Content uploaded by Mark A Schnitzler
Author content
All content in this area was uploaded by Mark A Schnitzler on Nov 28, 2017
Content may be subject to copyright.
http://pss.sagepub.com/
Psychological Science
http://pss.sagepub.com/content/9/2/139
The online version of this article can be found at:
DOI: 10.1111/1467-9280.00026
1998 9: 139Psychological Science
Joel Myerson, Mark R. Rank, Fredric Q. Raines and Mark A. Schnitzler
Race and General Cognitive Ability: The Myth of Diminishing Returns to Education
Published by:
http://www.sagepublications.com
On behalf of:
Association for Psychological Science
can be found at:Psychological ScienceAdditional services and information for
http://pss.sagepub.com/cgi/alertsEmail Alerts:
http://pss.sagepub.com/subscriptionsSubscriptions:
http://www.sagepub.com/journalsReprints.navReprints:
http://www.sagepub.com/journalsPermissions.navPermissions:
What is This?
- Mar 1, 1998Version of Record >>
by guest on February 20, 2013pss.sagepub.comDownloaded from
PSYCHOLOGICAL SCIENCE
VOL. 9, NO. 2, MARCH 1998 Copyright © 1998 American Psychological Society
139
Abstract—
The impact of education on racial differences in general
cognitive ability was assessed using data from the National Longitudi-
nal Survey of Youth. To control for attrition during the educational pro-
cess, we compared the scores of individuals who ultimately attained the
same level of education but who were tested at different points in their
educational careers. Multiple regression analyses revealed that educa-
tion can have a strong positive effect on cognitive ability in both whites
and blacks. Whites benefited more from education than blacks during
the high school years, but blacks benefited much more than whites from
a college education, substantially narrowing the gap between the
races. These findings contradict the hypothesis that racial differences
in intelligence are relatively immutable, in part because of the dimin-
ishing returns from increases in education.
On average, blacks in the United States tend to score significantly
lower than whites on intelligence tests (e.g., Jensen, 1980; Reynolds,
Chastain, Kaufman, & McLean, 1987), even when socioeconomic sta-
tus (SES) is statistically controlled (Herrnstein & Murray, 1994; Loeh-
lin, Lindzey, & Spuhler, 1975). According to Herrnstein and Murray
and other researchers (e.g., Jensen, 1969, 1985), this is because the
gap between blacks’ and whites’ test scores is largely due to genetic
differences. With respect to education and racial differences, Herrn-
stein and Murray argued that the relationship between education and
intelligence is negatively accelerated, and therefore efforts to improve
educational quality or increase the amount of education beyond 12
years will pay increasingly diminishing returns in terms of raising
intelligence or reducing racial differences. However, this claim is
based on assumptions that go well beyond current knowledge, and the
data cited by Herrnstein and Murray are open to quite different inter-
pretations (e.g., Fischer et al., 1996; Hauser, 1995).
Only recently has consensus emerged on the fact that education
influences intelligence (Neisser et al., 1996; for a review, see Ceci,
1991), and it is still unclear whether schooling has equivalent impact
on different subgroups. With respect to secondary and postsecondary
education in particular, analysis is complicated by the fact that amount
of schooling may be positively related to intelligence test scores, not
just because of the effects of education on cognitive ability, but also
because educational level may reflect selective attrition of the less able
(Herrnstein & Murray, 1994; Kaufman, 1990). Interestingly, although
Herrnstein and Murray reported a significant effect of schooling on
intelligence in their own analyses, they did not compare blacks and
whites in this regard, nor did they attempt to control for selective attri-
tion. Thus, their analyses shed no light either on the nature of the effect
of education on intelligence or on whether there are racial differences
in this regard.
From the perspective of educational policy, it is important to deter-
mine whether the relationship between education and intelligence is
nonlinear, as Herrnstein and Murray (1994) hypothesized, and pro-
vides diminishing returns for increases in education, and also whether
this relationship differs for blacks and whites. In conducting a test of
Herrnstein and Murray’s diminishing-returns hypothesis, we used the
same data set as they did; the same approach to the selection of
respondents, construction of variables, and modeling techniques; and
the same test of general cognitive ability, the Armed Forces Qualifica-
tion Test (AFQT), which was administered to an exceptionally large
nonmilitary sample as part of the National Longitudinal Survey of
Youth (NLSY). For the present purposes, the unique advantage of this
data set was that the longitudinal nature of the NLSY provided an
opportunity to minimize the confound of educational attrition.
