206 44 52MB
English Pages 324 [340] Year 1996
•
I
Bell
r
Cu
Claude S. Fischer, Michael Hout, Martin Sanchez Jankowski, Samuel R. Lucas, Ann Swidler. .
andKi
Tr
"
Inequality by Design
Inequality by Design CRACKING THE BELL CURVE MYTH
Claude
S.
Fischer,
Michael Hout,
Martin Sanchez Jankowski,
Samuel
Ann
R. Lucas,
Swidler,
and Kim Voss Department of Sociology University of California, Berkeley
PRINCETON UNIVERSITY PRESS PRINCETON, NEW JERSEY
Copyright
©
1996 by Princeton University Press
Published by Princeton University Press, 41 William Street,
New Jersey 08540 Kingdom: Princeton University Chichester, West Sussex
Princeton, In the United
Press,
All Rights Reserved
Library of Congress Cataloging-in-Publication Data Inequality by design bell
curve myth
/
Claude p.
cracking the
:
S.
Fischer
.
.
.
[et al.].
cm.
Includes bibliographical references and index.
—
ISBN 0-691-02899-0 (cl alk. paper). ISBN 0-691-02898-2 (pb alk. paper) :
:
1
3.
.
Intellect.
2.
Nature and nurture.
4. Intelligence levels
United States. 6.
—United —Social aspects
Intelligence levels
5.
States.
Educational psychology.
Herrnstein, Richard I.
Fischer,
Claude
J.
S.,
Bell curve.
1948-
BF431.I513 1996 96-2171 305.9'082—dc20 This book has been composed in Times
CIP
Roman
Princeton University Press books are printed
on acid-free paper and meet the guidelines
permanence and durability of the Committee on Production Guidelines for Book Longevity
for
of the Council on Library Resources Printed in the United States of America
3579 7
9
8642
10 10
(Pbk.)
8
TO OUR CHILDREN
We
have been quick
failures in
to seek explanations of our
what we are instead of what we do.
to the belief that our situation is a
problems and
We
seem wedded
consequence of our nature
rather than of our historical acts
.
.
.
—Kenneth Bock, Human Nature Mythology
* Contents *
Figures and Tables
ix
Preface
xi
Chapter
Why
1
Inequality?
Chapter
3
2
22
Understanding "Intelligence"
Chapter But
Is It
3
55
Intelligence?
Chapter 4
Who
Wins?
Chapter
Who
70
Loses?
5
The Rewards of the Game: Systems of
Inequality
102
Chapter 6
How
Unequal? America's Invisible Policy Choices
Chapter
7
Enriching Intelligence:
Chapter
129
More
Policy Choices
158
8
Race, Ethnicity, and Intelligence
171
Chapter 9 Confronting Inequality
in
America: The Power of Public
204
Investment
Appendix
Summary
1
of The Bell Curve
21
Appendix 2 Statistical
Analysis for Chapter 4
225
Notes
241
References
277
Index
303
Figures and Tables *
Figures 1.1
Changes
1.2
Explained Variance in Household Income Accounted for by
in
Household Incomes, 1959-1989, by Income Class
5
Intelligence 2.
15
Distribution of Original Scores on the
AFQT
and Distribution
of Scores as Transformed by Herrnstein and Murray
32
2.2
Different Interpretations of Predictive Validity
36
2.3
Interpreting Criterion Validity
38
4.1
Herrnstein and Murray's
Model of the Causes of Inequality
and Social Problems
73
4.2
Our Model of the Causes of
4.3
Probability That an
AFQT Score 4.4
in
1990 by
was Poor
in
1990 by
NLSY Respondent Was Poor in
1990 by
8
NLSY Respondent
85
Social Environment, and Formal Education
Probability That an
AFQT Score 4.7
Poor
Score and Social Background
Probability That an
AFQT Score, 4.6
NLSY Respondent Was
NLSY Respondent Was Poor in
87
1990 by
and Gender
89
Explained Variance in Household Income Accounted for by (a) Intelligence
Alone versus
(b) Social
Environment and
Gender Alone
100
Today
5.1
Estimates of Inequality from
5.2
Percentage of All Household Income Received by Highest-
1
800
to
107
Income 5 Percent and Lowest-Income 40 Percent of Households, 1930-1994
Who
Earned Enough
110
5.3
Percentage of Full-Time Workers
5.4
Keep a Family of Four Out of Poverty, 1964-1994 Income Changes Over a Decade for Men in Their Thirties and Forties,
to
1950-1993
Age Group, 1966-1994
Rates of Poverty by
5.6
Ratios of Earnings for High-, Median-, and Low-Earners in
Six Nations
1
19
122
Ratios of Incomes for High-, Median-, and
Households
112
114
5.5
5.7
74
and Family Background
Probability That an
AFQT 4.5
Inequality and Social Problems
in
Eight Nations
Low-Income 124
ix
FIGURES AND TABLES 6.
1
Percentage of Children
Who
Government Action,
Eight Nations
in
7.1
An
7.2
Probability that Students
8.1
A Model
Are Poor, Before and After 134
S-shaped Learning Curve
Were
161 in
College Track by Math Test
Score and Social Class of
How Low
164 Ethnic Position Causes
Low
Test
Scores 8.2
174
Index of Black Residential Isolation, Eighteen Northern Cities,
1890-1990
184
Tables Women,
the Unmarried, and Parents
4.1
Poverty Rates for
6.1
147
8.1
Change in Family Income, 1977 to 1990 Group Differences Around the World
A2.1
Descriptive Statistics of All Variables by Race
230
A2.2
Description of Variable Coding
231
A2.3
Logistic Regression of Likelihood of a Person Being in
A2.4
Logistic Regression of Likelihood of a Person Being in
A2.5
Logistic Regression of Likelihood of a
A2.6
Logistic Regression of Likelihood of a
192
Poverty in 1990 (Whites Only)
233
234
Poverty in 1990 (African Americans Only)
Interviewed in
Interviewed in
A2.7
Jail after
Jail after
AFQT
A2.8
Being
236
(Whites Only)
Logistic Regression of Likelihood of a
AFQT
Man
Being
Americans Only)
Woman
AFQT (African
Woman
237
Having an
(Whites Only)
Logistic Regression of Likelihood of a Illegitimate Child after
Man
AFQT (African
Illegitimate First Child after
90
238
Having an
Americans Only)
239
* Preface *
w
were
impelled to write
book by
this
the publication in late 1994 of
The Bell Curve. That immensely well publicized book was then the
latest
statement of a philosophy that gained extensive credence in the 1990s: The
widening inequalities among Americans century are inevitable. Because of
the market, because of the nature of
more and more by
sarily divide
that
human
modern
social class
ophy. Besides being morally complacent, foundation. Research has
shown
developed
in the last quarter-
nature, because of the nature of
it
society,
and is
Americans
race.
will neces-
We reject that philos-
a doctrine without scientific
that "nature" determines neither the level
of inequality in America nor which Americans in particular will be privileged or disprivileged; social conditions and national policies do. Inequalsense designed. Similarly, the market does not require us to
ity is in that
accept great inequalities for the sake of growth. Quite the reverse seems
do better the more equal
true; nations ties
can be
less
And modern socieothers are now and ours
their citizens are.
unequal than America
is in
1996;
has been in the past. The Bell Curve, in particular, as an emphatic statement
of this mistaken philosophy,
is
intelligence explain inequality.
wrong to claim that differences in native The social science evidence is clear. We
see our task as bringing such evidence to the attention of the wider public.
