Content uploaded by William Dickens
Author content
All content in this area was uploaded by William Dickens on Feb 05, 2021
Content may be subject to copyright.
Research Article
Black Americans Reduce the
Racial IQ Gap
Evidence From Standardization Samples
William T. Dickens
1
and James R. Flynn
2
1
Economic Studies, Brookings Institution, Washington, DC, and
2
Department of Political Studies, University of Otago,
Dunedin, New Zealand
ABSTRACT—It is often asserted that Black Americans have
made no IQ gains on White Americans. Until recently,
there have been no adequate data to measure trends in
Black IQ. We analyzed data from nine standardization
samples for four major tests of cognitive ability. These data
suggest that Blacks gained 4 to 7 IQ points on non-Hispanic
Whites between 1972 and 2002. Gains have been fairly
uniform across the entire range of Black cognitive ability.
No one can really trace the Black-White IQ gap in the United
States back to its origins. Estimates for 1917 and 1943 are based
on military data subject to a host of biases, and estimates since
1945 are based almost entirely on averaging studies, none of
which compared nationally representative samples taking the
same test at two different times (Herrnstein & Murray, 1994).
Rushton and Jensen (2005) recently stated that the IQ difference
between Black and White Americans stands at 1.1 standard
deviations and is as large today as itwas nearly 100 years ago. We
believe that the racial IQ gap may have been about 1.1 standard
deviations in the late 1960s. Two data sets from nationally rep-
resentative samples, the data of the 1965 Coleman Report
(Jensen, 1980, p. 479) and the standardization data for the 1972
Wechsler Intelligence Scale for Children–Revised (WISC-R;
Harcourt Assessment, 2005b), yield an average Black-White IQ
gap of 1.108 standard deviations. More recently, Flynn (1987)
analyzed military data and found that Blacks gained 3 points on
Whites between 1940 and 1960, but this estimate was tentative.
DATA
The inclusion of Blacks in recent standardization samples
means that better data are now available. To examine whether
U.S. Blacks have made any IQ gains on U.S. Whites, we obtained
results from the following: (a) the 1972, 1989, and 2002
standardizations of the Wechsler Intelligence Scale for Chil-
dren, called the WISC-R, WISC-III, and WISC-IV, respectively
(Harcourt Assessment, 2005b); (b) the 1978 and 1995 stand-
ardizations of the Wechsler Adult Intelligence Scale (WAIS),
called the WAIS-R and WAIS-III (Harcourt Assessment, 2005b);
(c) the 1980 and 1997 standardizations of the Armed Forces
Qualification Test (AFQT; U.S. Department of Defense, 2005);
and (d) the 1985 and 2001 standardizations of the Stanford-
Binet, called the SB-4 and SB-5 (Riverside Publishers, 2005;
Thorndike, Hagen, & Sattler, 1986, pp. 34–36). Results by race
are not available for previous standardizations. The publishers
of the Wechsler and the Stanford-Binet tests provided sample
sizes, mean IQs, and standard deviations by age group for
Whites and Blacks. The U.S. Department of Defense provided
sample weights, individual test scores, and information on sub-
jects’ age, race, and ethnicity for the two standardizations of the
AFQT. Summary data are in Appendix A.
The AFQT is not administered individually, but it is one of the
most highly g-loaded tests in use (grefers to the general intelli-
gence factor). Scores on the AFQT correlate with scores on indi-
vidual classic IQ tests more highly than scores on the classic tests
correlate with one another (Herrnstein & Murray, 1994, pp. 580–
585). The gmeasured by the AFQT is skewed toward crystallized
g(the tools and skills that intelligent people tend to acquire), but
the same is true of the Wechsler tests (Jensen, 1987, p. 96).
In estimating values for Black IQ, we used the convention that
sets the White mean at 100 and the White standard deviation at
15. For example, if prerise Blacks were 1.1 White standard
deviations below Whites, this convention puts their IQ at 83.5.
RACE AND SAMPLES
Whenever we refer to Whites, we mean non-Hispanic Whites.
Hispanics score below other Whites, and in recent years, their
Address correspondence to James R. Flynn, Department of Political
Studies, University of Otago, Dunedin, New Zealand, e-mail: jim.
flynn@stonebow.otago.ac.nz.
PSYCHOLOGICAL SCIENCE
Volume 17—Number 10 913Copyright r2006 Association for Psychological Science
numbers have increased dramatically in the United States. Were
they not excluded, the Black-White IQ gap would show a decline
irrespective of any gains made by Blacks. The data for the 1972
WISC-R and the 1978 WAIS-R give values for White IQ that do
not exclude non-Hispanic Whites. We adjusted for this by
raising the White means by 0.70 and 0.62 IQ points, respec-
tively. Appendix A describes how we arrived at these values.
The adjustments made little difference, increasing the estimated
rate of gain by less than 0.03 IQ points per year.
