See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/4901569

IQ and Income Inequality in a Sample of Sibling Pairs from Advantaged Family

Backgrounds

Article in American Economic Review · February 2002

DOI: 10.1257/000282802320191570 · Source: RePEc

CITATIONS READS

69 1,946

1 author:

Charles Murray
American Enterprise Institute
36 PUBLICATIONS 927 CITATIONS

SEE PROFILE

All content following this page was uploaded by Charles Murray on 14 April 2015.

The user has requested enhancement of the downloaded file.

American Economic Association

IQ and Income Inequality in a Sample of Sibling Pairs from Advantaged Family Backgrounds
Author(s): Charles Murray
Source: The American Economic Review, Vol. 92, No. 2, Papers and Proceedings of the One
Hundred Fourteenth Annual Meeting of the American Economic Association (May, 2002), pp.
339-343
Published by: American Economic Association
Stable URL: http://www.jstor.org/stable/3083428
Accessed: 08/03/2009 19:50

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=aea.

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].

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

http://www.jstor.org
 IQ and Income Inequalityin a Sample of Sibling Pairs
from Advantaged FamilyBackgrounds
By CHARLESMURRAY*

The Bell Curve (Richard Herrnstein and which the siblings had lived with both biologi-
Murray, 1994) presented data on the indepen- cal parentsfor at least seven years afterbirth,to
dent effect of IQ on a wide variety of social and minimize differential family backgroundsaris-
economic outcomes for members of the Na- ing from divorce and remarriage.Both limits on
tional Longitudinal Survey of Youth (NLSY). the sample were essential to preservethe virtues
To control for socioeconomic background,we of the sibling comparison.But notice what had
constructed an index using the standardthree been done in the process. In applying those
indicators:parentaleducation, occupation, and conditions, the sibling sample represented a
income. Among the many threads in the re- populationin which all parentswere wed when
sponse to The Bell Curve, the following ques- the child was born (zero illegitimacy) and all
tion arose: How much would the independent young children were brought up by both bio-
effect of IQ have been attenuatedif a broader logical parents during their most formative
set of family background variables had been years (zero early divorce).
used as controls?To test this, SandersKorenman Such a population is one that has achieved
and Christopher Winship conducted a fixed- much of what we ideally want to achieve
effects analysis of the large numberof siblings through social policy. The thought occurredto
within the NLSY, in effect controlling not just me: why not complete the process? Having al-
for socioeconomic status, but for everything in ready createda sample without illegitimacy and
the shared environmentof the family. The re- early divorce, why not slay as well the great
sults were that "[w]ith a few exceptions, the beast of social policy, poverty?To achieve that,
fixed-effects estimates for AFQT [the cogni- I lopped off the sibling pairs whose parents
tive test used in the NLSY] are remarkably were anywherein the bottom 25 percent of the
similarto the standardOLS and logit estimates" income distributionas of 1978-1979, when the
(Korenman and Winship, 2000 p. 146). The NLSY began. This produced a sample of 733
independent effect of IQ is robust across sibling pairs who grew up in households which,
methods. by 1978-1979, had a median parental family
I subsequentlyconductedmy own analysis of income of $64,586 and a minimum income of
the NLSY siblings for a publicationon income $30,486 (expressed in 2000 dollars). For prac-
inequalityand IQ (Murray,1998). In the course tical purposes, I had by this process nearly
of that work, anotheraspect of sibling analysis achieved the utopia of income distributionas
struckme. The procedure,which is used in this defined by many, one in which the lowest in-
paper as well, begins by identifying every sib- come is half the median income.
ling pair in which one sibling had scored in the I hereby dub these 733 pairs the utopian
normalrange, defined as a tested IQ of 90-109, sample. It is utopian not just because it has
and the other member of the pair had scored virtually no illegitimacy, divorce, and poverty.
somewhere outside that range. Those in the The way it has been selected has also necessar-
normal range were used as the reference group ily effected drastic improvementsin the neigh-
againstwhich sibling outcomes were compared. borhoods,peers, and educationalsystems which
I then limited the sample to full biological sib- the youths attended. The members of the uto-
lings, to take genetics out of the picture, and pian sample had a big edge in their potential
further limited the sample to sibling pairs in access to college, both economically and be-
cause the sample is highly selected for the kind
of parentswho actively encouragetheirchildren
* AmericanEnterpriseInstitute, 1150 17th Street, N.W., to continuetheireducations.The same selection
Washington,DC 20036. factors mean that I have created a sample in
339
340 AEA PAPERSAND PROCEEDINGS MAY2002