—came
we
—
members of Berkeley's Department of Sociolphenomenon and soon response from sociologists was in order. Some colleagues
In late 1994,
ogy
all
together to discuss The Bell Curve
agreed that a
suggested that The Bell Curve would fade from public consciousness. After all, its
arguments about the significance of intelligence had already been
dealt with.
(Almost a quarter-century ago, Christopher Jencks and his col-
leagues showed, in Inequality, that individuals' intelligence at best only
modestly affects their fortunes.) But the ideology The Bell Curve represents
is
too pervasive; the book's shock waves are too great to ignore.
social scientists, sity teachers,
we
we
feel responsible for correcting the record.
are painfully
aware
that
As citizens, we must participate much of our ongoing work to write
As
The Bell Curve has unsettled our
students.
in the national debate.
aside
this
Inequality by Design
is
As
univer-
So we
set
book.
a true collaboration. While particular individ-
uals took the lead in drafting specific chapters, everyone joined in outlin-
ing the basic argument, contributing ideas, and revising drafts. Fischer had
xi
PREFACE and giving the book
responsibility, in addition, for coordinating the project
a single authorial voice.
We benefited from the collaboration as well of several Berkeley graduate students. Richard
Arum was
so critical to our reanalysis of The Bell Curve
data that he shares authorship of chapters 3 and 4. Elizabeth Armstrong, Leslie Bell, Charlotte Chiu, Tally Katz,
helped us find some of the scientific
Amy
literature.
uted include Judy Haier and Maureen Fesler. tion" of several scholars
both.
We thank Robert
Christopher Jencks,
who gave
Schalet, and
Berkeley
We
for
contrib-
Adam
Hochschild,
Rob Macoun, Douglas Massey, Lee
Rainwater, Paul
We
especially thank
commenting us.
who
us either suggestions or comments, or
Bellah, Fred Block, Joe Harder,
on an early
closely
draft.
Troy Duster, Robert Hauser, and Chris Winship shared manuscripts with
Sean Stryker
had the "collabora-
also
Romer, Saul Rubin, David Vogel, and Alan Wolfe.
David Levine
staff
William Dickens,
their prepublication
Audiences who heard early versions of the work
at the
University of Arizona, University of Virginia, Princeton University, the University of California, Berkeley, and the
May
national Sociological Association's Research
1995 meeting of the
Committee on
Inter-
Stratification
and Mobility helped hone our arguments.
No
large grants funded this work, but Berkeley's
Department of Sociol-
ogy and Survey Research Center provided meeting rooms and occasional secretarial assistance. In addition, the stitute
Survey Research Center and the
of Industrial Relations supported a few of the graduate students
In-
who
helped us.
We
also very
much
appreciate the
commitment and energy
Dougherty, publisher of Social Science and Public Affairs University Press, gave this book.
McKenna and editor.
We
We
that Peter
Princeton
also thank, at the Press, Michelle
Jane Low. Anita O'Brien was our amazingly efficient copy
Chris Brest drew the figures.
hope
that this
book
will help redirect the public discussion
from the mistaken, helpless view equality are very
much
tion that, as great as
whom we
within citizens' control.
it is
away
that inequality is fated or necessary
toward the more accurate, empowering understanding
to
at
We
that opportunity
look forward to a na-
today, will be a yet fairer one
when our
dedicate this work, shoulder the burdens of
xn
and
and
its
children,
citizenship.
Inequality by Design
CHAPTER
*
Why
A, we .s
1
*
Inequality?
write, Americans are engaged
in a great
debate about the in-
equalities that increasingly divide us. For over twenty years, the
gaps have widened. As the American Catholic Bishops stated
economic
in late 1995,
economy sometimes seems to be leading to three nations living by side, one growing more prosperous and powerful, one squeezed by
"the U.S. side
stagnant incomes and rising economic pressures and one
behind
left
increasing poverty, dependency and hopelessness." Being prosperous 1
mean owning
in
may
a vacation home, purchasing private security services, and
having whatever medical care one wants; being squeezed
may mean 1 when
ing one modest but heavily mortgaged house, depending on 91
hav-
dan-
ger lurks, and delaying medical care because of the expense of copayments;
and being
left
rent, relying
may mean
behind
barely scraping together each month's
on oneself for physical
safety,
and awaiting emergency aid
Most Americans
an overcrowded public
clinic.
fragile their position
One missed mortgage payment
is.
jury might be enough to push
Few deny
thing can be done about
Some
them
that inequality has it,
middle
in the
know how
or one chronic in-
into the class that has
widened. 2 The debate
at
been
left
behind.
over whether any-
is
over whether anything should be done about
it.
voices call for an activist government to sustain the middle class and
uplift the poor.
Other voices, the ones that hold sway as
government ought
to
do
less,
They argue
not more.
we write,
argue that
for balanced budgets,
lower taxes, fewer domestic programs, minimal welfare, and less regulation.
These moves, they contend, would energize the economy and
way
help the middle class.
They would
in that
also help the poor, economically
and otherwise. Speaker of the House Newt Gingrich
in
1995 said of people
on welfare: "The government took away something more important than .
.
.
money. They took away
their initiative,
morality, their drive, their pride.
I
want
.
to help
.
.
their
freedom,
them get
.
.
that back."
.
3
their
As
to
some advocates of circumscribed cannot be changed, because inequality is natural; some
the increasing inequality of our time,
government say say
it
it
ought not be changed, because inequality drives our economy. At a
deeper
level, then, the
explains
its
the debate.
origin,
debate
is
about
what explains
its
how
to understand inequality
growth. That
is
where we
shall
— what engage
CHAPTER The arguments over
1
policy emerged from almost a quarter century of
economic turmoil and disappointment. Middle-class Americans saw the era of seemingly
ever-expanding affluence for themselves and ever-
expanding opportunities for
The
their children
come
to
an abrupt end in 1973.
cars inching forward in the gasoline lines of the mid-1970s foreshad-
owed
the next twenty years of middle-class experience.
prices rose, husbands to stay at
seemed tracted.
home
felt
worked longer hours, and even wives who preferred
pressed to find jobs. The horizons for their children
upward economic mobility conWhat was going on? What could be done about it?
