussed below. The environ? mental versus genetic breakdown assumes the figures in the third numerical column in Table 1.

which

the

regressors

include

contribution, including well as the direct effect It used

the AFQT test or a similar instrument). The full the effect of schooling on IQ and its effect on earnings as of schooling on earnings holding constant IQ is 0.12.

to be commonly assumed that once adequate measures of schooling the only effects of parental economic status on offspring quality were developed, effects on cognitive and schooling, earnings would operate through functioning But even as with the direct effect of parental status on offspring earnings vanishing. of school quality has improved over the years, the estimated of parental incomes on offspring earnings has turned direct effect (or earnings) out to be remarkably robust. For example, Mulligan (1999), using early 1990s data from the (U.S.) National Longitudinal the effect of Study of Youth, first estimated a change in the logarithm of parental earnings on offspring's logarithm of earnings without measures controlling of school for any other factors and then controlled for a number of and quality, as well as the AFQT and standard educational variables. He found that between two-fifths and one-half of the gross the measurement

demographic

statistical relationship of parental and offspring earnings remains (unconditional) even after controlling for the other factors. These results just reaffirm the black box data and methods: more than two-fifths of the puzzle using entirely different transmission coefficient is unaccounted for.4 intergenerational of the fact that the children of the well-to-do account Taking healthier are much

than poor children (Case, Lubotsky and Paxson, 2001) along with the fact that poor health has substantial effects on incomes later in life (Smith, 1999) would transmission pro? probably account for a substantial part of the intergenerational cess. The role of health in the process incomes by either parents appear to have strong impacts the health status of the parents and children. is particularly on child health striking because parental for that are not accounted similarity between

nor by the genetic

4 It is also true that we can typically statistically account for less than half of the variance of the earnings or income using the conventional variables described above. But this fact does not explain our limited success in accounting for the intergenerational correlation, as this correlation measures only that part of the variation of earnings that we can explain statistically by parental economic status.

18

Journal

of Economic Perspectives

Wealth

Effects

wealth

the inheritance of success can be passed on in a family through as well as inter vivos wealth transfers to children. Remarkably little scholarly in part because no representative has been given to this mechanism, attention of other earnings determinants exists for data set with adequate measures panel Economic of the age at which the inheritance has reached generation The only study of which we are aware that has been completed. typically to their deaths estimates addresses this problem by following the second generation the second wealth correlation than those reported by Mulli? higher intergenerational of wealth clearly matter for the 1979). But while inheritances gan, above (Menchik, of the income distribution, we doubt whether such transfers play an important top a much role for most families. Very few individuals receive inheritances on sufficient Mulligan magnitude. subject to inheritance of deaths (1997) estimates that estates passing tax in the United States constituted of significant wealth to be

which

wealth

between 2 and 4 percent Even though this figure leaves out some quite over the years 1960-1995. as well as transfers that occur during life, it seems unlikely substantial inheritances, status is transmitthat for most of the population a substantial degree of economic ted directly by the intergenerational transfer of property or financial wealth. It thus seems likely that the intergenerational of wealth reflects, at persistence least in part, parent-offspring similarities in traits influencing wealth accumulation, such as orientation toward the future, sense of personal efficacy, work ethic, attainment and risk taking. Some of these traits covary with the level of schooling

may be more likely to be risk averse, to of efficacy. Because of this correlation of to wealth accumulation, wealth with the traits conducive parent-offspring similarity in wealth may arise from sources independent of any bequests or transfers. wealth: discount for example, less well-off people the future and have a low sense Whatever parent-offspring their from for families with significant income source, fraction wealth similarities can contribute a substantial wealth, to the

methods of incomes. Using the same decomposition intergenerational persistence of parent income and child wealth as above, this contribution is the correlation times the normalized coefficient of wealth in an income equation. We regression use data from the Panel Study of Income Dynamics analyzed by Charles and Hurst between parent income and child wealth (both in natural (2002). The correlation in this data set is 0.24. The average age of the children is only 37 years, logarithms) does not capture inheritance of wealth at death of the parents. so this correlation

one way to proceed To get a rough idea of the normalized coefficient, regression is by starting with the percentage associated with a 1 percent change in income in wealth; this elasticity will range from virtually zero (for those with little change or no wealth) to one (for those with no source of income other than wealth). A plausible mean value (based on average factor income shares) for the U.S. popu? lation is 0.20. We convert this to a normalized coefficient by multiplying regression the ratio of the standard deviation of log wealth to the standard deviation of log by income, also from the PSID data set provided by Charles and Hurst (2002). This