Because data were available on individuals’ ultimate educational
attainment, it was possible to compare the test scores of individuals
who all attained the same level of education but who were tested at dif-
ferent points in their educational careers.
One analysis was conducted on the data from individuals who ulti-
mately graduated from high school but obtained no further schooling,
and who had completed 8, 9, 10, 11, or 12 years of schooling at the time
they were tested. A second analysis examined the data from individuals
who ultimately graduated from college but received no postgraduate
training, and who had completed between 8 and 16 years of education
at the time they were tested. For each sample, we used multiple regres-
sion techniques to evaluate the form of the relationship between score
on a test of general cognitive ability and educational level when tested.
In addition, these analyses examined whether the linear and nonlinear
components of the relationship between test score and number of
grades completed differed significantly between blacks and whites.
METHOD
The NLSY is a nationally representative longitudinal survey of
12,686 young men and women who were between the ages of 14 and
21 as of January 1, 1979, when the study began. The data were gath-
ered by the National Opinion Research Council under the supervision
of Ohio State University’s Center for Human Resources Research.
Interviews with participants were conducted annually, and the survey
maintained a retention rate of more than 90% through 1990. Informa-
tion from the 1989 round of interviews was used to determine partici-
pants’ ultimate educational attainment. The present analysis was
restricted to data for black and white (non-Hispanic) individuals only.
More than 93% of the NLSY participants took the AFQT in 1980,
the year that it was administered as part of the NLSY. The AFQT con-
sists of select portions of the Armed Services Vocational Aptitude Test
(ASVAT) that serve as the test of general cognitive ability for the
armed services of the United States. According to the Armed Forces
1989 convention, the AFQT score is the sum of the raw scores on the
word knowledge, paragraph comprehension, arithmetic reasoning, and
RACE AND GENERAL COGNITIVE ABILITY:
The Myth of Diminishing Returns to Education
Joel Myerson,
1
Mark R. Rank,
2
Fredric Q. Raines,
3
and Mark A. Schnitzler
4
1
Department of Psychology, Washington University;
2
George Warren Brown School of Social Work, Washington University;
3
Department of Economics, Washington University; and
4
Health Administration Program, School of Medicine, Washington University
Address correspondence to Joel Myerson, Department of Psychology,
Campus Box 1125, Washington University, One Brookings Dr., St. Louis,
MO 63130; e-mail: jmyerson@artsci.wustl.edu.
Research Report
by guest on February 20, 2013pss.sagepub.comDownloaded from
PSYCHOLOGICAL SCIENCE
140
VOL. 9, NO. 2, MARCH 1998
Race and General Cognitive Ability
mathematics knowledge subtests of the ASVAT. Herrnstein and Mur-
ray (1994) reported that the AFQT correlates highly with other stan-
dardized tests of general cognitive ability.
We conducted multiple regression analyses in which the dependent
variable was AFQT score and the independent variables were age at
the time of testing, parental SES, education (number of grades com-
pleted at the time of testing), and education-squared. In these and sub-
sequent analyses, AFQT score and SES were normalized so as to have
a mean of zero and unit variance in the total sample (i.e., they were
converted to standard deviation units). SES was based on the mean of
at least three of four indicators measured in standard deviation units:
mother’s education, father’s education, family income, and parental
occupation (Duncan’s Socioeconomic Index). Age was included to
control for possible maturational differences in AFQT score. Educa-
tion and education-squared terms were both included in the regression
models so that the nonlinear as well as the linear effects of this vari-
able could be assessed. A reduced model in which AFQT score was
regressed on SES, age, education, and education-squared was com-
pared with a full model that also included an indicator variable (set
equal to zero for participants who identified themselves as being white
and set to one for those who identified themselves as black) and both
linear and nonlinear interaction terms. Comparing the full and reduced
models provided a test of separate regressions for blacks and whites.