In the early 1980s,
solution
was
one explanation dominated public discussion and
The cause of the middle-class less
crisis
government
was government, and
—had wrecked
the
—indeed,
by rewarding the sluggards and penalizing the
swer was
to get
talented.
stunting
The
an-
government "off the backs" of those who generate eco-
nomic growth. "Unleash the market" and the lift all
the very size of
economy by wasting money and
initiative,
that will
its
government. Regulations, taxes, programs for the poor,
preferences for minorities, spending on schools
result
would be a
"rising tide
boats, yachts and rowboats alike."
won enough
This explanation for the economic doldrums
public support
be enacted. Less regulation, less domestic spending, and more tax cuts
for the wealthy followed.
class
had not eased;
it
By
the 1990s, however, the crisis of the middle
had just become more complicated. Figure
the trends in family incomes, adjusted for changes in prices,
1989 to
stagnated,
to shrink as the opportunities for 4
public policy:
to
Wages
(the trends continued into the 1990s).
new
The
1.1
shows
from 1959
richest families
to
had soared
heights of income, the poorest families had sunk after 1970, and the
middle-income families had gained terly misleading.
The middle
class
slightly.
But
managed
this slight
to sustain
gain was
bit-
modest income
growth only by mothers taking jobs and fathers working longer hours. Also, the slight gain could not
make up
for
and parents' anxiety that key elements of the education, a stable job, and an affordable
grasp of their children.
began
to
And
growing economic insecurity
"American Dream"
home
—were
slipping
—
college
beyond the
so the phrase "the disappearing middle class"
be heard.
Another puzzle now called for explanation: The 1980s had been a boom decade; overall wealth had grown. But average Americans were working harder to stay even.
Why had the gaps between the rich and the middle and How do we understand such
between the middle and the poor widened? inequality?
WHY INEQUALITY
9
Between 1959 and 1969, income per person grew for all households. Since 1970, income per person has continued to grow rapidly for the richest households, grown at a declining rate among middle-income households, and fallen slightly among poor house-holds. The result is significantly
more
inequality.
$60,000 M «J
Top 20%
O T3 ON on ON
.A $50,000
1—1
C CO
$40,000
^A"
CLh
4-» C/5
3 $10,000
Bottom 20%
T3
< $0
1959
1979
1969
1989
Year
1.1.
Changes
in
Household Incomes, 1959-1989, by Income Class (Note:
Household incomes are adjusted by dividing income per family member by the square root of the household
size.
"Demographic Change, Rising Earnings
An answer emerged the
same voices
that
in the public debate,
had offered the
Source: Karoly and Burtless,
Inequality," table 2)
forwarded for the most part by
earlier explanation: Inequality is a
"natural," almost inevitable, result of an unfettered market.
sary by-product of unleashing talent. sink. Eventually,
free market.
however,
all
The reason such wider
the market has not
The
will gain
skilled soar
It is
the neces-
and the unskilled
from the greater efficiency of the
benefits have yet to be delivered
been freed up enough; we need
still
less
is
that
government and
CHAPTER
1
then the wealth will flow to middle- and lower-income Americans. Sharp inequality
among
trade-off for
The
strongest recent statement that inequality
market came
free
the classes, these voices suggested,
American were
is
the natural result of a
The Bell Curve: Intelligence and Class Structure
in
in
published in 1994. Richard Herrnstein and Charles Murray
Life,
argued that intelligence largely determined rich
the necessary
is
economic growth.
rich mostly because they
how
well people did in
life.
The
were smart, the poor were poor mostly
because they were dumb, and middle Americans were middling mostly because they were of middling intelligence. This had long been so but was
becoming even more so
as
new and
inescapable economic forces such as
made
global trade and technological development
by
more impor-
more open economy, people rose or sank
tant than ever before. In a
levels largely fixed
intelligence
their intelligence.
essentially innate, this expanding inequality cannot be stopped.
It
damaging the national economy. Inequality
is
is
might be
slowed by government meddling, but only by also doing injustice talented and
to the
Moreover, because intelligence
to the
in these
ways
"natural," inevitable, and probably desirable.
The Bell Curve also provided an explanation for another troubling aspect of inequality in America
—
its
strong connection to race and ethnicity.
Black families, for example, are half as likely
to
be wealthy and twice as
be poor as white families. The questions of
likely to
racial disparities
Now,
decades.
is
—and
why
to
how
to understand
do about them have anguished the nation
was a new answer
there
newed): Blacks whites; that
and what
Latinos, too
(actually, a very old
—were by
answer
for re-
nature not as intelligent as
they did less well economically, and that
is
why
little
can or should be done about racial inequality. Yet decades of social science research, and further research sent here, refute the claim that inequality
equality
is fated.
it
have such disparate standards of is this: First,
will pre-
explain
why
who
people in different classes
living. Instead,
individuals' social milieux
what
—
better explains in-
family, neighborhood,
—
community provide or withhold the means for attaining higher positions in American society, in part by providing people with mar-
school, class
we
natural and increasing in-
Individual intelligence does not satisfactorily explain
ends up in which class; nor does
equality
is
ketable
skills.
by social
Much
policy.
of what those milieux have to offer
For example, the quality of health care
is,
in turn,
shaped
that families pro-
vide and the quality of education that schools impart are strongly affected
by government
action. Second, social policy significantly influences the
rewards individuals receive for having attained their positions in society.
WHY INEQUALITY? Circumstances earn,
— such
how much
sidized
—determine
living. In turn, these
government.
We
how much money
as
professional or manual workers
tax they pay, whether their child care or housing
sub-
is
manual workers' standards of
professionals' versus
circumstances are completely or partly determined by
do not have
such inequalities to sustain or ex-
to suffer 5
pand our national standard of living. Thus, inequality
is
not the natural and
inevitable consequence of intelligence operating in a free market; in substantial
measure
it
is
and
will
always be the socially constructed and
changeable consequence of Americans' political choices.
Our contribution
to the debate over
and why inequality arises and
to clarify
how
our argument by
first
growing inequality
persists.
We
initiate
is
challenging the explanation in The Bell Curve, the idea that inequality
we go on
natural and fated. Then,
to
show how
social
is
environment and
conscious policy mold inequality in America. If the
growing inequality
in
America
is
not the inevitable result of free
markets operating on natural intelligence, but the aftermath of circumstances that can be altered, then different policy implications follow from
those outlined in The Bell Curve. equalities
mount; we do not have
growth; and
we do
We
do not have
to accept
them
to fatalistically let in-
as the Faustian trade for
not have to accept heartlessness as the companion of
social analysis. Instead,
we can
anticipate greater equality of opportunity
and equality of outcome and also greater economic growth.
Explaining Inequality
Why
do some Americans have a
human
follows inevitably from
talent than others; the first tion. is
Many
people accept
lot
more than others? Perhaps,
nature.