which the incidence of good health care, child- From this perspective, any index intended to
hood nutrition, and nurturinghome environ- capture family background variables may be
ments are all high comparedto the populationat criticized for soaking up variance that is, for
large. example, not just "the effects of having money
I will use the utopian sample to address the in the house," but also "the effects of having
following policy question: How much differ- parentscapable of making money," and so on.
ence would it make to income inequality if, Any socioeconomic status(SES) index, no mat-
magically, every child in the country could be ter how basic, must cope with this problem.To
given the same advantages as the more fortu- augmentthe basic SES variableswith a laundry
nate of our children? This is a highly charged list of additionalindependentvariablesthat are
political issue, so perhaps I should begin with in part the result of parentalindividual differ-
common ground.No one doubts that some nar- ences only compounds the confounding. Psy-
rowing of inequalities would occur. The envi- chometriciansattackthis problem with a set of
ronmentin which a child is raised does make a techniques designed to partitionboth the envi-
difference.The questionis not whether,but how ronmentaland genetic components of the per-
much. sonal characteristicsand their contributionto
The controversy in estimates about how the observed social and economic outcomes.
much narrowing would occur is being played Unlike the econometricians,they have tended
out via two very different academic traditions. to conclude that the shared environment for
The first is the economics/sociology tradition. siblings (such things as parental income, the
The KorenmanandWinshippaperI cited earlier neighborhood, and school) constitutes a small
is an example. After the sibling analysis, which proportionof environmentalinfluences, which
yields results very similar to those presentedin instead are dominated by the nonshared envi-
TheBell Curve,the authorsembarkon analyses ronment (i.e., a host of subtle ways in which
that add many more independent variables to childrenexperience the same environmentwith
the regression equations (e.g., family arrange- opposite effects, or have differentformativeex-
ment when the child was 14 years old, whether periencesdespite living in the same family) (see
the respondent's family had a librarycard,num- e.g., Joseph Lee Rodgers and David C. Rowe,
ber of siblings, and age of motherat the child's 1985; D. A. Grayson, 1989; RobertPlomin and
birth),put the data throughtransformationsthat C. S. Bergman, 1991; Rowe, 1994).
the authorsconsider appropriate,and conclude I applaudefforts from both traditionsto cal-
that environmental background variables are ibratehow much narrowingin outcomes might
much more important than Herrnstein and I theoretically be achieved by different policy
thought. options. However, my proposition is that the
The other major line of inquiry draws from sibling data suggest that the current levels of
psychometrics, which approachesthe interpre- income inequality are likely to persist or, if
tation of such a regression equation much dif- temporarilyreduced,to reboundunder any pol-
ferently. Take parentalincome as an example. icies short of the Swedish model. The utopian
Whereas an econometric analysis of social and sample offers a way of thinking about why this
economic outcomes among offspringis likely to will be so. Two of the usual suspects, poverty
treat parentalincome as exogenous, a psycho- and single parenthood, have been stripped
metricianwill see it as partly an expression of away. How much difference should this change
parentaltraits. These traits may be transmitted make to the income distribution in the next
directly to the child through genetics, but also generation?In the one after that?
indirectly. The same personal qualities that en- First, I will provide some basic information
able the parents to hold a steady, well-paying about the data.The measureof IQ is the Armed
job, be they IQ or qualities such as industrious- Forces Qualification Test, a highly g-loaded
ness and interpersonal skills, also affect the paper-and-penciltest designed for administra-
likelihood that the parents will raise a child tion to teenage students(g is the general factor
differentlyfrom people who lack the character- in mental tests). In the NLSY, the average cor-
istics necessary to hold a steady, well-paying relation of the AFQT with classic full-scale IQ
job. tests administeredto the NLSY sample when
VOL.92 NO. 2 NATUREVS. NURTUREIN DETERMININGECONOMICOUTCOMES 341

