Schooling in Capitalist America Revisited Author(S): Samuel Bowles and Herbert Gintis Source: Sociology of Education, Vol

Home , Herbert Gintis

Schooling in Capitalist America Revisited Author(s): Samuel Bowles and Herbert Gintis Source: Sociology of Education, Vol. 75, No. 1, (Jan., 2002), pp. 1-18 Published by: American Sociological Association Stable URL: http://www.jstor.org/stable/3090251 Accessed: 10/08/2008 17:04

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/action/showPublisher?publisherCode=asa.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.

JSTOR is a not-for-profit organization founded in 1995 to build trusted digital archives for scholarship. We work with the scholarly community to preserve their work and the materials they rely upon, and to build a common research platform that promotes the discovery and use of these resources. For more information about JSTOR, please contact [email protected].

http://www.jstor.org I

Schooling in Captalist America Revisited

Samuel Bowes and Herbert Gintis Universityof Massachusetts and Santa Fe Institute

Recent research has entirely vindicated the authors' once-controversial estimates of high levels of intergenerational persistence of economic status, the unimportance of the heritabilityof IQ in this process, and the fact that the contribution of schooling to cognitive development plays little part in explaining why those with more schooling have higher earings. Additional research has supported the authors' hypotheses concerning the role of personality traits, rather than skills, per se, as determinants of labor market seuccess.Recent contributions to the study of cultural evolution allow the authors to be considerably more specific about how behaviors are learned in school.

U _ he project that eventually result- the contribution of schooling to individual ed in the publication of Schooling economic success could be explained only in CapitalistAmerica (Bowles and partly by the cognitive development fos- Gintis 1976) began in 1968, stim- tered in schools. We advanced the position ulated by the then-raging academic that schools prepare people for adult work debates and social conflicts about the rules by socializing people to function well structure and purposes of education. We and without complaint in the hierarchical were then, and remain, hopeful that edu- structure of the modern corporation. cation can contribute to a more productive Schools accomplish this goal by what we economy and a more equitable sharing of called the correspondenceprinciple, namely, its benefits and burdens, as well as a soci- by structuring social interactions and indi- ety in which all are maximally free to pur- vidual rewardsto replicate the environment sue their own ends unimpeded by preju- of the workplace. We thus focused atten- dice, the lack of opportunity for learning, tion not on the explicit curriculum but on or material want. Our distress at how woe- the socialization implied by the structureof fully the U.S. educational system was then schooling. Our econometric investigations failing these objectives sparked our initial demonstrated that the contribution of collaboration. The system's continuing fail- schooling to later economic success is ure has prompted our recent return to the explained only in part by the cognitive skills subject. learned in school. The three basic propositions of the book Second, we showed that parental eco- deal with human development, inequality, nomic status is passed on to children, in and social change. Concerning human part, by means of unequal educational development, we showed that while cogni- opportunity, but that the economic advan- tive skillsare important in the economy and tages of the offspring of higher social-sta- in predicting individual economic success, tus families go considerably beyond the

Sociology of Education 2002, Vol. 75 (January): 1-18 I 2 Bowles and Gintis superior education they receive. We used the erational persistence of economic status (see then-available statistical data to demonstrate the firstsection), the unimportance of the her- that the United States fell far short of the goal itability of IQ in this process (see the second of equal economic opportunity and that section), and the fact that the contribution of genetic inheritance of cognitive skill-as mea- schooling to cognitive development plays lit- sured on standard tests-explains only a small tle part in explaining why those with more part of the intergenerational persistence of schooling have higher earnings (see the third status within families. section). Some additional research has sup- Finally,our historical studies of the origins ported our hypotheses concerning the role of of primaryschooling and the development of personality traits, rather than skills per se, as the high school suggested that the evolution determinants of success in the labor market of the modern school system is not account- (see the fourth section). But progress has been ed for by the gradual perfection of a democ- halting in this area. We survey some of this ratic or pedagogical ideal. Rather, it was the recent research in recent and forthcoming product of a series of conflicts arising through articles (Bowles and Gintis forthcoming a, the transformation of the social organization forthcoming b; Bowles, Gintis, and Osborne of work and the distribution of its rewards. In 2001, forthcoming). In the fifth section, we this process, the interests of the owners of the turn to the socialization process of schooling leading businesses tended to predominate itself. In Schoolingin CapitalistAmerica, we did but were rarely uncontested. The same con- not explore the individual-level learning flict-ridden evolution of the structure and processes that account for the effectiveness of purposes of education was strikingly evident the correspondence principle. Contributions in higher education at the time we wrote, and to the study of culturalevolution (Bowles and we devoted a chapter to what we termed the Gintis 1986; Boyd and Richerson 1985, contradictions of higher education. Later, in Cavalli-Sforzaand Feldman 1981) allow us to Democracyand Capitalism(Bowles and Gintis be considerably more specific about how 1986), we developed the idea that schools behaviors are learned in school. and the public sector generally are loci of conflicts stemming from the contradictory rules of the marketplace, the democratic poli- INTERGENERATIONAL INEQUALITY ty, and the patriarchalfamily. How do we now view Schooling in At the time we wrote Schooling in Capitalist CapitalistAmerica? For most of the quarter of America, there was a virtual consensus that a century since it was published, we have the statistical relationship between parents' researched subjects that are quite removed and children's adult economic status is rather from the questions we addressed in that weak. The early research of Blau and Duncan book. In recent years, however, we have (1967), for instance, firmly supported this returned to writing about school reform; how view. Even 20 years later, researchershad not economic institutions shape the process of changed their minds. For instance, Becker human development; and the importance of and Tomes (1986) found that the simple cor- schooling, cognitive skill, and personality as relations between parents' and sons' income determinants of economic success and their or earnings (or their logarithms) averaged role in the intergenerational perpetuation of 0.15, leading the authors to conclude that, at inequality. least for white men, "[a]lmost all earnings In light of the outpouring of quantitative advantages and disadvantages of ancestors research on schooling and inequality in the are wiped out in three generations" (p. S32). intervening years, the statistical claims of the Indeed, Becker (1988:10) expressed a widely book have held up remarkablywell. In partic- held consensus when, in his presidential ular,recent researchby us and others using far address to the American Economics better data than were available in the early Association, he concluded that "low earnings 1970s has entirely vindicated our once-con- as well as high earnings are not strongly troversialestimates of high levels of intergen- transmitted from fathers to sons." Schooling in Capitalist America Revisited 3