Some people
succeed while the others
this explanation, but
it
inequality
are born with
fail in life's
more
competi-
will not suffice. Inequality
not fated by nature, nor even by the "invisible hand" of the market;
it
is
a social construction, a result of our historical acts. Americans have created
and type of inequality we have, and Americans maintain it. To answer the question of what explains inequality in America, we must
the extent
divide
it
who falls behind in the competiSecond, what determines how much people get for being
in two. First,
tion for success?
who
gets ahead and
ahead or behind? To see more clearly
that the
two questions
are different,
think of a ladder that represents the ranking of affluence in a society. Question
one asks why
this
person rather than that person ended up on a higher
or lower rung. Question
two asks why some
societies have
tall
and narrow-
CHAPTER ing ladders
—ladders
people
—while other
have huge distances between top and bottom
that
rungs and that taper off
1
top so that there
at the
room
is
distance between top and bottom and with lots of
little
people
all
way
the
One
the footrace:
is
room
for
many
who wins and who and rewards of the race. Some
question
is
loses; another question is
what are the
rules
races are winner-take-all;
some award
prizes to only the
others award prizes to
The answer
few
—ladders with
to the top.
(Another metaphor
we need
stand the race,
for only a
have short and broad ladders
societies
many
finishers,
even
to understand the rules
to the question
of
who
first
few
to all participants.
finishers;
To under-
and rewards.)
ends up where
that people's social
is
environments largely influence what rung of the ladder they end up on. 6
The advantages and disadvantages
that people inherit
from
their parents,
the resources that their friends can share with them, the quantity and quality
of their schooling, and even the historical era into which they are born
boost some up and hold others down. The children of professors, our children, have substantial
head
starts
over children
Young men who graduated from high school
who
greater opportunities than the ones
of, say, factory
in the
own
workers.
booming 1950s had
graduated during the Depression.
Context matters tremendously.
The answer to is more
why
the question of
rewards
political.
societies vary in their structure of
In significant measure, societies choose the
height and breadth of their "ladders."
them, by providing services to
By
loosening markets or regulating
citizens or rationing
all
them according
to
income, by subsidizing some groups more than others, societies, through their politics, build their ladders.
To be
straints
deny
remains
(see, especially, chapters 5
full
freedom of
Americans have
the inequality
result of policy choices
tives
—have made.
tinctively unequal.
sure, historical
action, but a substantial
and
—
or, at least,
In the United States, the result
Our ladder
is,
To
see
—and becoming more
how
—
means
that
measure, the historical
Americans' representais
a society that is dis-
history, unusually
extended
so.
policies shape the structure of rewards
outcomes), consider these examples:
marketplace
this
by the standards of affluent democracies
and even by the standards of recent American and narrow
democracy,
in significant
is,
Americans
6). In a
and external con-
freedom of action
Laws
(i.e.,
the equality of
provide the ground rules for the
rules covering incorporation, patents, wages,
ditions, unionization, security transactions, taxes,
working con-
and so on.
Some laws
income and earnings among people in the market; others narrow differences. Also, many government programs affect inwiden differences
equality
more
in
directly through, for example, tax deductions, food stamps,
8
WHY INEQUALITY? and corporate subsidies. Later
social security, Medicare,
in this
book,
we
Americans have taken, or chosen
will look closely at the various initiatives
not to take, that shape inequality.
To
see
how
and which
policies also affect
the
fall to
which particular individuals get
bottom of our ladder
to the top
the equality of opportunity),
(i.e.,
consider these examples: The amount of schooling young Americans receive heavily determines the jobs they get and the income they make. In turn, educational policies
— what
school resources are distributed (usually according to the
which children strongly affect
live),
way community in
sorts of schools are provided, the
teaching methods such as tracking, and so on
how much
schooling children receive. Similarly, local em-
ployment opportunities constrain how well people can do economically.
Whether and where governments promote jobs or influence
who
Claiming equalities
tell
do so
who
that other policies could
in-
change those inequalities
a novel idea in the current ideological climate.
us that inequality
will, in turn,
is not.
have significantly constructed the
that intentional policies
we have and
may seem
fail to
poised for well-paid employment and
is
So many voices
the result of individuals' "natural" talents in a
is
"natural" market. Nature defeats any sentimental efforts by society to re-
duce inequality, they say; such
efforts should therefore
Bock wrote
be dropped as
futile
common
and comforting. As Kenneth
in his study of social philosophy,
"We have been quick to seek what we are instead of what
and wasteful. Appeals to nature are
explanations of our problems and failures in
we do. We seem wedded
to the belief that
is
a consequence of
this case,
appeals to nature
our situation
our nature rather than of our historical acts." 7 In are shortsighted.
Arguments from nature
are useless for answering the question of
what
determines the structure of rewards because that question concerns differences in equality us
why
among societies. Theories of natural
inequality cannot
tell
countries with such similar genetic stocks (and economic markets)
as the United States, Canada, England, and
Sweden can vary
so
much
degree of economic inequality their citizens experience. The answer
in the lies in
deliberate policies.
Appeals tion:
Why
genetic
cannot satisfactorily answer even the
to nature also
do some individuals get ahead and some
endowment
and white helps
helps.
Being
tall,
makes them matter
and these
traits
More important
milieux in which people grow up and
ques-
traits are totally
or partly
matter to the degree that society
—determining how much,
white skin are rewarded.
first
behind? Certainly,
slender, good-looking, healthy, male,
in the race for success,
determined genetically. But these
fall
for example,
good looks or
yet than these traits are the social
live.
CHAPTER
1
much
Realizing that intentional policies account for
of our expanding
more accurate than theories of natural inequality; it is also more optimistic. We are today more unequal than we have been in seventy years. We are more unequal than any other affluent Western nainequality
is
not only
tion. Intentional policies
could change those conditions, could reduce and
reverse our rush to a polarized society, could bring us closer to the average inequality in the West, could expand both equality of opportunity and
equality of result. Still,
the "natural inequality" viewpoint
is
a popular one. Unequal out-
comes, the best-selling Bell Curve argues, are the returns from a cess that sorts people out according to
how
Bell Curve's explanation of inequality
is
assuming
human
that
intelligent they are.
inadequate.
fair pro-
But The
The authors
err in
can be reduced to a single, fixed, and essen-
talents
they label intelligence. They err in asserting that this
tially innate skill
largely determines
how
people end up in
life.
And
trait
they err in imagining
that individual competition explains the structure of inequality in society.
In this book,
we
use The Bell Curve as a starting point for really under-
standing inequality in America.
evidence, fated
we can
see what
is
By
exploring that book's argument and
wrong with
by nature and see instead how
the viewpoint that inequality
social milieux
its
is
and social policy create
inequality.
Generations of social scientists have studied inequality. Hundreds of
books and
articles
many factors that including among of research.
have appeared
affect
in the last
decade alone examining the
who gets ahead and who falls behind in our society,
those factors intelligence. We will draw on this treasury We will also show, using the very same survey used in The Bell
Curve, that social environment
is
more important
which American becomes poor than
is
in
helping determine
"native intelligence" most gener-
we will turn to the more profound question, the why the United States has the system of inequality it
ously estimated. Then,
second question, of does.
We
will
show
that although
much
results
from purposeful, and
of
it
some
inequality results from market
—and even many aspects of market
forces,
inequality itself
alterable, policy.