RESULTS
We first examined the effect of education on the sample of NLSY
participants who had graduated from high school but obtained no further
education as of 1989 and who had completed between 8 and 12 years of
schooling when they were tested in 1980. Comparison of the full and
reduced models indicated that the effect of education on AFQT score dif-
fered significantly for blacks and whites,
F
(2, 1995) = 128.35,
p
< .001,
and therefore only the results for the full model are reported (Table 1).
The effect of education predicted by the regression model is shown in
Figure 1. There was a steeper increase in AFQT score with grade level
for whites than for blacks, and for whites only the increase was nega-
tively accelerated. More specifically, the increase in AFQT score pre-
dicted by the regression model for whites who entered high school at age
14 and graduated 4 years later was approximately 0.5
SD,
but for blacks
the predicted increase was less than half that size. Thus, although high
school education had a generally positive effect on AFQT scores, the
effect was considerably greater for whites than for blacks, widening the
gap between the races until by the completion of high school, the esti-
mated white-black difference had increased to nearly 0.9
SD.
We next examined the effect of education on the sample of NLSY
participants who had graduated from college as of 1989 and who had
completed between 8 and 16 years of schooling when they were tested
in 1980. Again, comparison of the full and reduced models indicated
that with age and SES statistically controlled, the effect of education
on AFQT score differed significantly for blacks and whites,
F
(2, 715)
= 4.19,
p
< .01, and therefore only the results for the full model are
reported (Table 2). Figure 2 shows the predicted effect of education on
the sample of college graduates. Education had quite different effects
on the AFQT scores of black and white students during the high school
Table 1. Multiple regression model for high school graduates
Variable Coefficient t ratio p
Constant –3.175 –3.062 .002
Black –1.375 –0.724 .468
Socioeconomic status –0.171 –8.079 .000
Age –0.034 –1.985 .047
Education –0.537 –2.663 .008
Education-squared –0.019 –2.070 .039
Black × education –0.343 –2.114 .035
Black × education-squared –0.013 –2.087 .038
Note. N = 2,003, 23% black. R2 = .317.
Fig. 1. Scores of future high school graduates on the Armed Forces
Qualification Test as estimated by the multiple regression model. Esti-
mates were generated assuming students were 14 years old when they
entered high school (8 grades completed) and 18 years old when they
graduated (12 grades completed).
Table 2. Multiple regression model for college graduates
Variable Coefficient t ratio p
Constant –0.249 –0.368 .713
Black –2.568 –1.362 .173
Socioeconomic status –0.163 –6.418 .000
Age –0.083 –2.352 .019
Education –0.233 –1.801 .008
Education-squared –0.004 –0.796 .426
Black × education –0.638 –1.899 .058
Black × education-squared –0.028 –1.945 .052
Note. N = 723, 16% black. R2 = .324.
by guest on February 20, 2013pss.sagepub.comDownloaded from
PSYCHOLOGICAL SCIENCE
J. Myerson et al.
VOL. 9, NO. 2, MARCH 1998
141
years. More specifically, the increase in AFQT score predicted by the
regression model for whites who entered high school at age 14 and
graduated 4 years later was approximately 0.3
SD,
but for blacks there
was no increase. Although the scores for black future college gradu-
ates appear to decrease initially, this decrease is not significant, as all
the values for the high school years are well within the prediction
interval for the initial data point.
The difference in the effect of education on black and white stu-
dents reversed dramatically during college. Although the white stu-
dents made gains during the college years, the black students benefited
more than four times as much, increasing their scores more than 0.7
SD
from the time of entering college until the time of graduation 4
years later. The gap between whites and blacks was estimated at more
than 1
SD
when they entered college, but by the time of college gradu-
ation, this difference had shrunk by approximately one half. Note that
the reason for this shrinkage was not because continuing education
beyond high school yielded diminishing returns for white students, but
rather because the gains made by black students accelerated positively.
DISCUSSION
With selective attrition, SES, and age all statistically controlled,
education produced significant increases in the scores of both black
and white students, those who were college bound as well as those
who were not, on a test of general cognitive ability (i.e., the AFQT).