We examined whether shifting racial-group memberships
might affect our results. Black/non-Black marriages increased
from about 1% of marriages in 1970 to 4.5% in 1990 (Farley,
1995; Staples, 1985). This means that the percentage of Black
children with half-White ancestry would rise as we moved from
those who were tested in 1980 to those who were tested in 2000.
For the moment, take it on faith that the relevant Black-White IQ
gap is less than 15 points. If such children scored halfway be-
tween the Black and White means, they would be 7.5 IQ points
above the Black mean. Arithmetic shows that the increase of
such children (up 3.5%) would cause a rise of 0.263 IQ points
(0.035 7.5 50.263). In addition, perhaps Blacks who can
pass for White are less likely to wish to do so today. If children
whose parents passed for White decided to declare themselves
Black, and if their parents provided environments in which
those children matched the White mean of 100, each such child
would bring an extra 15 IQ points as he or she entered the Black
population. However, the number of such cases must be very
small. If 1% more Blacks fall into this category today than in
1972, they would cause a rise of 0.15 IQ points. In sum, shifting
group membership is probably not a very important influence on
any recent changes in the Black-White IQ gap.
The performance of the two races could also be affected by
changes in test content. The Wechsler and Stanford-Binet or-
ganizations assured us that no item or subtest has been added or
deleted with the intention of influencing the racial IQ gap. Be-
tween 1980 and 1997, the AFQT changed from a pencil-and-
paper test to a computerized test. However, in 1997, a large
sample was randomly allocated between the two tests, and Segall
(1997, pp. 192–193) found that the computerized test gave
neither Blacks nor Whites any statistically significant advan-
tage. Finally, Jensen (1992) has shown that Black-White dif-
ferences tend to be larger on tests that correlate more highly with
g. The correlation between test scores and grose by 12% from
the SB-4 to the SB-5 (Roid, 2003, p. 108), rose by 1% from the
WISC-R to the WISC-IV, declined by 5% from the WAIS-R to
the WAIS-III (Harcourt Assessment, 2005a), and was the same
in the 1980 and 1997 versions of the AFQT (U.S. Department of
Defense, 2005). If anything, there was a slight increase in the
gloading of the tests over time.
The Wechsler and Stanford-Binet manuals show meticulous
sampling of schools and careful weighting to ensure that
standardization samples matched census data. Up to age 15,
virtually all American children are in school and can be sampled
and counted. One qualification that should be noted is that
unlike the SB-4 sample, the SB-5 sample included special-
education and limited-English-proficiency groups. Because a
higher percentage of Blacks than Whites are in these categories,
their inclusion would lower the SB-5 Black mean and deflate the
SB estimate of Black IQ gains.
Adults pose sampling problems, but individual data on the
AFQT allow a test of their significance. Neal (in press) found that
the 1980 National Longitudinal Survey of Youth (NLSY) sample
contained a sizable group of undereducated Blacks who did not
attempt many items on the AFQT. No such group was present in
the 1997 NLSY sample, so a comparison of these two samples
would overestimate Black gains. However, our results are not
based on the NLSY samples. They are based on the older Profile
of American Youth (PAY) samples used to norm the AFQT. We
checked to see whether a similar bias was present in the PAY
data. If it were, we would expect to see disproportionately large
gains among the lowest-scoring Blacks. Another possible bias is
that over the years, more and more Blacks may have become too
isolated to locate. Therefore, fewer below-average Blacks might
have been present in the 1997 PAY sample than in the 1980 PAY
sample. Weighting against census data would not be a corrective
because there would also be fewer below-average Blacks located
by the census as time passed. If this bias were present, we would
expect the more recent sample to show greater gains below than
above the median, with gains tailing off the further above the
median we go.
Neither of these distortions is present in the data. Figure 1
presents Black IQ gains on Whites on the AFQT, giving a value
for each percentile of the Black AFQT distribution. Gains are
relatively uniform across the entire distribution of Black ability,
at least from the 3rd to the 88th percentiles. Only the bottom 2%
show heightened gains, but they are only slightly greater than
gains at other points in the distribution.
Fig. 1. Blacks’ IQ gains on Whites by percentile. Results shown are from
the Armed Forces Qualification Test from 1980 through 1997.
914 Volume 17—Number 10
Black Americans Reduce the Racial IQ Gap
We also noted the fact that the number of incarcerated Blacks
(mainly young males) increased between 1980 and 1997. This
might make the later sample more elite than the earlier sample
and inflate gains. We reviewed the NLSY data for a marked rise
in the Black-White gap at the ages of heavy incarceration and
found nothing.
TRENDS FROM THE STANDARDIZATION SAMPLES
We have made available on a Web site a version of this arti-
cle that contains an appendix absent here (see http://www.
brookings.edu/views/papers/dickens/20060619_IQ.pdf). That
appendix details the procedures used to construct the average
Black IQs presented in this study. The simpler procedures de-
scribed here give almost identical results and convey the ra-
tionale for the analysis.