= But the appearance of such high levels of y (1 - Ry)y + y yp + Ey (1) intergenerational mobility was an artifact of two types of measurement error: mistakes in We use subscript "p"to refer to parental mea- reporting income and transitory components sures, so y is an individual'seconomic status, in current income uncorrelated with underly- adjusted so that its mean, y, is that of the ing permanent income (Atkinson, Maynard, parental generation, Ryis a constant, yp is the and Trinder 1983; Solon 1992; Zimmerman individual'sparental y, and ey is a disturbance 1992). The low validity in both generations' uncorrelated with yp. Rearrangingterms, we incomes depressed the intergenerational cor- see that relation, and when corrected, the intergenerational correlations for economic status now y- y = gy (yp- y) + y (2) appear to be quite substantial, on the order of twice or three times the average of the U.S. that is, the deviation of the offspring'sincome studies surveyed by Becker and Tomes from the mean income is By times the devia- (1986). The intergenerational correlations tion of the parent from mean income, plus an surveyed by Mulligan (1997) for family con- error term. We term Bythe "Galton measure" sumption, wealth, income, and earnings of intergenerational persistence (Galton used average, respectively, 0.68, 0.50, 0.43, and it to study the intergenerational persistence of 0.34. The upward adjustment of the consen- height, which he found to be two-thirds). The sus estimates of the extent of intergenera- influence of mean income on the income of tional inequality has stimulated a revival of the offspring, 1 - By,measures what is called empirical research on the mechanisms that regression to the mean, for, as Equation 2 account for parent-offspring similarityin eco- makes clear, one may expect to be closer to nomic status (see Behrman, Pollak, and the mean than one's parents by the fraction Taubman 1995; Mulligan 1997). 1 - Yr.The relationship between the Galton Thus, Schooling in Capitalist America was measure and the intergenerationalcorrelation correct: The extent of intergenerational eco- is given by nomic status transmission is considerable. In the United States, knowing the income or ry= Ry S wealth of someone's parents is about as infor- mative about the person's own economic sta- where Sy is the standard deviation of y. We tus as is knowing the person's years of school- measure economic success using natural log- ing attained or score on a standardized cog- arithms, By is the percentage change in off- nitive test. spring's economic success associated with a 1 To show how we support this assertion, we percent change in parents' economic success. represent the income of a member of the cur- Table 1 presents estimates of the Galton mea- rent generation as the sum of the effect of the sure. The extent of persistence-especially for parents' income, the mean income in the sec- income, wealth, and consumption-is sub- ond generation, and an error term. stantial.

Table 1. Intergenerational Persistence of Some Economic Characteristics(Galton Coefficient)

EconomicCharacteristic Numberof Estimates Range Average Log familyconsumption 2 0.59-0.77 0.68 Log familywealth 9 0.27-0.76 0.50 Log familyincome 10 0.14-0.65 0.43a Log earningsor wages 16 0.11-0.59 0.34a Yearsof schooling 8 0.14-0.45 0.29a

Source:Mulligan (1999). alf recentstudies of the UnitedStates only are included,these averagesare 0.35, 0.33, and 0.38, respectively. 4 Bowles and Gintis

How different are the probabilities of eco- ure is 0.36, while the analogous correlation nomic success for the children of the poor for single year for each (1975 and 1993, and the well off? Can the measures of persis- respectively) is only 0.16. Though the under- tence in Table 1 be translated into probabili- lying intergenerational correlation of incomes ties of obtaining high or low incomes condi- is a modest 0.36, the differences in the likely tional on the income level of one's parents? life trajectoriesof the children of the poor and The intergenerationalcorrelation coefficient is the rich are substantial. The "twin peaks" rep- a greatly oversimplified measure and may be resent those stuck in poverty and affluence unilluminating about the probabilitiesof eco- (though we do not expect the term "afflu- nomic success conditional on being the child ence trap" to catch on). Point A, for example of poor, rich, or middling parents. Calculating indicates that a son born to the top decile has these conditional probabilitiesand inspecting better than a 1 in 5 chance of attaining the the entire transition matrix gives a more com- top decile, while Point B indicates that for the plete picture. The results of a study by Hertz son of the poorest decile, the likelihood is 1 in (2001 b) appear in Figure 1 with the parents 100. Point C indicates that sons of the poor- arranged by income decile (from poor to rich est decile have a 19 percent probability of moving from left to right) and with adult sons attaining the lowest decile. Hertz's transmis- arranged by income decile along the other sion matrix and other studies (Cooper, axis. The height of the surface in cell (ij) is the Durlauf, and Johnson 1994; Corak and Heisz probability that an adult (aged 30 or over) 1999; Hertz, 2001a) suggest that distinct whose parents are in the ith decile of income transmission mechanisms may be at work at will have an income in the jth decile. The various points of the income distribution. For income of sons was averaged of the years example, wealth bequests may play a major 1984-93, and the parents' income was aver- role at the top of the income distribution, age over the years 1975-93. The simple (age- while vulnerability to violence or other adjusted) correlation of parents' and sons' adverse health episodes may be more impor- incomes in the data set represented in the fig- tant at the bottom.

Figure1. IntergenerationalIncome Transition Probabilities. Source: Hertz (2001b), which includes the 10 transitionmatrix.

- - 24% --s -2- ) Piobabily ofsofhsof"---s " -R1 chestt&oirchest deci'e (22%) de:ileis in pooret decie (19% ) ' D : ' 18% W -T ------F A% . (0%

12% - - L--( - -O------

_ 120, 'L-: ..... 0%

6 8 10 iecie Parent Income Decile to 1CCl to ui) (low to high) (o in America Revisited 5 Schooling-- Capitalist --

INTERGENERATIONALSTATUS Correlations of IQ between parents and off- TRANSMISSION spring are substantial, ranging from 0.42 to 0.72, the higher figure referring to average What accounts for the transmission of eco- parental versus average offspring IQ and McGue The contribu- nomic status from parents to offspring? There (Bouchard 1981). tion of to has are only a few income-generating traits for cognitive functioning earnings which both economic relevance and similari- been established using survey data to esti- of and have been mate the natural logarithm of earnings y as a ty parents offspring empiri- function of a measure of economic demonstrated. them are parental cally Among cogni- social status tive the level of and and/or yP, years (and perhaps performance, schooling, other of and of assets. Our estimates measures) schooling s, perfor- ownership (Bowles mance on a score in U.S. and Gintis that the cognitive c-often, forthcoming b) suggest data sets, the Armed ForcesQualification Test. fact that have off- wealthy parents wealthy We located 65 estimates of the normalized spring plays a substantial role in the intergen- coefficient of a test score in an erational transmission of income. But here we regression earnings equation for the United States over focus on and schooling cognitive perfor- three decades. These estimates appear in mance as concerns more central to the soci- Figure 2, where the vertical axis is the esti- ology of education. mated coefficient and the horizontal axis We treat income as a trait influ- phenotypic gives the year to which the data apply. The enced the individual's and by genotype g mean of these estimates, 0.15, indicating that environment e. Genotypic and environmental a standard deviation change in the cognitive influences jointly determine individual skills score, holding constant the remaining vari- and other traits relevant to job performance. ables, changes the natural logarithm of earn- Among the environmental influences are cul- ings by about one-seventh of a standard devi- tural transmission from parents, schools, and ation. By way of contrast, the mean value of other learning environments. the normalized regression coefficient of years How important is the transmission of IQ in of schooling in these studies is 0.22, suggest- the intergenerational transmission process? ing a somewhat larger independent effect of

Figure 2. Normalized Regression Coefficient of Cognitive Score on the Logarithmof Income or Earnings by Year:65 Estimates from 24 Studies. Source: Bowles et al. (forthcoming).

fyc 0.35 i i 1 I 0.3- . .