The Inheritance

of Inequality

19

calculation dren

chil? suggests that the fact that higher income parents have wealthier contributes of incomes. correlation 0.12 to the intergenerational This figure, while substantial, as it is based on data may be an underestimate, that, for the reasons mentioned above, do not capture a key transmission process, of wealth upon the death of one's parents. inheritance the Moreover, namely be adjusted upward to take account of the fact that those with to have higher average returns to their wealth (Bardhan, greater Bowles and Gintis, 2000; Yitzhaki, 1987). Greater parental or own wealth may also raise the rate of return to schooling and other human investments, but we have no wealth tend way of taking account of wealth contribution of course. be nearly income distribution mentioned of this empirically. For a sample of very rich parents, the to the intergenerational correlation would be much higher, For a sample of families with very limited wealth, the contribution would zero. The difference in the contribution of wealth effects across the is a reflection earlier. of the heterogeneous Because of the very skewed (to which nature of the transmission of wealth, the is considerably distribution apply) estimate should

process family with the mean level of wealth wealthier than the median family.

our estimates

Group Thus

Membership

and

Personality

pins most determine

the production function which underfar, we have followed approach, economic to intergenerational to transmission, approaches seeking the contribution in ownership of factors of of parent-child similarity We have complemented the usual choice-based production. approach by including the influence of genetic inheritance. But other traits are persistent across genera? tions and are arguably as important?for number of siblings and others. children, earnings for women, while number of race, first language, example, For example, is a predictor of low obesity for men. Good looks predicts high earnings

predict whether

height for both men and women, the latter independently of earnings with the public (Hammermesh and Biddle, they hold jobs interacting evidence 1993). Bowles, Gintis and Osborne (2002) provide a survey of empirical these and many other nonskill determinants of economic success. concerning high Two such variables the intergenerational illustrate transmission the potential of nonskill factors in importance of economic status: group membership and

personality. success is influenced not only by a person's traits, but Suppose that economic also by characteristics of the group of individuals with whom the person typically interacts. Groups may differ in a variety of dimensions: average level of schooling, economic and wealth level. Groups may be residensuccess, cognitive functioning tial neighborhoods, ethnic or racial groups, linguistic groups, citizens of a nation or any other set of individuals who typically interact with one another. Group effects on economic including success are well documented conformist discrimination, effects and may arise for a number on behavior, differential of reasons, access to

20

Journal

of Economic Perspectives

information

and complementarities in production (Cooper, Durlauf, and Johnson, 1994; Durlauf, 2001; Borjas, 1995). Race apparently plays a significant role in the intergenerational transmission of economic success. This is suggested by the fact that for the United States, the correlation among brothers' earnings estimated by Bjorklund et al. (2002), namely falls by 0.10 when the sample is restricted to whites. Apparently, what brothers 0.43, almost always have in common, namely race, accounts for much of their similarity of income. The same is true of parents and their children. In the data set underFigure 1, the elasticity of offspring family income with respect to parents' income is 0.54, but the same elasticity for whites only is 0.43 and for blacks family is explained in only is 0.41 (Hertz, 2002). Parent-offspring similarity in income lying measure by the fact that "race" is transmitted across generations. important Using Hertz's estimates, we find that race (that is, the correlation of parents income with 0.07 to the intergenerational correlation. While this race) contributes offspring estimate is a bit lower that those suggested by the above data, it may nonetheless be an overestimate, as it is based on an income equation with the standard regressors, but without a measure of cognitive the inclusion of which would performance, lower the race coefficient somewhat. probably A second example of traits not found in a conventional function production but that contribute to intergenerational status transmission are dispositions such as a sense of personal efficacy, work ethic or a rate of time discount (present orienof these aspects of personality stems from the fact that in tation). The importance a large class of exchanges, the hiring of labor, borrowing and lending, or including the exchange of goods of uncertain to specify all relevant quality, it is impossible in a contract enforceable aspects of the exchange by the courts. Where this is the are influenced of trust, case, the actual terms of the exchange by the degree hard work and other dispositions of the parties to the exchange. For honesty, a very present-oriented employee will not value the employer's promise of in the future, conditional on hard work now. Instead, such employment an employee will require a higher wage to motivate hard work in the present and, example, continued is less likely to be employed. As another example, fatalistic workers who therefore, believe that the probability of job termination is unaffected by their own actions will be costly to motivate under this type of labor contract (Bowles, Gintis and Osborne, 2002). (Duncan The empirical importance of these traits is suggested in a number of studies and Dunifon, and Rubinstein, 1998; Heckman 2001; Kuhn and Wein-

berger, 2001; Heckman, forthcoming). Osborne has studied the economic and intergen? (forthcoming) importance erational of fatalism, as measured persistence by the Rotter Scale, a common measure of the degree to which individuals believe that important events in their lives are caused by external events rather than by their own actions. Her study of a of U.S men and their parents found that the score on the Rotter Scale sample before entry to the labor market has a statistically significant measured and large influence offspring. on earnings. the Rotter score is persistent from parents to Moreover, The normalized influence of the Rotter Scale on earnings in Osborne's