Figure 2 shows that Black IQ rose on each of our four tests. All
of these tests but the WAIS cover ages under 25 (6–16 for the
WISC, 3–23 for the SB, and 18–23 for the AFQT). Figure 2
includes both the WAIS trend for all ages (16–74) and the WAIS
trend for individuals under age 25. The latter is in accord with
the trends for the other tests. The WISC line links two data sets,
one tracing trends from the WISC-R to the WISC-III and the
other tracing trends from the WISC-III to the WISC-IV. Setting
aside the WAIS trend for all ages, we have a total of five measures
of IQ gains by Blacks under the age of 25. If Black IQ were
constant or falling, the probability of a rise in all five of these
comparisons would be less than or equal to 1 in 32 (.03).
The terminal IQ values for young Blacks are as follows: The
WAIS (under age 25) terminates at 88.08 in 1995, the AFQT
terminates at 85.61 in 1997, the SB terminates at 88.40 in 2001,
and the WISC terminates at 88.10 in 2002. These give an
average of 87.55, and if all trends are projected to 2002, the
average rises to 88.2. The average of the median ages is 15.
Differences between Black IQs from one standardization to
another give estimates of the rate of gain. The WAIS (under age
25) shows 3.22 points gained over 17 years (rate 50.189 points/
year). The AFQT shows 3.62 points gained over 17 years (rate 5
0.213). The SB shows 1.79 points gained over 16 years (rate 5
0.112). The WISC shows 1.51 points gained from 1972 to 1989
(rate 50.089) and 4.16 points gained from 1989 to 2002 (rate 5
0.320). Averaging these gives a rate of gain of 0.185 points per
year.
The data do not show when recent Black IQ gains began. If we
take 83.5 as the value at the start of the gains, the SB trend
indicates the gains began in 1957, the WAIS (under age 25)
indicates 1971, the WISC indicates 1984, and the AFQT indi-
cates 1987. The SB date is the least plausible in that the earlier
gains began, the more likely that previous scholars would have
noticed something.
Table 1 presents estimates of the Black rate of gain and its
standard error. These estimates are based on the pooled data,
calculated both with and without controls for test and age. Dif-
ferential gains on the various tests are not large enough to allow
us to reject the hypotheses that the gains were the same on all
tests or that the gains were constant over time. Therefore,
pooling the data to compute a single rate for all tests for the
entire period is appropriate. However, the average scores on
Fig. 2. IQ scores for Blacks across three decades (White average 5100).
Results are shown for the Armed Forces Qualification Test (AFQT), the
Stanford-Binet (SB), the Wechsler Adult Intelligence Scale (WAIS), and
the Wechsler Intelligence Scale for Children (WISC). For the WAIS, mean
scores for all ages (16–74) are indicated by the dashed line, and mean
scores for individuals under age 25 are indicated by the solid line.
TABLE 1
Annual Rate of Blacks’ Gain in IQ Points
Sample
Controls
None Age
a
Test Test and age
a
All ages 0.195 (0.046)
w
0.187 (0.025)
w
0.183 (0.031) 0.188 (0.021)
Under age 25 0.212 (0.057)
w
0.161 (0.033) 0.190 (0.033) 0.184 (0.025)
Note. Rates of gain are the coefficients of a year variable in a regression of Black IQs by age, test, and year on year
and other controls specified. The specifications of the regressions are described at http://www.otago.ac.nz/
politicalstudies/jim_flynn.html. Standard errors (given in parentheses) are the maximum of those from the gener-
alized least squares procedure (w) or White robust standard errors (unmarked).
a
Controls for age were average age of the group in years, average age squared, and average age cubed. When all ages
were included, a separate age polynomial was estimated for individuals over 24.
Volume 17—Number 10 915
William T. Dickens and James R. Flynn
different tests and at different ages showed statistically signifi-
cant differences. Therefore, the preferred estimates are those
with controls for both age and test.
For Blacks under age 25, Table 1 shows a rate of gain of 0.184
IQ points per year. The rate of gain for Blacks 25 and older in the
WAIS data is smaller, but our data yield no reliable estimate for
that age group. This is because the WAIS standardization sam-
ples included only a small number of individuals ages 25 to 74.
The estimated rate of gain has a 95% confidence interval of
0.129. That is too large to allow us to reject the hypothesis that
older Blacks had the same rate of gain as those under 25.
On the basis of projected values for Black IQ at two different
times (1972 and 2002), we derived estimates of Black IQ gains
on Whites. Using our test/year/age-group data points for each
age from 4 to 23, we projected results before 1987 back to 1972
and projected results from 1987 and later forward to 2002. To
each data point, we subtracted or added the annual rate of gain
(0.184) times the difference between the year the test was ad-
ministered and the year to which we were projecting. We also
adjusted each value by subtracting the coefficient of the corre-
sponding test indicator variable from that administration (with
those coefficients normalized to represent the deviation from the
average of all tests).