OF THESIBLINGSAMPLES
TABLE 1-CHARACTERISTICS TABLE2-MEDLAN FAMILYINcoME (IN $2000)

Complete Utopian Everybody Complete Utopian Everybody

sample subsample else sample subsample else
Characteristic (1,808 pairs) (733 pairs) (1,075 pairs) IQ group (1,808 pairs) (733 pairs) (1,075 pairs)
Ethnicity (percentages) Very bright siblings (120+) $68,300 $70,700 $65,100
Non-Latino white 54 65 47 Bright siblings (110-119) $60,200 $60,500 $59,900
Black 26 19 31 Reference group (90-109) $46,000 $52,700 $41,600
Latino 15 13 17 Dull siblings (80-89) $34,200 $39,400 $31,300
Other 4 3 5 Very dull siblings (<80) $22,000 $23,600 $18,400
Median parental
income (in $2000) $44,941 $64,586 $25,858 Note: Income representsthe average for the NLSY survey years of
Fathers with 16+ years 1993 and 1995.
of education Source: Author's analysis, National LongitudinalSurvey of Youth.
(percentage) 14 19 10
Mean parentalDuncan
score for
occupation 33 42 19
ParentalDuncan score
therebywill tend to mask evidence for my prop-
(percentage)'80 osition ratherthan inflate it.
(white collar or Table 2 shows median income of the adult
professional) 5 8 3
No father at birth sibling pairs, ages 30-38 when these data were
(percentage) 7 0 13 collected, brokendown by what Herrnsteinand
Broken home before
age 8 (percentage) 18 0 32 I called "cognitive classes." In Table 2 and the
subsequenttables, cell sizes are typically in the
Source: Author's analysis, National LongitudinalSurvey of Youth. hundreds. I note cells for which samples are
smaller than 50. The row labeled "reference
group" gives the data for the member of each
they were younger was 0.81, somewhat higher sibling pair who had an IQ between 90 and 109.
than the usually observed correlationsof those The otherrows show the results for the siblings
IQ tests with each other.The AFQT scores used of the reference subjects, groupedby their IQs.
in the analysis have been normalizedseparately Thus, for example, Table 2 says that, in the
for each year's birth cohort to a mean of 100 utopian sample, the reference subjects, all of
and a standarddeviationof 15. In all, the NLSY whom had IQs between 90 and 109, had a
contains 1,808 unique pairs of subjects who median family income of $52,700. The siblings
were partof the same household (siblings, half- of those reference subjects who had IQs of 120
siblings, or unrelated),of whom 733 pairs qual- or higher had a median family income of
ified for the utopian sample. $70,700.
Table 1 shows some of the basic characteris- I should note that the results presented in
tics of the full 1,808 pairs, with the numbers these tables have been checked for fallacies
shown for two subgroups,the utopian sample, arising from aggregationby IQ groups, using a
and "everybody else," meaning the 1,075 sib- variable that contrasts the reference and com-
ling pairs who failed one or more of the tests for parison sibling on a pairwise basis, and have
getting into the utopian sample. been cleared of that possibility.
The contrastbetween the utopiansample and Returning to the data on median income,
everybody else is marked on all the variables. these results offer a nice example of how the
Note that one cannotthink of the full 1,808-pair policy debate can hinge on what comparison
sample as being representativeof the national one chooses to focus on. Advocates for ex-
population. The NLSY oversampled a number pandedincome programscan legitimately point
of groups, and this is reflected in the ethnic to the sizable difference between the median
breakdown, in which blacks and Latinos are incomes of the utopian sample and the "every-
substantiallyoverrepresented.The bias this in- body else" sample, especially in the lower cog-
troduces, comparing the utopian sample with nitive classes. However, if one is asking how
one thatis more disadvantagedthana nationally much difference it would make if poverty and
representativeone would be, will tend to exag- illegitimacy were driven to near zero, the an-
gerate differences between the full sample and swer implied by these data is "not that much."
the utopian sample, not understatethem, and The column of numbersfor the utopian sample
342 AEA PAPERSAND PROCEEDINGS MAY2002