Samuel Bowles and Herbert Gintis

21

study is somewhat larger (in absolute value, namely ?0.2) ence of IQ in our meta-analysis of 65 studies discussed correlation of parental income with child fatalism is ?0.14. fatalism income channel to child

than the average influ? earlier. The estimated The contribution of the

to the intergenerational is the correlation of parent correlation fatalism multiplied the correlation from child fatalism to by 0.028?that income, is, (-0.2) (-0.14). subsequent Osborne (forthcoming) also studied a sample of women in England and found that measures

of social maladjustment taken at age eleven (the Bristol Social are strong predictors and withdrawal, of Scale), such as aggression Adjustment at age 33. The normalized of personality influence traits of aggression and earnings withdrawal on earnings is considerably of IQ. There are larger than the influence of intergenerational of personality traits in the Osborne's persistence data set, but other studies suggest that parent-child English similarity in measures of social maladjustment be quite high. For example, Duncan et al. (forthcom? may ing) found that deviant forms of behaviors of U.S. mothers were strong predictors of the same behaviors in daughters, including drug use, violent behaviors, early sex, from school and criminal convictions. Osborne's work thus suggests suspension that the intergenerational transmission channel cultural) may be an important tence of income. of personality explaining traits (whether genetic or the intergenerational persis? no measures

We know relatively little about the workings of the intergenerational transmis? for personality traits relevant to economic other than cognitive success, Kohn's However, (1969) functioning. study of child rearing values of parents sion process suggests in the workplace are traits, parents' experiences and passed on to children. his parent sample by the Kohn categorizes generalized that each experiences on the job, ranging from those degree of self-determination who are relatively unsupervised to those who are closely directed by superiors. Kohn found that parents emphasize children. with high levels of what he termed "occupational self-direction" and independence as values for their self-control, curiosity, happiness Those who are closely monitored at work emphasize by supervisors that at least for some

to external authority. Kohn concluded: or not, "Whether consciously conformity parents tend to impart to their children the lessons derived from their own social class and thus help prepare their children for a similar class position." The work by Osborne suggests that the degree of self-direction has significant effects on earn? ings later in life. Other work by Yeung, Hill and Duncan (2000) shows that parental church attendance, in social organizations and behavior, including membership such precautionary behavior as seat belt usage have significant on their impacts children's earnings.

Conclusion Recent evidence transmis? points to a much higher level of intergenerational than was previously thought to be the case. America may position

sion of economic

22

Journal

of Economic Perspectives

Table 3 The Main Causal Channel IQ, conditioned on schooling Schooling, conditioned on IQ Wealth Personality (fatalism) Race Total Intergenerational Correlation Accounted For Channels of Intergenerational Status Transmission in the U.S. Income 0.04 0.07 0.12 0.02 0.07 0.32

Earnings 0.05 0.10 0.03 0.07 0.25

Notes:For each channel, the entry is the correlation of parent income with the indicated predictor of offspring income, multiplied by its normalized regression coefficient in an earnings or income equation. The total is the intergenerational correlation resulting from these channels, in the absence of a direct effect of parents' status on offspring status. Source: Calculations described in text and Bowles and Gintis (2001).

still be the land of opportunity by some measures, but parental income and wealth are strong predictors of the likely economic status of the next generation. of intergenerational Our main objective has been to assess the extent transmission best been estimates and the mechanisms of the relative our for it. Table 3 summarizes accounting of the main causal channels we have importance

is the first, which able to identify. The only entry not previously explained of the correlation income and child IQ multi? between parental of the normalized effect of IQ on earnings, conditioned plied by our estimate other things, years of schooling. The estimates for IQ, schooling and on, among is an estimate personality to in the take income account column of the income adjusted in Bowles and Gintis (2001). Thus, normalized as described differences, suitably we do not take account of the way that these earnings determinants may affect the rate of return to one's wealth. By contrast, we assume that the race effect is of the same magnitude in determining the returns to both human capital and conventional earnings, be significantly greater). While the estimates in Table wealth (if the race effect on incomes its contribution to the intergenerational worked solely via an effect on correlation would earnings effect are simply those of earnings in the earnings differences on column