Figure 3 shows the projected Black IQ values for 2002 and for
1972 as a function of age, with an ordinary least squares re-
gression line fitted to each set of values. In accord with the
scoring convention used throughout this article, if White IQ at
all ages is set at 100, the values for Black IQ in Figure 3 show
that the Black-White gap widened with age in both 1972 and
2002. Indeed, the gap widened by about 11 points between ages
4 and 23.
However, across ages, Black gains on Whites over this 30-year
period were much the same. For example, Black IQ at age 4 was
90 in 1972 and 95.5 in 2002, for a gain of 5.5 points. Black IQ at
age 23 was 79 in 1972 and 84.5 in 2002, for a gain of 5.5 points.
The pattern is clear. Blacks’ gains on Whites over time did not
vary with age, but were steady at 5.5 points at all ages.
Figure 3 shows that in 2002, the mean IQ of Black 15-year-
olds was 88.8. Recall that 88.2 was the value suggested by our
rough calculations using the terminal values in Figure 2. We
derived our prerise estimate of 83.5 from the Coleman Report
(Jensen, 1980) and the WISC-R (Harcourt Assessment, 2005b),
and their subjects were 12.5 and 11, respectively, which gives an
average age of about 12 (11.75). Figure 3 puts 12-year-old
Blacks at 90.5 in 2002. This implies that Blacks had gained a
total of 7 points on Whites. But racial differences on the tests
used in the Coleman Report may not be comparable with racial
differences on the tests in our sample. It is safer to say that 12-
year-old Blacks today have a mean IQ of about 90.5, and that
young Blacks have gained 5.5 points on Whites over 30 years. It
is worth noting that the only data set (the WISC) that covers the
entire period from 1972 to 2002 gives a gain of 5.67 points.
The 95% confidence interval for the gains indicated by our
regression estimates is just under 1.5 points. This value was
derived from the standard error for the estimated rate of gain in
Table 1. The calculation was as follows: First, standard error of
0.025 was multiplied by 1.98, yielding 0.0495, to set the 95%
confidence limit of the rate of gain. Second, a rate of gain of
0.184 0.0495 put the rate between 0.1345 and 0.2335. Third,
multiplying those rates by 30 years put the total gain between
4.035 points and 7.005 points. So our best estimate is that
Blacks under age 25 gained 5.52 points (0.184 30) on Whites,
plus or minus 1.485 points.
Blacks gained on Whites even though Whites made their own
gains. From 1972 to 2002, 12 cases in which the same subjects
took a later and an earlier version of a Wechsler or Stanford-
Binet IQ test show an average gain for all Americans of 0.311
points per year (Flynn & Weiss, 2006). If both Blacks and
Hispanics (see Appendix A for an indication of Hispanic gains)
have been gaining at a faster rate than Whites, the rate of gain for
non-Hispanic Whites (about 75% of the population) would be
approximately 0.265. Therefore, the rate of gain for Blacks has
been about 0.45 points per year (0.265 10.184).
IQ GAINS AND gGAINS
Some researchers attribute the predictive validity of IQ scores to
their correlation with the gfactor. Whether or not this is true, it
raises the question of whether Black IQ gains on Whites reflect g
gains. We were able to compute gscores for the WISC, WAIS,
and AFQT by using subtest scores (there are no race data for the
SB subtests). However, the AFQT subtests had virtually iden-
tical gloadings, which rendered correlations between the g
loadings of the subtests and score gains on the subtests mean-
ingless. Therefore, we were restricted to the WISC and WAIS
data. To compute the ggap between Blacks and Whites, we took
their average difference on the standardized first principal
component of the subtest correlation matrix and multiplied by
15 (thus making the gscores equivalent to IQ scores). By
comparing the ggap on an earlier test (e.g., the WISC-R) with
Fig. 3. Projected IQ scores (with regression lines) of Blacks of various ages
in both 1972 and 2002 (White average 5100).
916 Volume 17—Number 10
Black Americans Reduce the Racial IQ Gap
the ggap on a later test (e.g., the WISC-IV), we estimated how
much the ggap closed.
However, to compare ggains with IQ gains, we had to recal-
culate our values for Black IQ. The two were noncomparable
because our estimates of IQ trends were based on the age-group
averages publishers gave us. To achieve comparability, we had
to compute IQ differences between Blacks and Whites by
summing subtest differences. In fact, recalculation had little
effect on the estimates of Black IQ. The reason they shifted at all
is that our method of aggregating subtests differed from that used
by the publishers because they had IQ data for individuals that
we lacked. To achieve comparability between ggains and IQ
gains, the correction of IQs for the presence of Hispanics in the
WISC-R and WAIS-R standardization samples had to be omitted.