TABLE3-MARRIAGE(PERCENTAGES)
AMONG
SUBJECTS TABLE 4-WOMEN WHO HAVE GIVEN BIRTH OUT
WITH CHILDREN OF WEDLOCK (PERCENTAGES)

Complete Utopian Everybody Complete Utopian Everybody

sample subsample else sample subsample else
IQ group (881 pairs) (469 pairs) (412 pairs) IQ group (594 pairs) (291 pairs) (303 pairs)
Very bright siblings Very brightsiblings
(120+) 86 87 82 (120+) 4 2(n = 47) 10(n = 10)
Bright siblings Brightsiblings
(110-119) 88 91 83 (110-119) 9 10 7 (n = 46)
Reference group Referencegroup
(90-109) 24 17 30
(90-109) 69 78 62 Dull siblings
Dull siblings (80-89) 37 33 40
(80-89) 53 63 48 Very dull siblings
Very dull siblings (<80) 62 44 (n = 27) 69
(<80) 40 43 39
Source: Author's analysis, National LongitudinalSurvey of
Source: Author's analysis, NationalLongitudinalSurvey of Youth.
Youth.

makes it obvious that the dispersion even utopian sample is great. Being smart is associ-
among these fortunatechildrengrown to adult- ated with very high levels of marriageamong
hood is extremely large. Moreover,that disper- those with children, and being dumb is associ-
sion across cognitive classes is likely to increase ated with very big drop-offs, even for compar-
in years to come, even for this first generation. isons with siblings in the utopian sample.
The lower cognitive classes are in predomi- Table 4 shows the percentageof female sib-
nantly low-skill jobs, and the higher ones in lings who have ever borne a child out of wed-
white-collarand professionaljobs. The income lock. I have noted the sample sizes for the "very
trajectoryfor low-skill occupationspeaks early bright"siblings, only three out of 57 of whom
at a fairly low wage, while the income trajectory had a child out of wedlock, but the inverse
for professionalsand seniormanagerspeaks late relationshipbetween IQ and nonmaritalchild-
at much higher wages. The differences in bearing is one that has been well-documented,
earned income as of 1995, when the NLSY and it is not simply a matter of smart girls
subjects were aged 30-38, must be expected to having more abortions (Carolyn T. Halpern et
increase substantially. al., 2000). The steep rise in the percentage of
The income results from the utopian sample siblings having nonmaritalbirths as one moves
lead to an obvious question: If the below- down the cognitive classes is striking, espe-
average IQ offspringof intact,nonpoorfamilies cially if one likes to think that growing up in a
fall so far short of their parents'economic suc- stable two-parenthouseholdhas positive social-
cess, what happensin the second generation?It izing effects.
does not require sophisticatedmodeling to see Taking everything together, the image con-
that the returnto the original level of inequality veyed by the utopian sample is for very large
is likely to be rapid. One might reach this con- income and family-structuredifferences inter-
clusion simply by extrapolatingfrom the dis- twined with systematically different distribu-
persion observed in the first generation. It is tions of intelligence-all of which amounts to
possible to do better than that, however, by the kind of cognitive stratificationthat was the
examining collateral indicators involving fam- topic of TheBell Curve.And these are levels of
ily structure,which in turnare relatedto earning inequality producedby the offspring of a pop-
power as an adult. ulation more advantagedin income and family
Table 3 limits the samples to those who had structurethan even the most optimistic social
children as of 1996 and shows the percentage reformercan hope to achieve in the real world.
who were marriedas of then. As in Table 2, the The difference between the siblings that pro-
data indicate that environment does count, to duced these inequalitiesis not race, birthorder,
some degree, but the range even within the poverty, parentalsocial status, parentalmarital
VOL.92 NO. 2 NATUREVS. NURTUREIN DETERMININGECONOMICOUTCOMES 343