The Bell Curve Controversy In late
1994 a publishing sensation burst upon America. The covers of
newsmagazines heralded a new study articles inside
and whites
in
suggested
—of
America. The
—perhaps
the definitive study, the
the differences in intelligence
between blacks
New Republic blared "Race & IQ" in enormous 10
WHY INEQUALITY? letters
—and
sold out
rack copies in Harvard Square. Newsweek's cover
all
man and
featured facial profiles of a black
back with the superimposed words "IQ.
New Book
Controversial
&
on Race, Class
Those who went beyond the
man
a white
Is It
standing back-to-
A
Hard Look
at a
book claiming
that
Destiny?
Success."
front covers read of a
blacks are not as smart as whites, most likely because the two groups'
genes
More
differ.
broadly, they read that intelligence
by nature unequally inequalities
inequality
best
among Americans. The
is
natural, then
wrongheaded and
at
political implications
in the first case; the
for National Public
as
was
it
publicized. Both The
New
it
is at
New
Repub-
York Times Magazine published a coverthat
he
is
a boor; an interviewer
Radio delivered almost every question
with a clear note of skepticism; the editorials against the
clear: If
their reports with critical sidebars, over a
one of the authors implying
story profile of
were
worst destructive.
and Newsweek bracketed
dozen
a gift distributed
governmental intervention to moderate
The book was attacked even lic
is
conception and that this distribution explains the
at
New
book; and so on.
to that author at least
two
book withstood
the
York Times published
And
yet the
attacks and sold hundreds of thousands of hardcover copies (perhaps a sales record for a
book with dozens of pages of
The Bell Curve, by Richard
ogy professor
at
statistical tables).
Herrnstein (who had long been a psychol-
J.
Harvard University
at his
untimely death shortly before
the book's publication) and Charles
Murray
well-known conservative
and resident
more
tanks), is
stance its
is
is
essayist,
substantial than
due not merely
to its
imposing array of graphs, a philosophy ages old:
demption; inequality talents,
is
Ph.D.
in political science,
at
conservative think
media representations suggest.
its
Its
sub-
mass, about 850 pages cover-to-cover, nor to tables, footnotes,
Human
fated;
(a
misery
is
and people
and references. At
its
base
and beyond human
natural
re-
deserve, by virtue of their native
the positions they have in society.
From
that ideological base,
Herrnstein and Murray build a case that critics cannot simply dismiss out
of hand. Herrnstein and Murray argue
—
relying on their
own
national survey, supplemented by an array of citations
analysis of a large
—
that individuals'
intelligence largely decides their life outcomes. Intelligence
unequally a
among
few people
distributed
people, in a distribution shaped like a "bell curve" with
at the
people clumped
is
lower end, a few people
in the
middle.
A
at the
upper end, and most
person's position in that distribution
heavily influences his or her position in the other distributions of
life
—
the
distributions of jobs, income, marriage, criminality, and the like.
The centerpiece of Herrnstein and Murray's evidence 11
is
the National
CHAPTER (NLS Y),
Longitudinal Survey of Youth
1
a massive survey of over ten thou-
sand young Americans involving repeated interviews over more than a de-
The
cade.
NLSY administered
Armed
the
to its subjects in 1980. Herrnstein
who
scored high on that
poorly. This
is
(AFQT)
NLSY
subjects
which the authors
test,
usually doing well ten years
Forces Qualifying Test
and Murray show
and those
later,
that
treat as
an "IQ"
who had low
test,
proof, they argue, that intelligence largely determines life
outcomes. Herrnstein and Murray also contend that intelligence tially fixed,
unchangeable
in
therefore also unchangeable. forts to alter this naturally
any significant fashion. People's
And
who have
not read
The Bell Curve
is
so must be social inequality
is
essen-
fates are
Ef-
itself.
unequal order waste money and undermine
efficiency and justice. (Appendix
those
were
scores ended up
summarizes The Bell Curve
1
its
in detail for
it.)
an inadequate explanation of where individual Ameri-
cans end up in the system of inequality.
some people end up higher than
others
Its
answer
to the question of
why
on the ladder of success vastly over-
estimates the relative importance of aptitude tests and underestimates the
importance of the social environment. Despite Herrnstein and Murray's self-congratulations that, in examining intelligence, they have dared to go
where no
social scientist has
that scores
on IQ and IQ-like
gone before, scholars long ago established tests
were only of modest importance com-
pared with social context in explaining individual attainment.
We reinforce
and expand that familiar conclusion by redoing Herrnstein and Murray's analysis of the
NLSY
the
AFQT is
largely a
We show that they made major errors that AFQT relative to social factors. For example,
survey.
exaggerated the role of the
measure of
Newsweek cover could
instruction, not native intelligence.
just as well read "Grades.
NLSY
Moreover, a correct analysis of the score
is
only one factor
among
these factors, the social ones are
More
how
more important than
AFQT
well people do; of
the test score.
fundamentally, The Bell Curve also provides an inadequate under-
standing of systems of inequality. the
Are They Destiny?")
survey reveals that the
several that predict
(The
American ladder
is
so
tall
Its
implied answer to the question of why
and narrow
natural market. This interpretation
is
is
that natural talent prevails in a
wrong,
in part
because
it
is histori-
cally naive. For example, during most of this century Americans became
substantially
more equal economically, but
since 1973 they have
become
substantially less equal. Understanding such fluctuations in inequality re-
quires a broader historical and international perspective than The Bell
Curve provides.
Why
We
try to provide
do we pay so much
such a broader perspective.
attention to
12
The Bell Curve? Some colleagues
WHY INEQUALITY^ The Bell Curve
told us that
is
so patently wrongheaded that
would be
it
quickly dismissed; that genetic explanations of inequality are old news,
having gained notoriety and disrepute
and
we would
earlier; that
we
so forth. But
go away.
The Bell Curve
felt that
seventy years ago,
thirty years ago,
only further publicize The Bell Curve; and not easily ignored.
is
It
will not
ideas and data, at least as transmitted by the media and by poli-
Its
many
provide a touchstone in policy debates for
ticians, will
Our
years.
Berkeley colleague Troy Duster notes that within weeks of The Bell Curve's publication, Charles Murray had been invited to address the newly elected Republicans in the
House of Representatives and
that an article in
The Chronicle of Philanthropy had speculated that charity for "people of 8 Shortly afterward, the president of lesser ability" might be a waste. Rutgers University faced an uproar when he apparently alluded to The Bell
Curve
in explaining
problems of black students.
Also, The Bell Curve's perspective on society, which reduces a complex
more than
reality to little
a footrace
among unequally
offends us as social scientists. Social reality set
—
swift individuals,
for example,
up the "race" and how they reward the runners
how
societies
—cannot be understood
through such reductionist thinking.
Nor were we
satisfied with the critical appraisals that
we undertook this
Some
project in late 1994.
had appeared when
reviewers, even as they casti-
gated The Bell Curve, accepted, or were perhaps intimidated by, tific
presentation.