Table 2 shows that the estimated ggains of Blacks on Whites
were 91.13% of the comparable IQ gains on the WAIS (2.57/2.82
5.9113) and 94.73% of the comparable IQ gains on the WISC
(4.67/4.93 5.9473). The average of the two is 93%. Multiplying
that value by the IQ gain indicates that between 1972 and 2002,
Blacks made a ggain on Whites equivalent to 5.13 points. Table
2 also shows that when we correlated subtest gains with subtest
gloadings (i.e., the correlations of subtest scores with gfactor
scores), we got negative rather than positive values. This means
that Black gains cannot be attributed entirely to changes in g.
From the perspective of a principal-components analysis, this
conundrum might be resolved by suggesting that Blacks lost
ground on factors other than g. We find this interpretation un-
convincing and suspect that the pattern of Black gains in various
areas is not related to either the gloadings or other factor
loadings in these areas. However the trends came about, the
brute fact remains: The standard measure of the ggap between
Blacks and Whites declined virtually in tandem with the IQ gap.
CONCLUSION
Other scholars have provided scores from Blacks and Whites
who took the same test some years apart or have analyzed trends
(Gottfredson, 2005; Lynn, 1996; Murray, 2005; Vincent, 1991;
Wicherts, 2005). In every case, the samples lacked the quality of
standardization samples. Nonetheless, all results from these
other studies are compatible with our estimate of an IQ of 90.5
for Black schoolchildren in 2002. Some of the studies show
little or no change during the periods they cover, and some
show Black children reaching that value well before 2002,
but none cast doubt on the contention that Blacks have matched
our estimate (see Appendix B). All existing data suggest that
since the 1960s, Black children have made large IQ gains
relative to Whites, even if the precise timing of those gains is
uncertain.
The constancy of the Black-White IQ gap is a myth and
therefore cannot be cited as evidence that the racial IQ gap is
genetic in origin. Blacks have gained 4 to 7 IQ points on Whites
over the past 30 years. Neither changes in the ancestry of the
individuals classified as Black nor changes in those who identify
themselves as Black can explain more than a small fraction of
this gain. Therefore, the environment has been responsible.
The past two decades have seen both positive and negative
developments for Blacks: Gains in occupational status and
school funding have been accompanied by an increase in the
number of Black preschoolers in single-parent homes and by a
decrease in income in those homes (Neal, in press). We believe
that further Black environmental progress would engender
further Black IQ gains.
Acknowledgments—We thank Rebecca Vichniac and Jennifer
Doleac for able research assistance and participants in seminars
at The Brookings Institution and the Psychology Department of
the University of Virginia for helpful comments. The unpub-
lished Wechsler data are copyright 2005 by Harcourt Assess-
ment, Inc., all rights reserved; the unpublished Stanford-Binet
data are copyright 2005 by Riverside Publishers, all rights re-
served; and the unpublished Wonderlic data are copyright 2006
by Wonderlic, Inc., all rights reserved. We would like to thank
the publishers for access to these data and also extend thanks to
J.M. Wicherts for the same courtesy.
TABLE 2
Blacks’ gGains (Converted to IQ Metric) on Whites Compared with Blacks’ IQ Gains on Whites
Gain or correlation
WAIS-R to WAIS-III
WISC-R to
WISC-IV
Full
sample
Individuals
under age 25
Gain in g1.17 2.57 4.67
Gain in Full Scale IQ (calculated from subtest scores) 1.20 2.82 4.93
Gain in Full Scale IQ (from publishers’ data) 1.09 2.60 4.96
Average correlation of White-Black difference with gloadings
a
.65 .74 .86
Average correlation of subtest gains with gloadings
a
.28 .73 .38
Note. WAIS-R and WAIS-III are the 1978 and 1995 standardizations of the Wechsler Adult Intelligence Scale (Harcourt
Assessment, 2005b); WISC-R and WISC-IV are the 1972 and 2002 standardizations of the Wechsler Intelligence Scale for
Children (Harcourt Assessment, 2005b).
a
Common subtests only were included in these calculations.
Volume 17—Number 10 917
William T. Dickens and James R. Flynn
REFERENCES
Farley, R. (1995). State of the union: America in the 1990s. New York:
Russell Sage Foundation.
Flynn, J.R. (1987). Flynn replies to Nichols. In S. Modgil & C. Modgil
(Eds.), Arthur Jensen: Consensus and controversy (pp. 234–235).
Lewes, England: Falmer Press.
Flynn, J.R., & Weiss, L.G. (2006). American IQ gains from 1932 to
2002: The significance of the WISC subtests. Manuscript submitted
for publication.
Gottfredson, L.S. (2005). Implications of cognitive differences for
schooling within diverse societies. In C.L. Frisby & C.R. Reynolds
(Eds.), Comprehensive handbook of multicultural school psy-
chology (pp. 517–555). New York: Wiley.
Harcourt Assessment. (2005a). [Factor analysis of Wechsler tests].
Unpublished raw data.