0.25 - - 0.2- . ? * - o : 0.15- * . * 0 0.1 - *

0.05 - * ~~~0~~~~~~~~~~~ -0.05 - * _n-. 1 it I l I i I 1960 1965 1970 1975 1980 1985 1990 1995

year 6 6 Bowles and nd GintiGintis schooling. There is no apparent trend in the tional correlation be if there was no genetic estimated importance of cognitive perfor- inheritance of IQ, that is, if the correlation of mance as a determinant of earnings, casting parental and child genotypic IQ was zero. some doubt on the widely held view that Inspecting the causal model in Figure 3, one cognitive skill is becoming an increasingly can see that it involves severing the genetic important determinant of economic success. link (r9) and then calculating the implied We investigated the sensitivity of the hypothetical correlation between parents' results just reported to a number of possible earnings and offsprings' earnings. The differ- sources of error. First,we tested for effects of ence between this hypothetical calculation the age at which the test was taken and espe- and the observed correlation is the genetic cially whether the respondent had completed contribution via IQ to the intergenerational schooling at the time. Forabout two-thirds of transmission of economic status. the estimates, we were able to determine if To answer this question, we need the the test was taken before or after school was answers to two further questions. First,what completed. For these estimates, there is no role does genetic inheritance of IQ play in the effect of the timing of the test on the mea- covariation of parents' and offsprings' cogni- sures reported earlier. Second, we investigat- tive performance? Second, how important is ed the importance of the type of test used cognitive performance as a direct and indirect and found that studies that used more com- (through educational attainments) determi- prehensive tests generally performed some- nant of earnings? The answer to the first what less well than did those that used more question depends on two factors: the heri- narrowly defined tests (often components of tability of IQ, which is probably about 0.5 but the more comprehensive test). However, the cannot be greater than unity, and the genet- estimated effects were not even marginally ic correlation (also 0.5). The answer to the significant (t-statistics less than unity) except second question depends on three factors: for the estimate of the contribution of the influence of IQ on educational attain- noncognitive traits to the returns to school- ment; the influence of educational attaining. Here, the more comprehensive tests ment on earnings; and the direct influence of yielded estimates about 10 percent larger IQ on earnings, independently of schooling. than the narrowertests. The causal paths on which this calculation What do these results imply about the role is based appear in Figure 3 as continuous of IQ transmission in status transmission? A arrows, and the others as dashed arrows. We way to formulate this question precisely is to used representative estimates from the litera- ask how much lower would the intergenera- ture (most of them summarized in Bowles et

Figure 3. A Causal Model of Intergenerational Earnings Transmission. Note: The causal paths generate the intergenerational status correlation rypr Solid lines indicate the causal paths used to calculate the genetic contribution (via IQ) to the similarity of incomes across generations.

Parental Parental IQ Income pyc r CP yp

X '" ssP" Parental gp \ I Schooling ,' gP rS I Parental Geno- Inco me typic IQ h,- g c P V Oenotypic IQ av f