Some
its
attacked the authors, the authors' funders, or the
authors' intellectual friends. Deserved or not, such attacks
Herrnstein and Murray's claims.
do not invalidate
Some commentators seemed
to be grasp-
ing at straws, picking one or two contrary studies reported in the
without noting that the authors had piled on
And some just admonished The
arguments. plications.
We
scien-
many
book
others to support their
Bell Curve for
its
political
im-
believed that the book deserved neither the deference nor
the unfair attacks.
It
could be challenged on scientific grounds. Also,
in
responding, critics generally accepted Herrnstein and Murray's framing of
why some people finish first and others last. 9 We do not. As academics, we have the impulse to contest every claim and statistic the 850 pages of The Bell Curve. There are certainly many errors and
the question:
in
10
contradictions in the details. dress:
What
is
ronment play set as
it
is?
intelligence? in
What
shaping
However, there are more basic issues
What
life
role
outcomes?
difference does policy
issues, the particular statistics usually history.
We
will
do individual
show
that
Why
to ad-
and social envi-
the structure of
outcomes
make? For resolving many of these
do not matter
The Bell Curve 13
is
talent
is
wrong
as
much
as logic and
statistically, that
it
is
CHAPTER
1
even more profoundly wrong logically and
historically,
and
that its impli-
cations are destructive.
One
statistic is
worth noting right away because
it
shows
The Bell Curve than some intimidated reviewers have
less to
that there is
realized: "ex-
plained variance." Near the end of their text, Herrnstein and Murray capsulize their
argument by asserting
on how people do that
AFQT scores,
their
powerful bearing
that "intelligence has a
in life" (p. 527).
However, 410 pages
measure of IQ, explain "usually
cent and often less than five percent" of the variance in
admit
earlier they
less than ten per-
how
people do in
What does "explained variance" mean? It refers to the amount variation in some outcome, like income, from zero to 100 percent,
life (p. 117).
of the that
can be explained by a particular cause or
set
of causes. To state that
intelligence explains 10 percent of the variance in, say, people's earnings is to
say that intelligence accounts for 10 percent of the differences
people in earnings, leaving 90 percent of the differences
unaccounted
for.
By
among
own statistical estimate, outcomes among respondents
Herrnstein and Murray's
5 to 10 percent of the differences in
life
odds that they became poor, criminal, unwed mothers, and so on accounted for by differences among them way,
90
if
to
we
95 percent of the inequality
The
1
we
see today.
What
F-axis represents the proportion of
shows
that virtually
2 percent
is,
we would
that
still
in the
United
in
to $150,000.
in 1993;
—had incomes of $25,000, about .01(1
shape of inequality
hardly changed. Because
line displays the actual distribu-
household income. The dashed if
line dis-
every adult in the
AFQT score accounts for, at best, only
of the variation in earnings, 11
it
solid
about
percent) had
United States had had identical intelligence as measured by the
for.
see
shown
American households. The
plays what that distribution would have looked like
counted
the
is
means
income
no households had zero income
incomes of $75,000; and so on. The solid tion or the
— —can be
scores. Put another
bottom are the incomes from zero
line
that
only
.2.
The
—
AFQT
figure displays the distribution of household
States in 1993. Across the
.02
in
could magically give everyone identical IQs,
graphically in figure
among earners
AFQT:
10 percent
leaves 90 percent of the variation unac-
In sum, intelligence, at least as
measured by the AFQT,
is
of
such minor importance that American income inequality would hardly
change even
if
everyone had the same
AFQT score.
(In a response to simi-
Murray backed away from explained variance as a criterion judging the importance of intelligence, but The Bell Curve argument
lar criticisms,
for
depends on
that criterion.)
12
As some economists have noted for policy
is
in
reviewing The Bell Curve,
neither total explained variance nor even whether
14
13,
the issue
it is
intelli-
WHY INEQUALITY
If all adults had the same test scores (but different family origins and environments), inequality of household
incomes would decrease by about 10 percent.
% of households with income close to average remains about the same
t
a
4 Household income,
V
U.S., 1993
*
All adults have the same AFQT
A
4
*P
o §\000 $25,000
$50,000
$125,000
$100,000
$75,000
$150,000
Household Income Level
1.2.
Explained Variance
gence (Note: See
in
Household Income Accounted
for
by
Intelli-
and notes for method of calculation)
text
gence or the social environment that explains more of the variation vidual outcomes.
It is
contribution to outcomes. In asserting that cognitive ability
determining individuals' fortunes but
Murray argue
that
in indi-
whether a given intervention can make a positive net
is
no intervention can pay
is critical
to
unchangeable, Herrnstein and off.
We
will see,
however,
that
cognitive abilities are malleable (chapters 2 and 7). In asserting that socio-
economic background
is
of
ual outcomes, Herrnstein and
importance
trivial
Murray
conditions cannot be effective.
We
influencing individ-
in
are claiming that
will see,
working on social
however, that socioeconomic
conditions matter a great deal, so that policy there can be effective in in-
creasing opportunity (see chapter 4).
Murray do not consider
the deeper
More important ways
yet, Herrnstein
that social policy
and
shapes both
individual competition and the structure of inequality (see chapters 5 and 6).
There
is
great leverage for policy there, as well.
15
CHAPTER The claim
1
is
the central
argument of The Bell Curve. But many
discussions in The Bell Curve wander from that argument. distraction
the discussion of ethnicity and IQ.
is
the argument over intelligence and inequality
is
The major such
a distraction because
It is
unchanged whether or not
Murray has
there are inherent racial differences in intelligence. Charles
admitted
end, whether genes or environment explain racial dif-
that, in the
IQ scores "doesn't much matter"
ferences in
on the
that intelligence accounts for individuals' locations
ladder of inequality
(italics in original).
We
14
agree (although the genes versus environment debate matters a great deal if
we want
topic of race
issue (see chapter 8).
and IQ so
we must
centrally,
But otherwise we intend
we
discussion. Finally,
some ways Herrnstein and Murray policy suggestions.) For example,
become
address the
on the main
to stay
line of
agree with Herrnstein and Murray on some matters.
(Secondhand readers of The Bell Curve may be surprised
since 1970
Because the
to explain racial differences in life circumstances).
media featured the
to learn that in
are not always conservative in their
we
agree with them that Americans have
increasingly polarized between rich and poor, and
we
agree with them that a guaranteed annual income ought to be considered as a possible national policy.
We raise
15
several arguments against The Bell Curve, any one of which
sufficient to dismiss
it.
If intelligence is not single, unitary,
intelligence can be altered; if test scores
gence
is
mismeasure
not the major cause of people's fortunes;
reward intelligence;
if
we respond
in detail to
intelligence; if intelli-
markets do not
ment and of
its
—
if
it
affords us
we
see in
stresses the importance of social environ-
policies that construct the social environment. That under-
standing, in turn, begins a realistic discussion of
and
fairly
fails.