3 are quite imprecise, the qualitative results are not likely to be affected by reasonable The results are somealternative methods. of are important to the inheritance what surprising: wealth, race and schooling economic and, as we have seen above, the status, but IQis not a major contributor, transmission of IQ is even less important. genetic A policymaker design interventions success playing to level the playing field might use these seeking that would loosen the connection the between results economic a level to

of parents and the economic prospects of their children. field entail no correlation and child between parental ? There are important values

But does incomes

forthcoming)

(Swift, of family life and privacy that would be

The Inheritance

of Inequality

23

the fortunes of parents and compromised by any serious attempt to disconnect children completely. Rather than pursuing an abstract (and to our minds unattraca better approach might be to correlation, tive) objective of zero intergenerational ask which mechanisms of intergenerational transmission seem unfair, and to direct policies accordingly. is clearly generation The role unfair. of race in transmitting status from generation to the strong correlation between Many people regard health as morally suspect, and many feel the same way Other

parental income and child about high levels of wealth inheritance. sate for inherited heritance not disabilities. of good looks, an appropriate target consensus could be formed

Large majorities favor policies to compenof persistence?the mechanisms genetic in? for example?strike most people as unobjectionable and for compensatory Even if some policy interventions. on which of these mechanisms are morally suspect, the the possible incentive on child success would

would be far from clear. For example, policy implications effects on parental behaviors of reduced parental influence have to be estimated and considered.

Appendix Decomposing Correlation Coefficients and Estimating Heritability

Suppose parental earnings yp directly affects offspring earnings y, but offspring is also affected with vx and v2, that are correlated earnings by two variables, if ry Vi and ry are the correlations of parental earnings parental earnings.5 Then, with vx and v2, respectively, of ypy vlf and if the normalized coefficients regression and v2 predicting we have y are given by fiy y, fiViy and jS^ respectively, \ / This is the correlation its direct effect effect ypy Pypy ypviPviy ypv2Pv2y

into between parental and offspring earnings, decomposed first term), the effect via variable vx (the second term) and the (the via variable v2 (the third term). To derive this equation, we write y all variables = + PypyyP PViyvx + pVtyv2 + ey, to have zero mean and unit variance, and e is the above expressubstituting

(2) where

are normalized

with the independent variables. Then, sion for y into the expectation E[ ypy], and noting that if two variables (e.g., y and between these variables is the yp) have zero mean and unit variance, the correlation value of their product, we get expected uncorrelated

5 For

previous treatments of this material, see Rao, Morton and Yee (1976), Cloninger, Rice and Reich (1979), Rao et al. (1982) and Otto, Feldman and Christiansen (1994).

24

Journal

of Economic Perspectives

(3)

= = E [ypyp]Pypy + E^yJjS^ rypp E [ypy] = 1, E[vxy]

+ Efay]^,. = rvxy, and E[v2y] =

Since, given our normalization, l.6 we arrive at equation We now apply this method identical degrees Suppose genotypes, (4) where e is uncorrelated and fraternal twins.

E[ypyp]

rv^y,

to estimating heritability using data on similarity of A more general treatment, using pairs of varying in Feldman, Otto and Christiansen additively yx and y2, depend e1 and e2. Thus, for i = 1, 2, (2000). on their

of relatedness, is developed a family has two sons whose

earnings, gx and g2, and their environments, yt = $ee{ + hg{ + eyi

such that the variance

to unity. Note square root of the heritability of earnings. We assume the environment i depends both on his genotype g{ and the common family environment have for i=

variables in the model and is chosen with the independent of y{ is unity. The variances of e{ and g{ are also normalized that the normalized coefficient of genotype is then h, the regression e{ of brother E. We thus

(5) where such ee is uncorrelated that the variance earnings,

^feJJ+jS^+e*

1,2,

parental

earnings offspring in the coefficient effect of genes on environment the genotype gi of brother i is determined Finally, given (6) where by &= Pggf+

with the independent variables in the model and is chosen E as including the effect of of e{ is unity. We interpret factor that affects education and any other environmental the full and is shared by brothers. For simplicity, we include with E. j8^, so g{ is uncorrelated of the parents, the genotypes by

&ggn>

of father and mother, and (3g is the normalized gj and gm are the genotypes son's of father's (or mother's) coefficient predicting regression genotype (path) The structure of this model is illustrated in Figure 3. genotype.