~~IQ IQ~ySchoong Schooling Schoolingo i in Capitalist America Revisited 7 al. forthcoming; see Bowles and Gintis forth- (1998), in collaboration with the Department coming b for details of the calculation). We of Education, which asked, "When you con- conclude that the estimate of the normalized sider hiring a new nonsupervisory or produc- effect on earnings of the child's IQ (both tion worker, how important are the following directly and indirectly via schooling) is sub- in your decision to hire?" Employers ranked stantial: 0.266. We take this to be the relevant "industry based skill credentials" at 3.2 on a value for the parents' generation as well. We scale of 1 (unimportant) to 5 (very impor- estimate the genetic contribution to the cor- tant), with "years of schooling" at 2.9, "score relation of parental and offspring's incomes as on tests given by employer" and "academic a maximum of 0.035, assuming that IQ is per- performance" both at 2.5. By far, the most fectly heritable, or 0.018 making the more important was "attitude" (ranked 4.6), fol- widely accepted assumption that about half lowed by "communication skills" (ranked the variation in IQ is due to genetic inheri- 4.2). tance. The second example is from the far-more- If the genetic inheritance of IQ were the detailed Employers' Manpower and Skills only mechanism accounting for the intergen- Practices Survey of 1,693 British employers erational income correlations, then Figure 1 reported in Green, Machin, and Wilkenson would represent a set of poorly laid bricks on (1998). Of the somewhat more than a third a barely tilted surface, rather than the moun- of the establishments that reported a "skill tainous terrain it actually resembles. The like- shortage," personnel managers identified the lihood that a child of the richest decile would recruitment problem as the "lackof technical attain the top income decile would exceed skills" in 43 percent of the cases, but "poor that of the poor by 12 percent, assuming IQ attitude, motivation, or personality" in a to be 50 heritable, rather than by the 16-44 remarkable62 percent of the cases. Poor atti- times observed in Figure 1. tude was by far the most important reason given for the recruitment difficulty. The importance of motivation relative to technical skill was even the full HOW SCHOOLINGAFFECTS greater among sample. The third example is from a series of stud- LABOR MARKETSUCCESS ies (Cameron and Heckman 1993; Heckman forthcoming; Heckman, Hsee, and Rubinstein Individuals a vector of possess personal capa- 1999) on the labor market impact of the GED and sell these on bilities, c, capabilities the (general equivalency diploma), a diploma labor market at with hourly prices p, hourly gained by a test of cognitive skills taken by a w = The earnings pc. common assumption is large fraction of dropouts from U.S. high that c consists of skills cognitive that depend schools. GED holders exhibit substantially on an individual's innate ability and level of better cognitive performance than other high schooling. We argued in Schooling in school dropouts. But behavioral and person- CapitalistAmerica that cognitive skillsare only ality problems, evidenced by delinquent and a part of what is in c and that schooling does illegal behaviors, account for the fact that the more than enhance cognitive skills. wages of GED holders are barely higher than Until recently, this message has been wide- those of other, less cognitively skilled ly ignored. The availabilityof data on cogni- dropouts and are perhaps 10 percent below tive performance scores on dozens of test the levels that would be predicted on the instruments appears to have crowded out basis of their cognitive skills and other con- other reasonable hypotheses concerning less ventional determinants of earnings. Heckman copiously measured individual attributes. The and his coauthors reasoned that the GED is a following are three examples of the impor- "mixed signal," indicating to employers that tance of noncognitive traits that are impor- the individual had the cognitive skill to com- tant for success in the labor market. The first plete high school but lacked the motivational is from a recent survey of 3,000 employers or behavioral requisites. Their data are also conducted by the U.S Bureau of the Census consistent with the view that the economic 8 Bowles and Gintis8 returns to schooling depend on "seat time"; that are homogeneous with respect to level of that is, being there may be more important cognitive skill(Gintis 1971). A positive answer than learning the new curriculum. in a well-specified model suggests that Sociological accounts frequently stress the schools contribute to earnings by means non-skill-related determinants of earnings other than their contribution to cognitive and of the contribution of schooling to the skill. economy, often under the heading "socializa- An approximation of this test is available. tion for work" (Dreeben 1967; Parsons 1959). Suppose that the income-generating struc- Until recently, economists have ignored this ture for a given demographic group is literature,arguing that an employer would be no more willing to pay a premium for the ser- y = EsS + Rbb + I,c + 6, (3) vices of a "well-socialized" worker than a shopper would be to pay a higher price for where y, s, b, and c measure earnings, school- the fruit of a "well-socialized" grocer. ing, parental socioeconomic background, However, this reason for ignoring noncogni- and cognitive skill level, and ? measures sto- tive traits is inconsistent with modern labor chastic influences on earnings uncorrelated economics, which recognizes that the with the other explanatory variables. Many employment relationship is generally contrac- estimates lack measures of cognitive skill and tually incomplete and hence that an employ- hence estimate ee's effort (and hence the delivery of productive services to the employer) depends on y = R'ss + R'bb + E', (4) how the employee responds to the various types and levels of incentives the firm pre- with c' representing the stochastic influences sents to the employee. stated earlier, plus the influences of cognitive Several examples of this dependence come skilloperating independently of demographic to mind. First,a reduction in the employee's grouping, socioeconomic background, and rate of time preference-that is, a greater ori- schooling. We can compare two estimated entation toward the future-raises the impor- regression coefficients for a years-of-school- one in an like tance to the employee of retaining the job in ing variable, equation Equation in which a measure of skill also the future and thus of avoiding any behavior 3, cognitive and another like in that may result in termination. Second, indi- appears (Rs) Equation 4, which the measure is absent viduals differ greatly in the strength of their cognitive (1's). The ratio of the first to the which we sense of personal efficacy, a trait frequently second, write as measured (inversely) by the so-called Rotter scale. Highly fatalistic, low-efficacy persons a=s(5) believe that their actions have little impact on RIs (5) the outcomes they experience, so that by comparison with those with a greater sense of is an estimate of the contribution of traits personal efficacy, more fatalistic people other than those measured on the cognitive believe that their work effort has less effect on tests to the estimated return to schooling. We the probabilityof their job termination. Thus, call this the "noncognitive component of the the threat of dismissal and the promise of returns to schooling." reward have little incentive effect on those If schooling affected earnings solely with high Rotter scores, and they will make through its contribution to cognitive capaci- poor employees. The third example is how ties (assuming these capacities to be ade- helpful or disruptive an employee is in inter- quately measured by the test scores used), a acting with other employees. would be zero because the regression coeffi- The most direct test of the proposition that cient of years of schooling would fall to zero the contribution of schooling to the develop- once the cognitive level of the individual is ment of cognitive skillsaccounts for the effect accounted for, there being (by hypothesis) no of schooling on earnings is to ask if earnings contribution of schooling to earnings beyond covary with years of schooling in populations its effect on cognitive functioning. By con- America Revisited 9 Schoolingi in Capitalist trast, if the contribution of schooling to cog- The mean value of a in our studies is 0.82, nitive skill explained none of schooling's con- meaning that introducing a measure of cog- tribution to earnings, a would be unity. The nitive performance into an equation using estimates involved are, of course, subject to educational attainment to predict earnings biases, and we address this question at some reduces the coefficient of years of education length in Appendix 2 (available from the by an average of 18 percent. The median for authors). The most obvious potential prob- a was 0.84, and the range was 0.48 to 1.13. lem-that the cognitive score may be mea- This finding suggests that a substantial por- sured with considerably more error than the tion of the returnsto schooling are generated schooling variable and hence that a is biased by effects or correlates of schooling that are upward-is almost certainly not the case. substantially unrelated to the cognitive For the United States from the late 1950s capacities measured on the available tests.2 to the early 1990s, we were able to locate 25 In Figure 4 we present these data, along studies, allowing 58 estimates of the relation- with the years to which the earnings data per- ship between R, and R'sand thus an estimate tain. In a regression using categorical vari- of a. The data sources underlying this and the ables to take account of the demographic other figures in this article are described in groups studied, there is no statisticallysignifi- Appendix 1 (available from the authors). cant time trend in the noncognitive compo- Methods of estimation differ, of course, and nent of the return to schooling. This evidence the demographic groups covered and the gives no support to the commonly held view years for which the data apply vary consider- that the role of measured cognitive traits in ably. We surveyed these studies and selected the contribution of schooling to earnings has what we considered the best-specified esti- increased over the past three decades. mates in each study. For example, we favored These data suggest that a major portion of estimates using measurement error correc- the effect of schooling on earnings operates tion and instrumental variables estimation or in ways independent of the contribution of other techniques to take account of endo- schooling to measured cognitive functioning. geneity of the explanatory variables. We have Correspondingly, the contribution of cogni- included all studies available to us.1 tive functioning to earnings is substantially

Figure 4. The Noncognitive Component Fraction, a, of the Private Return to Schooling over Time: A Summary of 58 Estimates from 25 Studies

I I I~~~~~ % (a)120 I I I 0 110 0

I I 0 100 * 0 0 90 - * 0 * 9 I 80 I S.0* 70 -

0 60 - 0

40 I I I I I I 1960 1965 1970 1975 1980 1985 1990 1995 year 10 Bowles and Gintis independent of schooling. This being the ous objective and subjective indicators of job case, it may be thought that cognitive scores performance (Barrick and Mount 1991; may explain a substantialfraction of the resid- Ghiselli and Barthol 1953; Schmidt and ual variance in the standard earnings equa- Hunter 1998; Tett, Jackson, and Rothstein tion, that is, one including years of schooling 1991) suggested that some dimensions of but not cognitive scores. But this is not the personality, particularly those captured on case. We located 57 of these estimates in 24 what are termed integrity tests, and one of studies. The estimated values of AR2(using, the "big five" personality traits, "conscien- in most cases, "corrected" R2) and the years tiousness," are strong predictors of success in to which the earnings data pertain appear in occupations. In the most recent meta-analysis Figure 5. The mean value of AR2 is 0.0104, (Schmidt and Hunter 1998), these two traits the median is 0.007, and the range is -0.015 were found to be uncorrelated with general to 0.04. Regressing the estimates of AR2on cognitive performance, with average normal- the years to which they pertain, we find no ized regression coefficients of 0.41 and 0.31 time trend in its value (see Bowles et al., predicting job performance. The many individual studies we consulted forthcoming, for details). yielded highly variable results, however. A large number of studies have indicated the importance of and other WHICH TRAITS ARE REWARDED personality noncognitive traits as determinants of earn- IN THE LABOR MARKET? ings (Andrisanni and Nestel 1976; G. J. Duncan and Dunifon 1998a; Filer 1981; If the role of cognitive performance in the Goldsmith, Veum, and Darity 1997; Jencks determination of earnings is modest, what 1979; Murnane et al. 1997; Osborne 2000; individual traits may account for the large Rosenbaum, DeLuca, and Miller 2000). unexplained variance of earnings among Jencks's (1979) survey of research made it demographically similar individuals with the quite clear that personal traits-industrious- same years of schooling? Four meta-analyses ness, perseverance, leadership, and others- of personality measures as predictors of vari- self-assessed and reported by others, as well

Figure 5. Contribution of Cognitive Differences to Residual Inequality by Year: Estimates from 57 Estimates in 24 Studies. AR2is the change in adjusted R2when a cognitive test is added to the regression.