Between the beginning and end of high school, the scores of white
future high school graduates increased more than twice as much as the
scores of black future high school graduates. The scores of white
future college graduates also increased significantly during the high
school years, whereas the scores of black future college graduates
showed no increase during this period. In contrast, it was the black
college students who made the largest gains between the end of high
school and college graduation, with their test scores increasing more
than four times as much as those of white college students.
It is important to consider any possible explanations of these find-
ings in the context of the whole pattern of results. If one were to con-
sider only the high school graduates, for example, a possible
explanation for the racial differences might be that blacks do not bene-
fit as much from high school because they are of lower cognitive abil-
ity. However, consideration of both the high school and the college
graduates included in this study argues strongly against that possibil-
ity. Blacks who went on to graduate from college showed no increase
in their test scores during the high school years, yet these students
were presumably of high ability.
Similarly, if one were to consider only racial differences in the
impact of a college education, a possible explanation might be that
black college students benefit more because those who graduate col-
lege are a select subset of the blacks who enter college, whereas whites
who graduate college are a less select group. It is certainly true that
there is greater attrition among black college students than among
white college students in general, as well as among NLSY participants
(Herrnstein & Murray, 1994). However, this fact only makes the failure
to profit from high school by the highly select group of black future
college graduates all the more remarkable, and raises the possibility
that the increases they showed in college resulted from the removal of
whatever may have been handicapping them during high school.
What might be the source of such a handicap? Our tentative answer
to this question is based on the fact that the quality of secondary educa-
tion differs for blacks and whites, even after controlling for socioeco-
nomic differences (e.g., Card & Krueger, 1992a, 1992b; Coleman, 1990;
Jaynes & Williams, 1989; Yinger, 1995). Much of the difference in sec-
ondary school environments may be traced back to a pattern of de facto
racial residential segregation that results in substantial segregation in
public education (Orfield, 1993; Rivkin, 1994). The pattern of racial res-
idential segregation tends to hold even when SES is taken into account,
so that students from black and white families with the same income
may be exposed to very different public-education experiences (Farley
& Frey, 1994; Massey & Denton, 1993; Massey & Hajnal, 1995).
It is not surprising, then, that as black and white students complete
more grades in high school environments that differ in quality, the gap
in cognitive test scores widens. At the college level, however, where
black and white students are exposed to educational environments of
comparable quality (National Center for Education Statistics, 1995),
many blacks are able to make remarkable gains, closing the gap in test
scores. Others, of course, may enter college with a disadvantage that
cannot be overcome rapidly enough, contributing to the higher dropout
rate among black students.
Herrnstein and Murray (1994) suggested that education in the
United States has reached the point where further increments in
amount and quality will pay diminishing returns, particularly with
respect to racial differences. In contrast, the present findings suggest
that variations in amount and quality of education that are within the
range commonly observed in this country can have a profound impact.
In some cases (e.g., secondary education), these variations may act to
exacerbate racial differences in intelligence test scores, whereas in
other cases (e.g., postsecondary education), such variations may act to
substantially reduce racial differences.
Fig. 2. Scores of future college graduates on the Armed Forces Quali-
fication Test as estimated by the multiple regression model. Estimates
were generated assuming students were 14 years old when they
entered high school (8 grades completed) and 22 years old when they
graduated college (16 grades completed).
by guest on February 20, 2013pss.sagepub.comDownloaded from
PSYCHOLOGICAL SCIENCE
142
VOL. 9, NO. 2, MARCH 1998
Race and General Cognitive Ability
Unfortunately, the present analyses also suggest that standardized
tests of academic ability used in college admissions are taken at the
point when racial differences are most pronounced. In the present data,
the largest difference between white and black students who would go
on to graduate college was estimated to occur in the senior year of
high school, precisely when admissions tests are commonly taken. As
indicated by the much higher scores (and smaller racial differences)
observed by the end of college, the high school test scores of the future
black graduates greatly underestimated their potential.
In conclusion, the present results reveal that the relationship
between grades completed and cognitive test scores is not negatively
accelerated, at least for black students. In fact, for the black students
who would later graduate from college, the relationship between the
number of grades completed and cognitive test scores appeared to be
positively accelerated. Although blacks gained much less than whites
during high school, they improved at a much greater rate than whites
during college, dramatically narrowing the gap in test scores. One can
only speculate what the results might have been if the experiences of
blacks and whites prior to college had not been so different. Neverthe-
less, the present findings indicate that, rather than producing diminish-
ing returns, increasing the amount and the quality of education shows
considerable promise as a means of reducing racial differences.