The Bell Curve because
an opportunity to explain what does account for the inequality
America today. That explanation
is
fixed; if
patterns of inequality are socially constructed
any of these arguments holds, The Bell Curve case In the end,
if
and
harmful
how
to
reduce inequality
effects.
Overview of the Argument If
one asks why some people get ahead and some people
fall
behind, an-
swers concerning natural differences in ability are woefully inadequate.
We can
see that by looking closely at "intelligence."
in intelligence is a abilities are
One
poor explanation of class inequality
much more complex,
is
reason inequality that individuals'
variable, and changeable than
16
is
sug-
WHY INEQUALITY
1
gested by the old-fashioned notions of intelligence upon which The Bell
Cur;e
rests.
Concretely, the basic measure of intelligence that Herrnstein
and Murray use, the quick-wittedness.
AFQT,
It is
instead a test of what people have been taught, es-
pecially in high school, of
they
make
in the test.
actually not a test of genetic capacity or of
is
how much
they recall, and of
Another reason
how much
explanation of individual success or failure
effort
an adequate
that intelligence is not
as social scientists
is that,
have
known for decades, intelligence as measured by such tests is only one among many factors that affect individuals' success or failure. In the NLSY, respondents' AFQT scores in 1980 do not explain well how they ended up at the end of the 1980s. We show that, instead, aspects of respondents' social environments explain the If
outcomes more
fully.
one asks the more basic question of what determines the pattern of
inequality,
answers concerning individual intelligence are largely
vant. Societies
and
historical
epochs vary greatly
in the nature
irrele-
and degree
much more than any variations in intelligence, or the market, can account for. Some of that variability lies in technological, economic, and cultural changes. But much of it lies in specific of their inequality; they differ
policies concerning matters such as schooling, jobs, and taxes.
we can change
In the end,
American
inequality.
have reduced inequality
policies
We in
improving the economic fortunes of the elderly equality in others
—
of the already advantaged.
shows
that there is
do
to
much more
for
in-
that
And
the experience of other nations
can be done
to
reduce inequality
if
we
so.
Policies also affect ity.
example, spheres — —and have expanded
for example, with tax expenditures that advantage
many
choose
have changed inequality.
many
where individuals end up on the ladder of inequal-
Policies help construct social environments. Policies even alter cog-
nitive skills, particularly in the
age here
lies
ways we
structure schooling.
The
lever-
not with the episodic compensatory programs over which
there has been
much
debate, but with the everyday structure of schools in
America. Finally,
what about race? Arguments
Americans have done poorly intelligent than whites are
in
Americans and Latino
United States because they are less
completely backward. The experiences of low-
caste groups around the world
worse
in the
that African
show
that subordinate ethnic minorities
do
schools and on school tests than do dominant groups, whatever
the genetic differences or similarities
European Jews
in
1910
New
between them. Whether
York, the Irish in
or Afrikaaners in South Africa, being of lower caste or status
17
it
is
England, Koreans
Eastern
in Japan,
makes people
CHAPTER seem "dumb." The States
fits
status;
1
particular history of blacks and
the general pattern. It
is
Mexicans
in the
United
not that low intelligence leads to inferior
that inferior status leads to low intelligence test scores.
it is
The Plan of This Book Chapter 2 examines
Murray draw
how
the psychometricians
for their psychology
have sought
upon
is
is
themselves:
tests
Intelli-
the statistical core (labeled "g") of those tests. In other words,
intelligence nition
The
to study "intelligence."
psychometric concept arises largely from the IQ
gence
whom Herrnstein and
what IQ
is
by showing
tests
We show how problematic that defiAFQT largely measures how much math and
measure.
that the
English curricula teenagers have learned and display. But there are other, better
ways
to think about intelligence.
We discuss
formation-processing" perspective, one which
is
as an
more
example the
Chapter 3 examines The Bell Curve's specific evidence about
AFQT. Scores on
gence: scores on the
"in-
realistic. intelli-
school achievement tests are, of
course, important in a society that rewards people according to
how
well
they do in school, but they are not what most people would consider as
We
"intelligence" per se.
"massage" the ity
and
utility
ways Herrnstein and Murray arguments. They overstate the valid-
also explore the
AFQT data to fit their AFQT scores. Yet
of the
degree that such
to the
measure how well we educate our children,
outcomes
testifies to
how
educational policy
critical
test scores
their ability to predict life is
for
American
in-
equality.
Chapter 4 addresses the analyses purporting to
fattest section
show
that
NLSY
of The Bell Curve, respondents'
— and presumably, respondents' mines — what becomes of them. We review
predict
the
so,
its statistical
AFQT
intelligence
scores best
most
deter-
critical errors Herrnstein and
Murray made
come
—
in their analysis;
we
as other scholars have, also
ronment
is
more, not
less,
reanalyze the identical data; and
—
we
to opposite conclusions: Social envi-
important than
test
scores in explaining poverty,
likelihood of incarceration, and likelihood of having a child out of wedlock.
For economic outcomes, gender, a
nored, matters most of
conditions
—
specific data, tists.
We
all.
trait
Other social factors
Herrnstein and Murray ig-
—education and community
are at least as important as test scores. Stepping back
we
point out that these findings are not
have long understood
that a person's
news
economic fortunes are hos-
tage to his or her gender, parents' assets, schooling, marital status,
18
from the
to social scien-
commu-
WHY INEQUALITY? nity's
economy, stage
one item on such a
and so on; intelligence
in the business cycle,
list.
This chapter
settles the issue
of
why some
is
what
get ahead of others in the race for success; the next chapter looks at the racers
win or
lose.
Chapter 5 turns attention
not whether individuals are
how how
will see
particularly
how extreme nations.
The
systems of inequality, showing
to
they vary across history and
We
just
people
among
nations.
more or
less equal, but
American
widened since the 1970s. And we
inequalities
greatly
whether societies
the degree of inequality fluctuated in
5
is
are.
history,
will see
compared with other advanced industrial America today is not "natural" but in great mea-
the United States
inequality in
how
The question of chapter
is
sure the result of policies that tolerate wide inequalities. Ironically, those policies are, despite assertions
by interested
parties, not necessary for eco-
nomic growth; indeed, inequality may well retard economic growth. Chapter 6 turns to several explicit national policies that structure equality in America.
Some
policies
social security, Medicare, food stamps, etc. rate subsidies, the
so on. little
—while some widen —corpoit
mortgage deduction, laws concerning unionization, and
Compared with America's economic
to equalize people's
economic fortunes
competitors,
a result of the policies
Americans have
Chapter 7 turns to policies that shape individual
economic inequality
—
relatively
or even their economic op-
at least tacitly
intelligence. Individuals' cognitive skills
we do
—
portunities. This explains our charge that the inequality
that determine
in-
and programs narrow inequality
Americans have
is
chosen.
abilities, specifically,
—those supposedly
fixed talents
are indeed changeable.
We
show,
using the examples of the school year, tracking in schools, and the structure
of jobs, that learning environments alter
how and how
Policies help construct those learning environments.
well people think.