I I I I I ' Ak2 0.04 0 I ' 0 ; i 0 0.03

0.02 I 0

0 0 ? 0 0.01 I** *

I t I i 0 I

-0.01

-0.02 I II I I I 19)60 1965 1970 1975 1980 1985 1990 1995

XY:lar SchoolingSonnaatm in Capitalist Americaaesd Revisited 11 as study habits and other behavioral patterns schooling coefficient occasioned by the addi- in school, influence subsequent occupational tion of a cognitive score (see Figure 4). status attainment and earnings independent- In addition, Osborne (2000) using the ly of parental socioeconomic background, (U.S.) National Longitudinal Survey of Young scores on cognitive tests, and years of school- Women (NLSYW) and the (U.K.) National ing. For example, in an equation predicting Child Development Study (NCDS), found hourly earnings in a large representative sam- that behavioral traits have a significant influ- ple, the normalized regression coefficient on ence on the earnings of women, controlling a composite measure of noncognitive traits is for standard human capital variables. The four times the size of the analogous coeffi- Rotter locus of control was the only personal- cient for a test score, twice that of family ity variable considered from the NLSYW.It background, and 50 percent larger than that was designed to measure the externality of an for years of schooling (reported in Jencks individual, or the degree to which the indi- 1979, Table 5.8, Equation 5). vidual believes that outcomes are the result of G. J. Duncan and Dunifon's (1998a, luck or fate, rather than hard work. The 1998b) study, using the Panel Study of NLSYWcollects measures of externality by Income Dynamics, suggested that the effects using the 11-item abbreviated Rotter scale, of incentive-enhancing preferences are and Osborne used measures of personal con- robust. These researchersstudied men whose trol, evaluated from 4 of these 11 questions. motivational and behavioral traits had been From the NCDS, Osborne extracted two measured 15 to 25 years prior to the obser- orthogonal personality variables using princi- vations of their current earnings. Among the pal components from a 146-item and 12-syn- motivational traits measured were preference drome inventory of social adjustment evaluat- for challenge over affiliation, fear of failure, ed when the respondents were aged 11. The sense of personal efficacy, and degree of trust. inventory was evaluated during school by an The behavioral measures included church outside investigator. attendance, participation in social clubs, tele- Osborne's study also addressed the two vision viewing, newspaper reading, and (as econometric issues most troubling in this lit- discussed in the Introduction)an interviewer's erature: measurement error and the endo- assessment of the cleanliness of the respon- geneity of personality and outcome variables. dent's home. These variables, along with a Measurement error in each variable was cor- score on a cognitive test, a measure of years rected by augmenting the correlation matrix of schooling completed, and an unusually using reliability estimates from paired rich set of other controls, were then used to responses within the data set or external predict the average of the log of hourly wages sources when the data did not allow. These between 1988 and 1992. reliabilityestimates allowed "corrected" cor- In separate regressions estimated by relations to be used in regressions of wages Duncan and Dunifon at our request, the fol- on personality and human capital variables. lowing results were generated. First, the In addition, Osborne used exogenous reduction in the unexplained variance associ- instruments for adult personality,thereby pre- ated with the introduction of the motivation- venting the overestimation of the coefficient al and behavioral variables (to an equation on personality because of the positive covari- including all background controls, schooling, ance between personality and the error term. and the cognitive score) was 0.05, a figure to The first technique uses measures of person- be compared with the average of 0.01 for the ality prior to labor market experience as an reduction in the unexplained variance associ- exogenous instrument for adult personality, ated with adding a cognitive score (in Figure and the second technique creates an instru- 5). The introduction of the attitudinal and ment for adult personality that is indepen- behavioral variables reduced the estimated dent of wages yet highly correlated with adult coefficient on the years of schooling variable personality measures. In a regression analysis by 37 percent, which may be compared with of the NLSYW,we found that there is a signif- an average of a 18 percent reduction in the icant negative sign on the Rotter score, indi- 12 Bowles and Gintis cating that the belief that outcomes are the determinants of earnings is in its infancy, result of fate or luck has a negative influence there is some evidence that motivational and on earnings, with a 1 standard deviation behavioral traits are predictors of higher pay. increase in the Rotterscore associated with an It is impossible to know, of course, whether almost 7 percent decrease in wages. The these traits are simply proxies for (or perhaps coefficient is statistically significant, and the contributors to the acquisition of) unmea- results are similarto that found by Andrisanni sured skills or are valued as such by employ- (1978) and G. J. Duncan and Dunifon ers. (1998a). Using the NCDS, we found that the estimated coefficients on personalityvariables are statistically significant and suggest that a CULTURALEVOLUTION AND THE 1 deviation in is percent change aggression CORRESPONDENCEPRINCIPLE associated with an almost 8 percent decrease in wages, and a 1 standard deviation increase in withdrawal is associated with more than a The correspondence principle, which consti- tuted the of our of the 3 percent decrease in wages. In addition, the centerpiece analysis schools future increase in the total explained variance of way produce workers, may seem to be based on the notion that schools wages from including personality (0.014) is socialize students to larger than the mean increase in the accept beliefs, values, and forms of behavior on the basis of author- explained variance from including cognitive rather than the students' own critical scores in wage determination models report- ity, of their interests. Socializationthe- ed earlier (for details, see Bowles et al. forth- judgments has been criticized for coming; Osborne 2000). ory, however, broadly Osborne (2000) also found evidence of sex two reasons. First, it treats the process of and new behaviors as a and occupational status differences in the adopting rejecting black it does not how individuals returns to personality. The results indicate box; explain learn what. variants of social- that in high-status occupations, women face Second, many ization to the individual significantly larger penalties than do men for theory appear place in an a being aggressive, while men are more heavi- entirely passive role, mere receptacle the of than ly penalized for being withdrawn. Indeed, a 1 of content socialization, rather an standard deviation increase in aggression is active participant in the process. For this rea- associated with a decrease in women's earn- son, socialization theory appears to be incom- ings by more than 7 percent, while the same patible with widely accepted notions of change is associated with an average increase human agency that stress our rationality, in men's earnings by almost 25 percent. intelligence, and capacity to make choices Similarly,a 1 standard deviation increase in that are informed by knowledge of the con- withdrawal is associated with a decrease in sequences of such choices for achieving men's wages by 17 percent and 15 percent goals. In particular, if socialization theory for high- and low-status occupations, respec- were correct, social movements that question tively. For women, these same changes in dominant institutions (for instance, the withdrawal are associated with a 6 percent women's, antiwar, and civil rights movements increase in wages for high-status women and that were strong when Schooling in Capitalist a 6 percent decrease in wages for women in Americawas written) could not occur at all. low-status occupations. Of course, just as We were certainly aware of this critique when Osborne found that specific personality traits we wrote the book and, indeed, Gintis (1975) contribute to earnings in different ways, made exactly this point in an interchange depending on the job and the sex of the indi- with the sociologist Talcott Parsons. We have vidual, it may be that the traits found to be since devoted considerable research effort important in her study using British data toward developing an adequate theory of cul- would not have the same explanatory power ture and cultural change, and we sketch here in the United States or some other country. how the correspondence principle may be Thus, while the study of nonskill traits as fleshed out without assuming the "oversocial- Schoolingi in Capitalist America Revisited 13 ized" conception of the individual that is (assuming that the parents have identical inherent in socialization theory (Wrong, traits). They are subsequently paired with a 1961). cultural model (a teacher, that is) who may Our reformulation embodies two basic have the same or a different array of cultural principles. First, schools influence which cul- traits. Confining attention to a single trait, tural models children are exposed to. Second, suppose the teacher has the same trait as the schools immerse children in a structure of parents. Then the youth is assumed to retain rewards and sanctions. Concerning the first, the trait. But if the parents and the teacher we note that a huge body of evidence attests have different traits, the youth considers to the fact that a society's values are passed which one to adopt, surveying the experi- from generation to generation through a ences of those he knows (his classmates) for process of transmission that may be vertical guidance in making the switch. Among the (from parents) or oblique (from others in the experiences the youth may find salient are the prior generation) and involves a psychologi- rewards and punishments associated with the cal internalization of values (Boyd and particularstructure of schooling. The reward Richerson 1985; Cavalli-Sforzaand Feldman structure underlying the workings of the cor- 1981, 1982; Chen, Cavalli-Sforza, and respondence principle includes the close Feldman 1982; Grusec and Kuczynski1997). association, documented in Schooling in The school system is an unusual form of Capitalist America, between the personality oblique transmission whereby a particular and behavioral traits associated with getting group of people who are often unrepresenta- good grades in school and the traits associat- tive of the population of parents (teachers) ed with garnering high supervisor rankingsat occupy privileged positions