Harcourt Assessment. (2005b). [Standardization data for Wechsler
tests]. Unpublished raw data.
Herrnstein, R.J., & Murray, C. (1994). The bell curve: Intelligence and
class structure in American life. New York: Free Press.
Jensen, A.R. (1980). Bias in mental testing. London: Methuen.
Jensen, A.R. (1984). The black-white difference on the K-ABC:
Implications for future tests. Journal of Special Education,18,
377–408.
Jensen, A.R. (1987). The gbeyond factor analysis. In R.R. Ronning,
J.A. Glover, J.C. Conoley, & J.C. Witt (Eds.), The influence
of cognitive psychology on testing (pp. 87–142). Hillsdale, NJ:
Erlbaum.
Jensen, A.R.(1992). Spearman’s hypothesis: Methodology and evidence.
Multivariate Behavioral Research,27, 225–233.
Lynn, R. (1996). Racial and ethnic differences in intelligence in the
United States on the Differential Ability Scale. Personality and
Individual Differences,20, 271–273.
Murray, C. (2005). The inequality taboo. Retrieved November 30,
2005, from http://www.commentarymagazine.com/production/files/
murray0905.html
Naglieri, J.A., & Jensen, A.R. (1987). Comparison of black-white
differences on the WISC-R and the K-ABC: Spearman’s hypoth-
esis. Intelligence,11, 21–43.
Neal, D. (in press). Why has black-white skill convergence stopped?
In E. Hanushek & F. Welch (Eds.), Handbook of economics of
education. New York: Elsevier.
Raven, J. (2000). Manual for Raven’s Progressive Matrices and Vocabu-
lary Scales: Research supplement no. 3. Oxford, England: Oxford
Psychologists Press.
Riverside Publishers. (2005). [Standardization data for the 2001
Stanford-Binet]. Unpublished raw data.
Roid, G.H. (2003). Stanford-Binet Intelligence Scales (5th ed.), techni-
cal manual. Itasca, IL: Riverside Publishing.
Rushton, J.P., & Jensen, A.R. (2005). Thirty years of research on race
differences in cognitive ability. Psychology, Public Policy, and
Law,11, 235–294.
Segall, D.O. (1997). Equating the Cat-asvab. In W.A. Sands, B.K.
Waters, & J.R. McBride (Eds.), Computerized adaptive testing:
From inquiry to operation (pp. 181–198). Washington, DC:
American Psychological Association.
Staples, R. (1985). Changes in black family structure: The conflict
between family ideology and structural conditions. Journal of
Marriage and Family,47, 1005–1013.
Thorndike, R.L., Hagen, E.P., & Sattler, J.M. (1986). The Stanford-
Binet Intelligence Scale: Fourth Edition—technical manual.
Chicago: Riverside Publishing.
U.S. Bureau of the Census. (1976). Statistical abstract of the United
States, 1976, 97th annual edition. Washington, DC: U.S. Gov-
ernment Printing Office.
U.S. Bureau of the Census. (1979). Statistical abstract of the United
States, 1979, 100th annual edition. Washington, DC: U.S. Gov-
ernment Printing Office.
U.S. Bureau of the Census. (1999). Table 8. Race and Hispanic origin of
the population by nativity: 1850 to 1990. Retrieved August 22,
2006, from http://www.census.gov/population/www/documentation/
twps0029/tab08.html
U.S. Department of Defense. (2005). [Standardization data for the
Armed Forces Qualification Test from the Profiles of American
Youth of 1980 and 1997]. Unpublished raw data.
Vincent, K.R. (1991). Black/white IQ differences: Does age make the
difference? Journal of Clinical Psychology,74, 266–270.
Vincent, K.R., & Cox, J.A. (1974). A re-evaluation of Raven’s Standard
Progressive Matrices. Journal of Psychology,88, 299–303.
Wicherts, J.M. (2005). [Comparison of White and Black groups in the
U.S. research samples for the Woodcock-Johnson-R and the
Woodcock-Johnson-III]. Unpublished raw data.
Wonderlic. (2006). [Norms by race and age for the Wonderlic Personnel
Test N: Normative studies done in 1970, 1983, 1992, and 2001].
Unpublished raw data.
(RECEIVED 9/30/05; REVISION ACCEPTED 2/13/06;
FINAL MATERIALS RECEIVED 2/23/06)
APPENDIX A: SUMMARY DATA
Table A1contains the summary data from the test publishers and
the U.S. Department of Defense. Breakdowns by age are avail-
able upon request. The WISC-R and WAIS-R scores for Whites
have not been corrected for the inclusion of Hispanics (see the
next two paragraphs). Riverside Publishers requested that we
note the following: First, controlling for levels of parental edu-
cation substantially reduces IQ differences between ethnic
groups. Second, the SB-5 standardization sample included
special-education and limited-English-proficiency groups not
included in the SB-4 sample. Our response to the second point is
that because a higher percentage of Blacks than Whites are in
these categories, the inclusion of these groups in the SB-5
standardization sample would lower the SB-5 Black mean and
deflate the estimate of Black IQ gains.