To show that j8g is 1/2, suppose my is the correlation of maternal and paternal that the total effect of the additivity (meaning genes. Since we are assuming is the sum of the effects of each gene), we can derive ]8g for a single locus. genome to x it contributes We label each possible gene at this locus with the amount x so that E[x] = 0 and E[ x2] = 2. By basic genetics, a son earnings. We normalized

6 Note that the same argument holds if we replace the expectations, which refer to population values, with the sample means, variances and covariances. In this case, the statistical independence of the error terms and the independent variables is assured by construction, whereas on the population level this independence is assumed.

Samuel Bowles and Herbert Gintis

25

Figure 3 The Earnings

of Brothers

Notes:In this diagram, gj and gm are the genotypes of father and mother, g1 and g2 are the genotypes of brothers, E is the common environment of brothers, e1 and e2 are the total environment of brothers and y-^and y2 are the earnings of brothers. Here, my is the genetic relatedness of parents based on assortative mating and as explained below, Pg= 1/2, while h2 is the heritability of earnings. The path labeled fige represents the tendency of genes to affect the environments (fige > 0 means that identical twins experience more similar environments than fraternal twins).

inherits

one copy of the gene at the locus from each parent, say Xj from the father and xm from the mother. The value of genes at this locus for a son is then (x* + effect on economic success, xm)/2, assuming that both genes have equal expected

In addition to Xp the father has another gene we do here and throughout.7 with value z^ at this locus, with the same mean 0 and variance 2. The corresponding value for the father is then (ay + Z/)/2, The where Xj and Zj are uncorrelated. value for the mother is ( xm + zm) / 2, where zm is the mother's other corresponding Because of assortative mating, gene at this locus, and xm and zm are uncorrelated. each gene of the father Xp Zp is correlated with each gene of the mother xm, zm. my = + zw)2/4] The variance of the parents' genetic value at this locus is E[(xm = 1, and the covariance of father and son is E[ ( Xj + zj) ( Xj + E[ ( Xj + Zy)2/4] = (1 + the correlation of father's and son's genetic Therefore, xm)/4i] my)/2. value at this locus is the quotient of the previous two expressions, or which mv

(7)

;= P

+ M=r+-r.

The first term in this expression the direct path from father's genome represents to son's, and the second is the correlation of father's and mother's genetic value at the locus, my, multiplied by the direct path from mother to son at that locus. To see this, recall that the least squares estimator of bx in the regression equation

The actual value of a pair of genes at a locus can be higher or lower than their average value, of course, as when one gene is dominant or recessive.

26

Journal

of Economic Perspectives

to mean zero and variance b1x1 + b2x2 + ?, where x1? x2 and y are normalized with x1 and x2, is given by (Goldberger, and where e is uncorrelated 1991): unity, r W *1 = ? r 1 _ r-' r

= = r in the and rXiX2= rar Substituting In our case, ?x = j3g, r (1 + my)/2, above expression, we get fig = 1/2. we multiply the the correlation of fraternal twins' genotypes, To determine = 1, 2, and take expectations, giving right sides of (6) for i = = + (1/2) + 2(l/2mgmgf]

rf;ig2

E[glg2\

(1/2)*E[$

WJ

= (l/2)2(2

2my)

= (1 + my)/2,

which, consulting (7), confirms the standard result in genetics that fathers and sons brothers with the same parents on the other are on the one hand and nonidentical equally multiply related. of fraternal twins, we of environments To determine the correlation the right sides of (5) for i ? 1,2 and take expectations, giving = = 0l + r%gfi% 0* + (1 + m,)fi%/2. we get

<* Finally, multiplying

the right sides of (4) for i = 1, 2 and taking expectations,

which

expands =

to

(8)

rfiy2

#(/3|+

(1 + my)Pl/2)

+ h\\

+ my)/2

+ (1 + m,)Pfiffh. but now the correlation

In the case of identical of genotypes

twins, the same figure is relevant, of brothers is r1* = 1. We then

and id _ n2 id + 72 id + n ujd ^ n foid Per exe2 ^ n ' glg2 ^ Penr elg2 + Pe"<re2gl> ryiy2 which becomes =

(9)

r?,

0K01

+ 0p

+ h* + 20 j3?A.