as behavioral work. models for children (Marx 1852/1963:125). In this view, culture thus evolves by some Concerning the second principle-the individuals (those paired with an unlike rewards and sanctions involved in the social- model) shifting from what they take to be ization process-we model individuals as at lower- to higher-payoff cultural forms. The times treating culture more instrumentally- formal analysis of this process is presented in as a set of social practices that may be adopt- Bowles (2001) and Gintis (2001 a, 2001 b) on ed, abandoned, and transformed in organiz- the basis of the technique of evolutionary ing social interactions (Bowles and Gintis modeling called replicator dynamics. In this 1986; Gintis 1980). The rewards and sanc- model, it is possible for a school system or any tions associated with particular behaviors in other system of socialization to promote the the school setting are part of this process. spread of a culturaltrait that would otherwise Gellner (1985) noted the central role of spe- not proliferate, suggesting that schools do cialized personnel as the key feature of mod- more than simply reproduce the reward ern systems of culturaltransmission (which he structure of the rest of the society. Schooling termed exo-socialization because of the may thus promote prosocial traits even if important part played by outsiders, rather these traits are not individuallyadvantageous. than parents and neighbors, in the process). By like reasoning, schooling can also promote Marx, in a passage we quoted in Schoolingin traits that are advantageous to one group Capitalist America, depicted the process of (the group determining the structure of cultural modernization as a conflict between schooling) even if they are not generally two competing forms of oblique transmis- advantageous. sion: "the modern and the traditional con- To see the validity of this assertion, consid- sciousness of the [early 19th-century] French er a group whose members can adopt either peasant contended for mastery [in] . . . the cultural trait A or cultural trait B. Trait B is form of an incessant struggle between the superior in the sense that B types have payoff schoolmasters and the priests." 1, compared with trait A, whose users have A simple model of this process is the fol- payoff 1-s, where 0 < s < 1. We assume that lowing. Children initiallyacquire culturaltraits during childhood, A types and B types (those by vertical transmission from their parents who have provisionally received these traits 14 Bowles and Gintis via vertical transmission from their parents) These propositions show the importance are paired with a cultural model (teacher) of such oblique cultural institutions as who may be of either type. As we mentioned schools, which are necessary to stabilize cul- earlier, those paired with a like type retain tural forms, such as the legitimacy of being their type. Those who are paired with an subservient in the workplace, that benefit one unlike type then may switch their type, and group, in this case, the employers, at the the likelihood of their doing so increases with expense of another, the employees. In light of the difference in net rewards that the individ- this result, our analysis of the capital-labor ual observes. conflicts of the content and form of schooling Oblique transmission, as we have noted, is is understandable without recourse to the structured in a particularway in a modern theory of socialization as presented in stan- school system: Teachersare the major cultural dard sociology. models, more than neighbor elders, religious figures, and the like, and the rewards and penalties that drive the updating process are CONCLUSION structured by such things as the correspondence between the traits associat- personality The main scientific findings of Schooling in ed with and good grades employers' Capitalist America have remained plausible, We now believe that we have approval. may and their validity has even been strengthened the rewards associated with overemphasized over the past quarter century. We believe that future work rather than future roles as roles, the correspondence principle is also, by and citizens, family members, and the like, but this large, correct. bias does not bear on the of the logic model. Over the years Schooling in Capitalist Schooling can affect the direction of cultural Americahas received a considerable amount in if evolution two ways. First, most teachers of criticalattention, for which we are grateful. are As, then the children of A parents will One reading of our book-that it presented a rarelyswitch, while B children will virtuallyall functionalist argument-is sufficiently mis- have the occasion (a mismatch) to consider a guided to deserve a brief comment here. A switch. Second, if the reward structure of the functionalist argument explains something, school favors those with A traits (even if the Bs such as the structure of schooling, by the may do better in adult life), then a significant benefits it confers on some group, for number of B children will become As. instance, the profits accruing to employers For full developments of this and related from a well-socialized labor force, without models, see Bowles (2001) and Gintis (2001 a, providing any causal explanation of the man- 2001b). Here, we present only a few major ner in which these consequences account for implications. First, in the absence of the the thing to be explained. We devoted three oblique transmission of the disadvantaged chapters of Schooling in CapitalistAmerica to cultural form, the advantaged cultural form the history and evolution of education pre- always drives out the disadvantaged form. cisely to illuminate the process by which the Second, when oblique transmission of the correspondence principle and other aspects disadvantaged trait is present, a positive fre- of the structure of schooling came about. The quency of this trait can persist even when benefits (correctly) anticipated by employers some fraction of agents are switching to the loom large in this account. But this does not advantaged form to increase their payoffs. make the argument functionalist. We suspect Depending on the specific assumptions of the that some readers were surprised that overt model and the specific value of the parame- class conflict over the content and structure ters, there can either be two stable "homoge- of schooling played such a minor role in our neous" cultural equilibria involving high fre- account of the history of U.S. education, but quencies of either the advantaged or disad- we did not then, nor do we now, believe that vantaged trait or a single, stable "heteroge- the historical record supports this more tradi- neous" equilibrium involving a moderate fre- tional Marxianinterpretation. Our dissatisfac- quency of both culturalforms. tion with Schoolingin CapitalistAmerica in this in Revisited Schooling-- Capitalist-I America 15 respect is not that we downplayed class con- stormy days when Schooling in Capitalist flict or that we failed to provide a causal Americawas written, schools express the con- mechanism, but that we may have misunder- flicts and limitations, as well as the hopes, of stood the causal mechanism. Our interpreta- a heterogeneous and unequal society. tion gave insufficient attention to the contra- Schools continue to be both testing grounds dictory pressures operating on schools, par- and battlegrounds for building a society that ticularly those that emanate from the labor extends its freedoms and material benefits to market, which we stressed, and from the all. democratic polity, which we should have emphasized more. We present a more adequate view in Bowles and Gintis (1981, NOTES 1986). The main of in shortcomings Schooling 1. We located five additional studies, America reflect the times in which Capitalist an additional six estimates, in which we wrote. The 1960s economic boom allowing long the variable is a measure of occu- and the antimaterialist countercultural cur- dependent status rather than rents that it fostered led us to under- pational earnings: Bajema perhaps (1968), Conlisk 0. D. Duncan the value of in con- (1971), (1968), emphasize schooling Porter and Sewell, and to The (1974), Haller, tributing productive employment. Ohlendorf (1970). The mean value of a in more we think, is important shortcoming, these studies is 0.89, and the lowest is 0.81. We for the most programmatic. avoided, These results are not in 4. the of what schools should reported Figure part, question be, 2. These data concern the United instead on what schools are only focusing actually States, and we do not draw inference and do. We also to devote much any neglected from them about the returns to in attention to how economic other schooling systems other economies. We and there is than better facilitate the suspect, capitalism may some evidence et al. achievement of the of (Alderman 1996; enlightened objectives and Sabot Glewwe We took it as obvious that a Boissiere, Knight, 1985; schooling. system and Leboucher that of 1996; Lavy,Spratt, 1997) democratic, employee-owned enterprises, in societies where is more limited in coordinated both markets and schooling by govern- its the in the mental was both and eco- scope, cognitive component policies, politically returns to be viable as an alternative to schooling may considerably nomically capital- than in the United States. ism. We remain convinced of the attractive- larger However, to Moll in a of ness of such a but are less according (1998), sample system, sanguine black workers in South the value of a about its and more convinced that Africa, feasibility for returns to is for reforms of be the most primary schooling 0.73; capitalism may likely and for to the that we secondary schooling, 0.67; higher way pursue objectives 0.92. These values are well within embraced at the outset. the book education, Although the of estimates in 4. endorses the idea that radicals-even revolu- range presented Figure tionaries-must also be reformers,we provid- ed little guidance to policy makers, teachers, or students who are seeking practical positive REFERENCES steps to bring about long-term improvements in educational structure and practice. Alderman,Harold, Jere R. Behrman,David Ross, and RichardSabot. 1996. "The Returnsto Partlybecause we are now reasonably cer- Human in Pakistan'sRural tain that we had the facts right, we remain Endogenous Capital committed to our overall to school- Wage Labor Market." Oxford Bulletin of approach Economicsand Statistics the of education in 58(1):29-55. ing-embedding analysis Andrisanni, Paul J. 1978. "Internal-External the structure of evolving the economy and Attitudes, Personal Initiativeand LaborMarket the polity and giving attention to the Experience."Pp. 101-34 in WorkAttitudes and noncognitive as well as cognitive effects of Labor Market Experience,edited by Paul J. education. Today, no less than during the Andrisanni.New York:Praeger. 16 Bowlesand Gintis