REFERENCES
Card, D., & Krueger, A.B. (1992a). Does school quality matter? Returns to education and
the characteristics of public schools in the United States.
Journal of Political Econ-
omy, 100,
1–40.
Card, D., & Krueger, A.B. (1992b). School quality and black-white relative earnings: A
direct assessment.
Quarterly Journal of Economics, 107,
151–200.
Ceci, S.J. (1991). How much does schooling influence general intelligence and its cogni-
tive components? A reassessment of the evidence.
Developmental Psychology, 27,
703–722.
Coleman, J.S. (1990).
Equality and achievement in education.
Boulder, CO: Westview Press.
Farley, R., & Frey, W.H. (1994). Changes in the segregation of whites from blacks during
the 1980s: Small steps toward a more integrated society.
American Sociological
Review, 59,
23–45.
Fischer, C.S., Hout, M., Sanchez Jankowski, M., Lucas, S.R., Swidler, A., & Voss, K.
(1996).
Inequality by design: Cracking the bell curve myth.
Princeton, NJ: Princeton
University Press.
Hauser, R.M. (1995). The bell curve.
Contemporary Sociology, 24,
149–153.
Herrnstein, R.J., & Murray, C. (1994).
The bell curve: Intelligence and class structure in
American life.
New York: Free Press.
Jaynes, G.D., & Williams, R.M. (1989).
A common destiny: Black and American society.
Washington, DC: National Academy Press.
Jensen, A.R. (1969). How much can we boost IQ and scholastic achievement?
Harvard
Educational Review, 39,
1–123.
Jensen, A.R. (1980).
Bias in mental testing.
New York: Free Press.
Jensen, A.R. (1985). The nature of the black-white difference on various psychometric
tests: Spearman’s hypothesis.
Behavioral and Brain Sciences, 8,
193–263.
Kaufman, A.S. (1990).
Assessing adolescent and adult intelligence.
Needham Heights,
MA: Allyn & Bacon.
Loehlin, J.C., Lindzey, G., & Spuhler, J.N. (1975).
Race differences in intelligence.
San
Francisco: W.H. Freeman.
Massey, D.S., & Denton, N.A. (1993).
American apartheid: Segregation and the making of
the underclass.
Cambridge, MA: Harvard.
Massey, D.S., & Hajnal, Z.L. (1995). The changing geographic structure of black-white
segregation in the United States.
Social Science Quarterly, 76,
527–542.
National Center for Education Statistics. (1995).
Minority undergraduate participation in
postsecondary education
(Report No. 95166). Washington, DC: U.S. Department of
Education.
Neisser, U., Boodoo, G., Bouchard, T.J., Jr., Boykin, A.W., Brody, N., Ceci, S.J., Halpern,
D.F., Loehlin, J.C., Perloff, R., Sternberg, R.J., & Urbina, S. (1996). Intelligence:
Knowns and unknowns.
American Psychologist, 51,
77–101.
Orfield, G. (1993).
The growth of segregation in American schools: Changing patterns of sep-
aration and poverty since 1968.
Alexandria, VA: National School Boards Association.
Reynolds, C.R., Chastain, R.L., Kaufman, A.S., & McLean, J.E. (1987). Demographic char-
acteristics and IQ among adults: Analysis of the WAIS-R standardization sample as a
function of the stratification variables.
Journal of School Psychology, 25,
323–342.
Rivkin, S.G. (1994). Residential segregation and school integration.
Sociology of Educa-
tion, 67,
279–292.
Yinger, J. (1995).
Closed doors, opportunities lost: The continuing costs of housing dis-
crimination.
New York: Russell Sage.
(R
ECEIVED
5/8/97; A
CCEPTED
9/16/97)
Acknowledgments—
We would like to thank Lisa Jenkins and Jeremy Manier
for their valuable comments.
by guest on February 20, 2013pss.sagepub.comDownloaded from