All White samples consisted of non-Hispanic Whites except
the 1972 WISC-R and the 1978 WAIS-R samples. Census and
Current Population Survey data show that the percentage of the
U.S. population that was Hispanic was 4.46 in 1970 (U.S. Bur-
eau of the Census, 1999), 5.13 in 1973 (U.S. Bureau of the
Census, 1976, Table 41, p. 34), and 5.57 in 1978 (U.S. Bureau of
the Census, 1979, Table 35, p. 33). The last date corresponds
exactly with the WAIS-R. If 5.57% of the total sample was
Hispanic in 1978, then 6.31% of Whites were Hispanic. The
percentages for 1970 and 1973 were interpolated to give a
Hispanic percentage of 4.96 in 1972, at the time of the WISC-R;
therefore, 5.86% of Whites in this sample were counted as
Hispanic.
918 Volume 17—Number 10
Black Americans Reduce the Racial IQ Gap
The Coleman Report of 1965 (Jensen, 1980, p. 479) showed
Hispanics ages 8 to 17 at 12.79 IQ points below non-Hispanic
Whites; the SB-4 of 1985 (Thorndike et al., 1986, pp. 34–36)
showed Hispanics ages 2 to 23 at 8.87 points below. Interpol-
ation of these values gave an estimate for Hispanics at the time of
the WISC-R (1972)—specifically, that they were 11.42 points
below non-Hispanic Whites. Multiplying 11.42 by the per-
centage of Whites that were Hispanic (i.e., 11.42 0.0586) put
the reduction in the White score due to the inclusion of His-
panics at 0.67 IQ points.
The Coleman Report showed 17-year-old Hispanics at 12.975
points below non-Hispanic Whites; the SB-4 showed Hispanics
ages 12 to 23 at 8.16 points below. Interpolation of these values
gave an estimate for Hispanics at the time of the WAIS-R
(1978)—specifically, that they were 9.85 points below non-
Hispanic Whites. Multiplying 9.85 by the percentage of Whites
that were Hispanic (9.85 0.0631) put the reduction in the
White score due to the inclusion of Hispanics at 0.62 IQ points.
Adding the Hispanic corrections to the White means made
little difference in the final results. The estimated rate of Black
IQ gains rose by less than 0.03 IQ points per year in all re-
gressions described in Table 1.
APPENDIX B: TRENDS FROM OTHER SAMPLES
We have made an effort to locate every study in the literature that
is informative about the change in the Black-White IQ gap on
specific tests over the last generation. Vincent (1991) presented
data on Raven’s Progressive Matrices for two samples. The 1973
sample consisted of 215 Whites and 135 Blacks from a re-
habilitation unit in Houston, Texas (Vincent & Cox, 1974). From
the data he presented, it is clear that one race or the other had
more education than the group it represented, but it is not clear
which race this was. The 1985 sample consisted of 631 Whites
and 209 Blacks from Decatur, Alabama (Raven, 2000, pp. 19–
21). Between 1973 and 1985, Black IQ (normed on Whites) went
from 84 to 93, a huge gain. However, the first sample had a
median age of 29, and the second had a median age of 9. Our
estimates indicate that Blacks lose about 10 points on Whites
between ages 9 and 29 (the rate slows down after age 24).
Therefore, the 1973 mean must be raised to 94 (84 110). This
study sets a pattern that is repeated by others: Blacks show no
gain on Whites over time, but somehow reach an IQ level well
above the traditional estimate of 83.5.
Gottfredson (2005) concluded that the Black-White gap is
between 0.8 and 1.2 standard deviations. She reported results
from many of the samples we have analyzed, but did not have
access to our most recent data, and none of her values from those
samples preclude our 2002 estimates. She reported trends on
one test we have not yet discussed, namely, the Wonderlic
Personnel Test, and we review here all existent Wonderlic data.