Andrisanni, Paul J., and Gilbert Nestel. 1976. -forthcoming b. "The Inheritance of "Internal-ExternalControl as Contributor to Inequality."Journal of EconomicPerspectives. and Outcome of Work Experience."lournal of Bowles,Samuel, Herbert Gintis, and MelissaOsborne. AppliedPsychology 61:156-65. 2001. "Incentive-Enhancing Preferences: Atkinson, A. B., A. K. Maynard, and C. G. Trinder. Personality,Behavior and Earnings."American 1983. Parents and Children:Incomes in Two EconomicReview 91:155-58. Generations.London: Heinemann. -. Forthcoming. "The Determinants of Individual Bajema, C. J. 1968. "A Note on the Interrelations Earnings:Skills, Preferences, and Schooling." Among Intellectual Ability, Education journalof EconomicLiterature. Attainment, and Occupational Achievement: Boyd, Robert, and Peter J. Richerson. 1985. Culture A Follow-Up Study of a Male Kalamazoo and the Evolutionary Process. Chicago: Public School Population." Sociology of Universityof Chicago Press. Education41:31 7-19. Cameron, Steven, and James Heckman. 1993. Barrick,Murray R., and Michael Mount. 1991. "The "The Non-equivalence of the High School Big Five PersonalityDimensions and Job Per- Equivalents. Journal of Labor Economics formance: A Meta-Analysis."Personnel Psych- 11:1-47. ology 44:1-26. Cavalli-Sforza,Luca L., and Marcus W. Feldman. Becker,Gary S. 1988. "FamilyEconomics and Macro 1982. "Theory and Observation in Cultural Behavior."American Economic Review 78:1-13. Transmission."Science 218:19-27. Becker, Gary S. and Nigel Tomes. 1986. "Human Cavalli-Sforza,Luigi L., and Marcus W. Feldman. Capital and the Rise and Fall of Families." 1981. Cultural Transmission and Evolution. Journalof LaborEconomics 4(3):S1-S39. Princeton, NJ:Princeton UniversityPress. Behrman, Jere R., Robert A. Pollak, and Paul Chen, Kuang-Ho, Luigi L. Cavalli-Sforza, and Taubman. 1995. From Parent to Child: Marcus W. Feldman. 1982. "A Study of IntrahouseholdAllocations and Intergenerational Cultural Transmission in Taiwan." Human Relations in the U.S. Chicago: University of Ecology10:365-82. Chicago Press. Conlisk, John. 1971. "A Bit of Evidence on the Blau, Peter, and Otis Dudley Duncan. 1967. The Income-Education-AbilityInteraction." Journal American Occupational Structure. New York: of HumanResources 6:358-62. John Wiley & Sons. Cooper, Suzanne, Steven Durlauf, and Paul Boissiere, M., J. B. Knight, and R. H. Sabot. 1985. Johnson. 1994. "On the Evolution of "Earnings, Schooling, Ability, and Cognitive Economic Status Across Generations." Skills."American Economic Review:1016-30. AmericanEconomic Review 84:50-58. Bouchard,T. J., Jr.,and M. McGue. 1981. "Familial Corak, Miles, and Andrew Heisz. 1999. "The Studies of Intelligence." Science212:1055-59. Intergenerational Earnings and Income Bowles, Samuel. 2001."Individual Interactions, Mobility of Canadian Men: Evidence from the Group Conflicts, and the Evolution of Longitudinal Income Tax Data." Journal of Preferences."Pp. 155-190 in Social Dynamics, HumanResources 34:505-33. edited by Steven N. Durlauf and H. Peyton Dreeben, Robert. 1967. On What Is Learned in Young. Cambridge, MA: MITPress. School. Reading, MA:Addison-Wesley. Bowles, Samuel, and Herbert Gintis. 1976. Duncan, Greg J., and Rachel Dunifon. 1998a. Schooling in Capitalist America: Educational "Long-Run Effects of Motivation on Labor- Reformand the Contradictionsof EconomicLife. Market Success." Social Psychology Quarterly New York:Basic Books. 61:33-48. - 1981. "Contradiction and Reproduction in --.1998b. "'Soft-Skills' and Long-Run Market Educational Theory." In Schooling, Ideology, Success." Research in Labor Economics and Curriculum,edited by Len Barton. Sussex, 17:123-50. England: FalmerPress. Duncan, Otis Dudley. 1968. "Ability and - 1986. Democracy and Capitalism: Property, Achievement." EugenicsQuarterly 15:1-11. Community,and the Contradictionsof Modern Filer, Randall Keith. 1981. "The Influence of Social Thought.New York:Basic Books. Affective Human Capital on the Wage -. forthcoming a."The Inheritance of Economic Equation." Research in Labor Economics Status: Education, Class and Genetics." In 4:367-416. International Encyclopediaof the Social and Gellner, Ernest. 1985. Nations and Nationalism. Behavioral Sciences: Genetics, Behavior and Ithaca, NY:Cornell UniversityPress. Society, edited by Marcus Feldman and Paul Ghiselli, Edwin E., and Richard P. Barthol. 1953. Baltes.New York:Oxford UniversityPress. "The Validityof Personality Inventories in the Schoolingi i in Capitalist America Revisited 17