The Inheritance

of Inequality

27

In the text, we assume r^ = 0.9 for identical twins (although our results are not ? The two equations for the sensitive to this assumption), so fte = V0.9 /3?. very correlations of brother earnings, (8) and (9), together with the observed values of h and f3e for various and (9) imply that the difference between (8) Equations of identical and fraternal twins is given by earnings correlations, = (1 my)(h + 0e0^)2/2. these allow us to determine values of f3g. the correlations of

(10) Note that

r?,

r??

of h2, while assuming greater assortative mating raises the estimate a stronger tendency for genes to effect environment (raising f5ge) has the assuming as one would expect. In the literature, it is often assumed that my = opposite effect, = 0, in which case we 0 and ji for estimating get the standard equation heritability: ** = 2(r?, -

(ID

rfiy2). from this equation fie. and then use this

If this is the case, we can estimate h2 directly estimate of h2, together with (8), to estimate ? We would

like to thank Jere Behrman, Anders Bjorklund, Kerwin Kofi Charles, De Long, Williams Dickens, Marcus Eeldman, James Heckman, Tom Hertz, Bradford Erik Hurst, Arjun Jayadev, Christopher Jencks, Alan Krueger, John Loehlin, Casey Mulligan, Suresh Naidu, Robert Plomin, Cecelia Rouse, Michael Waldman and Elisabeth Wood for their contributions to this paper, Bridget Longridge and Bae Smith for research assistance support. and the John D. and Catherine T. MacArthur Foundation for financial

References Arrow, Kenneth, Samuel Bowles, and Steven Durlauf. 1999. Meritocracy and EconomicInequal? ity. Princeton, N.J.: Princeton University Press. Ashenfelter, Orley and Alan Krueger. 1994. "Estimates of the Economic Return to School? ing from a New Sample of Twins." American Economic Review. December, 84:5, pp. 1157? 172. Atkinson, A. B., A. K. Maynard and C. G. Trinder. 1983. Parents and Children:Incomes in Two Generations. London: Heinemann. Bardhan, Pranab, Samuel Bowles and Herbert Gintis. 2000. "Wealth Inequality, Credit Constraints, and Economic Performance," in Handbookof IncomeDistribution. Anthony Atkinson and Francois Bourguignon, eds. Dortrecht: NorthHolland, pp. 541-604. Becker, Gary S. 1988. "FamilyEconomics and

28

Journal

of Economic Perspectives

Macro Behavior." American Economic Review. March, 78:1, pp. 1-13. Becker, Gary S. and Nigel Tomes. 1986. "Human Capital and the Rise and Fall of Fam? ilies." Journal of Labor Economics.July, 4:3, pp. Sl-39. Bjorklund, Anders, Markus Jantti, and Gary Solon. Forthcoming. "Influences of Nature and Nurture on Earnings: An Early Progress Report on a Study of Various Sibling Types in Sweden," in Unequal Chances:Family Backgroundand Eco? nomicSuccess. Samuel Bowles, Herbert Gintis and Melissa Osborne, eds. New York: Russell Sage Foundation. Bjorklund, Anders et al. 2002. "Brother Cor? relations in Earnings in Denmark, Finland, Norway and Sweden Compared to the United States." Journal of Population Economics.Forth? coming. in Blalock, Hubert. 1964. CausalInferences NonResearch.Chapel Hill, N.C.: Univer? experimental sity of North Carolina Press. Blau, Peter and Otis Dudley Duncan. 1967. The American OccupationalStructure.New York: Wiley. Borjas, George. 1995. "Ethnicity, Neighborhoods, and Human Capital Exernalities." Ameri? can EconomicReview. June, 85:3, pp. 365-90. Bouchard, T. J. Jr. and M. McGue. 1981. "Familial Studies of Intelligence." Science.212: 4498, pp. 1055-059. Bowles, Samuel. 1972. "Schooling and In? equality from Generation to Generation." Jour? nal of Political Economy. May/June, 80:3, pp. S219-51. Bowles, Samuel and Herbert Gintis. 2001. "The Inheritance of Economic Status: Educa? tion, Class and Genetics," in International Encyclopedia of the Social and BehavioralSciences:Genetics, Behavior and Society,Volume6. Marcus Feldman and Paul Baltes, eds. New York: Oxford Univer? sity Press and Elsevier, pp. 4132-141. Bowles, Samuel and Valerie Nelson. 1974. "The 'Inheritance of IQ' and the Intergenera? tional Reproduction of Economic Inequality." Reviewof Economicsand Statistics.February, 56:1, pp. 39-51. Bowles, Samuel, Herbert Gintis, and Melissa Osborne. 2002. "The Determinants of Individual Earnings: Skills, Preferences, and Schooling." Journal of EconomicLiterature.December, 39:4, pp. 1137-176. Bowles, Samuel, Herbert Gintis, and Melissa Osborne, eds. Forthcoming. Unequal Chances: Family Background and Economic Success. New York: Russell Sage Foundation. Case, Anne, Darren Lubotsky and Christina