The Wonderlic is a 12-min test with 50 items. Between 1970
and 2001, that test was normed four times on samples of job
applicants. Increasing reluctance to record race reduced the
TABLE A1
IQ Means and Standard Deviations for Whites and Blacks
Test
Mean IQ Standard deviation
Number of
observations
White Black White Black White Black
Stanford-Binet
SB-4 103.6 90.0 15.37 13.86 3,691 711
SB-5 102.9 92.1 13.93 14.47 2,070 384
WISC
WISC-R 102.3 86.4 14.08 12.63 1,870 305
WISC-III 103.5 88.6 13.86 12.83 1,543 337
WISC-IV 103.2 91.7 14.52 15.73 1,402 343
WAIS (all ages)
WAIS-R 101.4 86.8 14.65 13.14 1,664 192
WAIS-III 102.6 89.1 14.81 13.31 1,523 247
WAIS (under age 25)
WAIS-R 101.2 87.0 14.28 13.54 519 72
WAIS-III 102.6 90.9 14.59 12.31 413 93
AFQT
1980 100.0
a
82.0 15.00
b
13.63 5,533 2,298
1997 100.0
a
85.6 15.00
b
13.23 2,880 1,191
Note. The SB-4 and SB-5 are the 1985 and 2001 standardizations of the Stanford-Binet (Riverside Publishers, 2005; Thorndike,
Hagen, & Sattler, 1986, pp. 34–36); the WISC-R, WISC-III, and WISC-IV are the 1972, 1989, and 2002 standardizations of the
Wechsler Intelligence Scale for Children (Harcourt Assessment, 2005b); the WAIS-R and WAIS-III are the 1978 and 1995 stand-
ardizations of the Wechsler Adult Intelligence Scale (Harcourt Assessment, 2005b); the AFQT is the Armed Forces Qualification
Test (U.S. Department of Defense, 2005).
a
This value is 100 by construction.
b
This value is 15 by construction.
Volume 17—Number 10 919
William T. Dickens and James R. Flynn
number of identified Whites from 123,000 to 15,600 and the
number of identified Blacks from 34,000 to 2,933. E.R. Long
(personal communication, January 5, 2006) noted that more
Americans ages 16 to 24 remain in education (rather than
seeking work) today than in the past, and that a larger number of
older workers of high quality are working intermittently. He also
noted that these changes affect the races unequally, and might
have a differential effect on the performance of the races from
one normative sample to another.
Wonderlic data from the 1970, 1983, 1992, and 2001 nor-
mative samples (Wonderlic, 2006) show a variety of trends.
Blacks ages 16 to 24 began at an IQ of 84.6 in 1970, rose to 87 in
1992, and remained there in 2001. Blacks of all ages (16 and
above) began at 85 in 1970 and were again at 85 in 2001. The
Wonderlic value of 87 for Blacks ages 16 to 24 circa 2001 tallies
with our results, which show an IQ of 86 for 20-year-old Blacks
circa 2002. But although we found Blacks made gradual gains
between 1970 and 2001, the Wonderlic data show a sudden rise
in 1992, with no change in any other year.
Murray (2005, footnotes 41 and 44) noted that standardiza-
tions of the Kaufman Assessment Battery for Children (K-ABC)
in 1983 and 2004 showed Black IQs of 93 and 92.1, respec-
tively. The K-ABC subtests were selected to minimize the Black-
White gap (Jensen, 1984) and reduce gloadings in favor of
measuring short-term memory (Naglieri & Jensen, 1987). Jen-
sen (1984) argued that the 1983 sample contained too great a
range of ability, thus yielding an inflated standard deviation and
a diminished Black-White gap (measured in standard deviation
units).
The Woodcock-Johnson was normed on excellent standard-
ization samples in 1987 and 1999. However, the race data do not
come from the full samples. Rather, research samples were
taken from the full samples, and subsamples of these research
samples (those who took all subtests used to compute IQ) were
used to calculate gscores (the Woodcock-Johnson is peculiar in
that it weights subtests by their gloading to calculate IQ scores).
These subsubsamples included 90% of the 1987 research
sample and just over 50% of the 1999 sample. Thus, all racial
comparisons must be based on Blacks and Whites in the sub-
subsamples, and Blacks were more likely than Whites to be
excluded from these subsubsamples in both years (fewer Blacks
than Whites take all of the subtests). Wicherts (2005) gave us
data for ages 1 to 65 in these two subsubsamples: Black IQ was
steady at 88 between 1987 and 1999 for all ages combined; for
individuals under age 25, it stood at 90 in 1987 and 88.70 in
1999.
Our analysis of nine Wechsler, Stanford-Binet, and AFQT
standardization samples showed Black IQ (age 12) rising from
85 in 1972 to 90.5 in 2002 (see Fig. 3). For young Blacks, the
four tests we have analyzed in this appendix give IQs of 93 in
1983 (K-ABC), 93 in 1985 (Raven’s), 90 in 1987 (Woodcock-
Johnson), and 87 in 1992 (Wonderlic). Averaging these values
puts Black IQ at 91 circa 1987. This is an almost perfect match
for our 2002 value, but it was attained 15 years too soon, and no
gains were recorded from 1987 to the present. In sum, these
imperfect data support the contention that Black schoolchildren
have attained an IQ of 90.5, but have them reaching that value
much earlier than our results from the nine standardization
samples.
Lynn (1996, p. 272) used results by age to infer trends for the
Black-White IQ gap. Age patterns do not chart trends over time,
but rather reflect an altering Black-White gap as cohorts age.
His value of 85.83 (our convention) for Blacks ages 6 to 17 in
1986 is close to our value of 84 for the WISC-III in 1989.
920 Volume 17—Number 10
Black Americans Reduce the Racial IQ Gap