Selection of Employees." Journal of Applied Mulligan, Casey. 1997. Parental Priorities and Psychology37:18-20. Economic Inequality. Chicago: University of Gintis, Herbert. 1971. "Education,Technology, and Chicago Press. the Characteristics of Worker Productivity." - 1999. "Galton vs. the Human Capital AmericanEconomic Review 61:266-79. Approach to Inheritance." lournal of Political -. 1975. "Welfare Economics and Individual Economy107:S184-S224 (Part 2). Development: A Reply to [T]alcott [P]arsons." Murnane, Richard, John Willett, Jay Braatz, and QuarterlyJournal of Economics89:291-302. Yves Duhaldeborde. 1997. "Does the Self- -. 1980. "Theory, Practice, and the Tools of Esteem of High School Aged Males Predict Communicative Discourse." Socialist Review Labor Market Success a Decade Later? 50:189-232. Evidencefrom the NLSY."Unpublished manu- -. 2001a. "The Hitchhiker'sGuide to Altruism: script, Graduate School of Education, Harvard Genes and Culture, and the Internalizationof University,Cambridge, MA. Norms" (Working paper). Santa Fe, NM: Osborne, Melissa. 2000. "The Power of Personality: Santa Fe Institute. LaborMarket Rewards and the Transmissionof -. 2001 b. "The Puzzle of Human Prosociality." Earnings." Unpublished manuscript, Available on-line at http://wwwunix.oit.umass.- Department of Economics, University of edu/-gintis Massachusetts, Amherst. Glewwe, Paul. 1996. "The Relevance of Standard Parsons, Talcott. 1959. "The School Class as a Estimates of Rates of Return to Schooling for Social System." Harvard Educational Review Education Policy: A Critical Assessment." 29:297-318. Journalof DevelopmentEconomics 51:267-90. Porter, James N. 1974. "Race, Socialization, and Goldsmith, Arthur, Jonathan Veum, and William Mobility in Education and EarlyOccupational Darity. 1997. "The Impact of Psychological Attainment." American Sociological Review and Human Capital on Wages." Economic 39:303-16. Inquiry35:815-29. Rosenbaum, James E., Stephanie DeLuca, and Green, Francis, Stephen Machin, and David Shazia Miller. 2000. "Are Noncognitive Wilkenson. 1998. "The Meaning and Behaviors in School Related to Later Life Determinants of Skill Shortages." Oxford Outcomes?" Unpublished manuscript, Bulletinof Economicsand Statistics60:165-88. Northwestern University,Evanston, IL. Grusec, Joan E., and Leon Kuczynski. 1997. Schmidt, FrankL., and John E. Hunter. 1998. "The Parenting and Children's Internalization of Validity and Utility of Selection Methods in Values:A Handbook of ContemporaryTheory. Personnel Psychology: Practical and New York:John Wiley & Sons. Theoretical Implications of 85 Years of Heckman, James. Forthcoming. The GED. Research Findings." Psychological Bulletin Heckman,James, Jingjing Hsee, and Yona Rubinstein. 124:262-74. 1999. "The GED as a Mixed Signal." Sewell, WilliamH., ArchibaldP. Haller,and George Unpublished manuscript, Department of W. Ohlendorf. 1970. "The Educationand Early Economics,University of Chicago. Occupational Status Achievement Process." Hertz, Thomas. 2001a. "Essayson Education and AmericanSociological Review 35:1014-27. Poverty in South Africa." Unpublished Ph.D. Solon, Gary R. 1992. "Intergenerational Income dissertation, Universityof Massachusetts. Mobility in the United States." American -. 2001 b. "Intergenerational Transition EconomicReview 82:393-408. Probabilitiesin the PSID."Princeton, NJ: Center Tett, Robert P., Douglas N. Jackson, and Mitchell for Healthand Well-being,Princeton University. Rothstein. 1991. "Personality Measures as Jencks, Christopher. 1979. Who Gets Ahead? New Predictorsof Job Performance:A Meta-Analytic York:Basic Books. Review."Personnel Psychology 44:703-42. Lavy, Victor, Jennifer Spratt, and Nathalie U.S. Bureau of the Census. 1998. "First Findings Leboucher. 1997. "Patterns of Incidence and from the EQW National Employer Survey" Change in Moroccan Literacy."Comparative (EQW Catalog Number RE01). Washington, EducationReview 41:120-41. DC: Author. Marx, Karl.1963. The EighteenthBrumaire of Louis Wrong, Dennis H. 1961. "The Oversocialized Bonaparte.New York:International Publishers. Conception of Man in Modern Sociology." (Original work published 1852) AmericanSociological Review 26:183-93. Moll, Peter. 1998. "PrimarySchooling, Cognitive Zimmerman, David J. 1992. "Regression Toward Skills and Wages in South Africa." Economica Mediocrity in Economic Stature." American 65:263-84. EconomicReview 82:409-29. 18 Bowles and Gintis

Samuel Bowles, Ph.D., is Professorof Economics, Universityof Massachusetts, Amherst, and Directorof the EconomicsProgram, Santa Fe Institute,Santa Fe, New Mexico.

HerbertGintis, Ph.D., is Professorof Economics,University of Massachusetts,Amherst, and on the ExternalFaculty of the Santa Fe Institute.

The authors thank those who commented on their work at recent meetings commemoratingthe 25th anniversaryof its publication under the auspices of the American EducationalResearch Associationin Seattle in March 2001 and the EasternSociological Association in Philadelphiain March2001. Theyalso thank the John D. and CatherineT. MacArthur Foundation for financialsup- port. Thisarticle is dedicated to the memory of MartinKessler, the authors' editor at Basic Books, who encouragedand guided them through the making of Schoolingin CapitalistAmerica. Address all correspondenceto Samuel Bowles (e-mail: [email protected])or Herbert Gintis (e- mail:[email protected]),Department of Economics,University of Massachusetts, Thompson Hall,Amherst, MA 01003.