Paxson. 2001. "Economic Status and Health in Childhood: The Origins of the Gradient." NBER Working Paper No. W8344, June. Charles, Kerwin Kofi and Erik Hurst. 2002. "The Correlation of Wealth Across Genera? tions." University of Chicago Working Paper. Cloninger, C. Robert, John Rice and Theodore Reich. 1979. "Multifactorial Inheritance with Cultural Transmission and Assortative Mat? ing. II. A General Model of Combined Polygenic and Cultural Inheritance." AmericanJournal of Human Genetics.29:31, pp. 176-98. Cooper, Suzanne, Steven Durlauf and Paul Johnson. 1994. "On the Evolution of Economic Economic Re? Status across Generations." American view. 84:2, pp. 50-58. Corak, Miles and Andrew Heisz. 1999. "The Intergenerational Earnings and Income Mobil? ity of Canadian Men: Evidence from the Longitudinal Income Tax Data."Journal of Human Re? sources.34:3, pp. 505-33. Devlin, Bernie, Michael Daniels and Kathryn Roeder. 1997. "The Heritability of IQ." Nature. July 31, 388, pp. 468-71. Duncan, Greg J. and Rachel Dunifon. 1998. "'Soft-Skills'and Long-Run Market Success." Re? searchin LaborEconomics.17, pp. 123-50. Duncan, Greg, et al. Forthcoming. "The Apple Does Not Fall Far from the Tree," in Un? equal Chances:Family Background and Economic Success. Samuel Bowles, Herbert Gintis and Melissa Osborne, eds. New York: Russell Sage Foundation. Durlauf, Stephen. 2001. "A Framework for the Study of Individual Behavior and Social Interactions." SociologicalMethodology. 31:1, pp. 47-87. Erikson, R. and J. H. Goldthorpe. 1992. The ConstantFlux: A Study of Class Mobilityin the ln? dustrial Societies. Oxford: Oxford University Press. Eysenck, H. J. 1994. "Intelligence and Introversion-Extraversion,"in Personalityand In? Robert J. Sternberg and Patricia Ruzgis, telligence. eds. Cambridge: Cambridge University Press, pp. 3-31. Feldman, Marcus W., Sarah P. Otto and Freddy B. Christiansen. 2000. "Genes, Culture, and EconomicIn? and Inequality," in Meritocracy equality. Kenneth Arrow, Samuel Bowles and Steven Durlauf, eds. Princeton, N.J.: Princeton University Press, pp. 61-85. Floud, Roderick, Kenneth Wachter and Annabel Gregory. 1990. Height, Health and History: Nutritional Status in the United Kingdom, 1750-1980. Cambridge: Cambridge University Press.

Samuel Bowles and Herbert Gintis

29

Fong, Christina. 2001. "Social Preferences, Self-Interest, and the Demand for Redistribu? tion." Journal of Public Economics.82:2, pp. 22546. Fong, Christina, Samuel Bowles and Herbert Gintis. 2002. "Reciprocity and the Welfare State," in Handbook on the Economicsof Giving, Reciprocity and Altruism. Jean Mercier-Ythier, Serge Kolm, and Louis-Andre G'rard-Varet,eds. Amsterdam: Elsevier. Forthcoming. Goldberger, Arthur S. 1989. "Economic and Mechanical Models of Intergenerational Trans? mission." AmericanEconomicReview.June, 79:3, pp. 504-13. Goldberger, Arthur S. 1991. A Coursein Econometrics. Cambridge, Mass.: Harvard University Press. Hammermesh, Daniel S. and Jeff E. Biddle. 1993. "Beauty and the Labor Markets." NBER Working Paper 4518, November. Heckman, James. Forthcoming. The GED. Heckman, James and Yona Rubinstein. 2001. "The Importance of Noncognitive Skills: Lessons from the GED Testing Program." AmericanEco? nomicReview.May, 91:2, pp. 145-49. Hertz, Thomas. 2001. "Education, Inequality and Economic Mobility in South Africa." Unpublished Ph.D. Dissertation, University of Mas? sachusetts. Hertz, Thomas. 2002. "Intergenerational Economic Mobility of Black and White Fami? lies in the United States." Paper presented at the Society of Labor Economists Annual Meeting, May. Kohn, Melvin. 1969. Class and Conformity. Homewood, 111.: Dorsey Press. Kuhn, Peter and Catherine Weinberger. 2001. "Leadership Skills and Wages." Institute for So? cial, Behavioral and Economic Research, December. Loehlin, John and Robert Nichols. 1976. HeAustin, Texas: Environment,and Personality. redity, University of Texas Press. Mazumder, Bhashkar. Forthcoming. "Earn? ings Mobility in the U.S.: A New Look at In? tergenerational Inequality," in Unequal Chanc? es: Family Background and Economic Success. Samuel Bowles, Herbert Gintis and Melissa Osborne, eds. New York: Russell Sage Founda? tion. Menchik, Paul. 1979. "Inter-Generational Transmission of Inequality: An Empirical Study of Wealth Mobility." Economica.November, 46: 184, pp. 349-62. Mulligan, Casey. 1997. Parental Prioritiesand EconomicInequality. Chicago: University of Chicago Press.