status, or any other systematic difference in inequality in abilities as a driving force behind
family background.The sibling analysis com- inequality in the distributionof social and eco-
bined with the utopianselection criteriatakes all nomic goods.
of those explanations out of the equation, plus
many others. The one thing that differs system- REFERENCES
atically about these siblings is their scores on a
paper-and-penciltest, which in turn is a valid Grayson,D. A. "Twins Reared Together:Mini-
and reliable measure of differences in cogni- mizing Shared EnvironmentalEffects." Be-
tive functioning. Whatever the source of those havior Genetics, 1989, 19(4), pp. 593-608.
differences in cognitive functioning, genes or Halpern, Carolyn T.; Joyner, Kara; Udry, J. R.
environment,they have proved remarkablyre- and Suchindran, C. "Smart Teens Don't Have
sistantto efforts to narrowthem (for a literature Sex (or Kiss Much Either)." Journal of Ado-
review of attempts to raise IQ, see Herrnstein lescent Health, 2000, 26, pp. 213-25.
and Murray[1994 pp. 389-416]). Herrnstein, Richard and Murray, Charles. The
People of different political viewpoints may bell curve: Intelligence and class structurein
reasonably respond to the data on the utopian American life. New York: Free Press, 1994.
sample with policy prescriptionsthat are in po- Korenman, Sanders and Winship, Christopher.
lar opposition. In many ways, the left has the "A Reanalysis of The Bell Curve: Intelli-
easier task. No one earns a high IQ. It is a gift. gence, Family Background,and Schooling,"
These data are tailor-made for the conclusion in Kenneth Arrow, Samuel Bowles, and
that a Rawlsian redistributivestate is the only Steven Durlauf, eds., Meritocracy and eco-
answer to unfair personal inequalities. The nomic inequality. Princeton, NJ: Princeton
right, for its part, must state forthrightlywhy it University Press, 2000, pp. 137-78.
thinks that a free society that tolerates large Murray, Charles. Income inequality and IQ.
differences in income, significantly caused by Washington,DC: American EnterpriseInsti-
unfair inequalities, is preferable to a more tute Press, 1998.
highly regulated society that reduces those Plomin, Robert and Bergman, C. S. "The Nature
differences. of Nurture:Genetic Influence on 'Environ-
In the meantime, it seems fair to conclude mental' Measures." Behavioral and Brain
from these datathat,thoughthe answersmay be Sciences, 1991, 14, pp. 373-427.
different for those of competing political per- Rodgers, Joseph Lee and Rowe, David C. "Does
suasions, the challenge is common to all. We all ContiguityBreed Similarity?A Within-Fam-
want a society in which everyone can find a ily Analysis of NonsharedSources of IQ Dif-
valued place. But in trying to develop policies ferences between Siblings." Developmental
to achieve that goal, too many scholars who Psychology, 1985, 21(5), pp. 743-46.
deal in policy have pretendedthey live in a Lake Rowe, David C. The limits of family influence:
Wobegon world where everyone can be above Genes, experience, and behavior. New York:
average. It is time to face the reality of human Guilford, 1994.

View publication stats