Mulligan, Casey. 1999. "Galton vs. the Human Capital Approach to Inheritance." Journal ofPolitical Economy. December, Part 2, 107:6, pp. S184-224. Osborne, Melissa A. Forthcoming. "Personal? ity and the Intergenerational Transmission of Economic Status," in Unequal Chances:Family Backgroundand EconomicSuccess.Samuel Bowles, Herbert Gintis and Melissa Osborne, eds. New York: Russell Sage Foundation. Otto, Sarah P., Marcus W. Feldman and Freddy B. Christiansen. 1994. "Genetic and Cul? tural Transmission of Continuous Traits." Stanford University, Morrisson Institute Work? ing Paper No. 64. Plomin, Robert. 1999. "Genetic and General Cognitive Ability." Nature. December, 402:2, pp. c25-c29. Plomin, Robert, et al. 2000. BehavioralGenetics. New York: W. H. Freeman and Company. Rao, D. C., N. E. Morton and S. Yee. 1976. "Resolution of Cultural and Biological Inheri? tance by Path Analysis." American Journal of Hu? man Genetics.28, pp. 228-42. Rao, D. C. et al. 1982. "Path Analysis Under Generalized Assortative Mating. II: American Research. 187-98. 39, IQ." Genetical Smith, James. 1999. "Healthy Bodies and Thick Wallets: The Dual Relation between Health and Economic Status."Journal ofEconomic Perspectives. Spring, 13:2, pp. 145-66. Solon, Gary R. 1992. "Intergenerational In? come Mobility in the United States." American EconomicReview. June, 82:3, pp. 393-408. Solon, Gary R. 1999. "Intergenerational Mo? bility in the Labor Market,"in Handbookof Labor Economics, Volume 3A. Orley Ashenfelter and David Card, eds. Amsterdam: North-Holland, pp. 1761-800. Sternberg, Robert J. et al. 1995. "Testing Com? mon Sense." American November, 50: Psychologist. 11, pp. 912-27. Swift, Adam. Forthcoming. "Justice, Luck and Family Values: The Intergenerational Transmission of Economic Status from a Normative Perspective," in Unequal Chances:Family Background and Economic Success. Samuel Bowles, Herbert Gintis and Melissa Osborne, eds. New York: Russell Sage Foundation. Taubman, Paul. 1976. "The Determinants of Earnings: Genetic, Family, and Other Environ? ments; A Study of White Male Twins." American EconomicReview.December, 66:5, pp. 858-70. Thorndike, Edward L. 1919. "Intelligence and Its Uses." Harper'sMonthlyMagazine.December/ January, 140, pp. 227-35.

30

Journal

of Economic Perspectives

Williams, Wendy M. and Robert J. Sternberg. 1995. Success Actsfor Managers.Florida: Harcourt Brace. Winship, Christopher and Sanders Korenman. 1999. "Economic Success and the Evolution of Schooling with Mental Ability," in Earning and Learning: How SchoolsMatter. Susan Mayer and Paul Peterson, eds. Washington, D.C.: Brookings Institution, pp. 49-78. Yeung, Jean, Martha Hill and Greg Duncan.

2000. "Putting Fathers Back in the Picture: Pa? rental Activities and Children's Adult Attainments." Marriage and Family Review. Part I, 29: 2-3, pp. 97-113. Yitzhaki, Shlomo. 1987. "The Relation Be? tween Return and Income." Quarterly Journal of Economics. February, 102:1, pp. 77-95. Zimmerman, David J. 1992. "Regression Toward Mediocrity in Economic Stature." American Economic Review. June, 82:3, pp. 409-29.