Even
the inequality of
ability is subject to social shaping.
Chapter 8 turns to race and ethnicity tion, albeit
—
a topic
we
on standardized
was
a distrac-
Why
do blacks and La-
tests? This turns out to
be not a biological
an incendiary one, in The Bell Curve.
tinos score lower
believe
question but a social one. Around the world,
members of disadvantaged
groups usually score lower than members of advantaged groups, whatever their racial identities. In
many
cases, both the higher- and lower-status
groups are of the same race. Also hard to reconcile with the point
is
the
way
ethnic groups seemingly
become smarter
racialist
view-
after they have
succeeded. For example, in Japan Koreans are "dull," while in the United States
Koreans are "bright"; Jews
years ago but are
among
in
America were "dull" seventy-five
the "cognitive elite" today.
19
We
describe three
CHAPTER ways
1
that ethnic subordination in a caste or castelike
school and
test
system leads to poor
performance: One, subordination means material depriva-
which
tion for students,
in turn impairs their achievement; two, subordi-
nation usually involves group segregation and concentration, which, by
multiplying disadvantage and drawing
all
group members into
difficult
learning situations, undercuts academic achievement; and three, subordination produces a stigmatized identity of inferiority,
which
in turn breeds
resignation or rebellion, both of which limit academic achievement. histories of African
rent conditions,
Americans and Latino Americans,
more than
suffice to explain
lower than whites on
to score
race for success.
tests
why
fits
members
their
and also why they do
The American case
The
as well as their cur-
tend
less well in the
the global pattern;
it
is
not
genes but caste positions that explain the apparent differences in cognitive performance.
Chapter 9 concludes with a consideration of what the intellectual and the practical implications are of understanding inequality in these historical
and sociological ways.
Concluding Comments by Way of Introduction
A comment
on the "burden of proof:
to contradictory studies about
may
how
Many
by now accustomed
readers,
certain foods
do or do not cause heart
among dueling Ph.D.s' claims many specific issues of evidence in The Bell Curve. But more important is how the basic questions are framed and the historical breadth of evidence examined. From such a fundamental perspective, we find that intelligence, broadly understood, does affect Americans' fates but is just one factor among many. It is not the key disease or cancer,
about inequality. In
feel
this
unable to decide
American inequality nor
to
ences
among
we
book,
to
contest
American
social problems; indeed, differ-
individuals altogether are not the key.
together as citizens, choose to structure our society.
The key
We
do
is
how we,
not, of course,
have unlimited freedom of action; we are constrained by material circumstances, social traditions, and political institutions. But
freedom
book
will show, than admitted
a lot
more
by those who coun-
acceptance of the growing inequalities in our society. The challenge
sel
to
to act, this will
we have
make
In thinking about those choices,
This nation draws tions.
is
those choices.
The
its
it
may
help to go back to
moral precepts from
its
biblical
first
principles.
and republican
tradi-
Bible repeatedly enjoins us to help the needy; the Declaration of
20
WHY INEQUALITY? Independence announces that
dowed by
should presume that ises.
"all
men
are created equal, that they are en-
their Creator with certain inalienable Rights." 16
The burden
is
have us sorted out proof is on those
its
people
on those
come
who would
at birth into the
Such
a nation
fulfill
those prom-
contend otherwise,
who would
equally equipped to
worthy and the unworthy. The burden of
who would contend that some
of us are hopeless and fated
only for piteous charity. Absent conclusive proof of that claim, Americans
should assume an equality of worth and can's horizon.
21
move
to
expanding every Ameri-
CHAPTER
*
*
2
Understanding "Intelligence" "When tone, "it
/ use a
means
word," Humpty Dumpty
just
what
I
choose
it
to
said, in a rather scornful
mean
—
more nor
neither
less."
"The question
mean
so
many
different things."
"The question ter
—
"whether you can make words
is," said Alice,
is," said
Humpty Dumpty, "which
—Lewis he word here
Carroll,
mas-
Through the Looking Glass
"intelligence." For those
is
who
believe that inequality
of talent explains inequality of fortunes, intelligence talent of
to be
is
that's all."
them
all.
For those
who
the immutability of intelligence
believe that talent
dooms any
comes. But arguments that intelligence
way of understanding
most important
the
largely fixed at birth,
effort to alter inequality of out-
immutable
rest
on a particular
approach called psychometrics (the
intelligence, an
measurement of mental
is
is
is
we will show that this outhow complex and flexible people's cogniunderestimates how much such skills can be
traits).
In this chapter,
dated approach underestimates tive skills are
and thereby
improved. Other,
newer schools of psychology
offer better approaches to under-
standing intelligence and also offer sensible hope for improving Americans' abilities.
They show us
cating people,
young and
more more
acutely, to
make more of
intelligent population
did generations ago.
We
we can beneficially invest more in eduwe can train people to use their minds their abilities; and that we can produce a
that
old; that
—just as the establishment of mass education one example of such alternatives
will take, as
psychometrics, the information-processing perspective.
It
psychometrics but extends our knowledge of intelligence Contrasting psychological schools
differences are akin to the difference between seeing the
orb in the sky and seeing lieve that the
always aim
moon
at the
is
it
fixed,
wrong
as an orbiting
companion of
you could never land on
place. If
you
encompasses
much
further.
not a mere academic exercise.
is
realize that
it
it is
to
moon
The
as a fixed
Earth. If
you be-
because you would
in
motion, the
moon
can become the point of departure for exploration of the cosmos. So with
22
UNDERSTANDING "INTELLIGENCE" understandings of intelligence:
we can
target our efforts
When we
understand
on expanding and enriching
how it.
malleable
To
is,
it
the degree that
people's cognitive skills affect their chances of success, then understanding
how
malleable those skills are underlines
how much
equality of oppor-
tunity can be changed.
The Bell Curve, Herrnstein and Murray imply
In
no
that
self-respect-
ing psychologist questions the psychometric perspective. But not only do
many
established psychologists question
come from
into intelligence
it,
most of the important insights
perspectives that directly contradict funda-
mental claims of psychometricians.
1
we
In this chapter,
ferences between the schools, showing
how
describe the dif-
limited psychometrics
is
for
understanding the role of intelligence in inequality.
Psychometrics,
how
we
will show, has not
people think or solve problems.
It
been centrally concerned with
has been concerned instead with
developing tools to rank, to differentiate, people by
how
successfully they
solve academic problems. Psychometricians have spent decades refining
minutely graded
tests so that they yield scores for individuals that are reli-
able and that correlate with success on other tasks, such as progress in
school and in the job market. The specific problems that comprise a less
important to psychometricians than that the
ple.
Among
the difficulties with this approach
test is
"works"
test are
to rank peo-
that the test items psy-
chometricians typically use deal with school subjects or are school-like,
such as math questions. They thereby confound intelligence with schooling.
(We
will see this in detail below.)
Another profound
difficulty is that
the psychometric effort to rank people requires that the tests produce fine distinctions, differences in scores that correspond
among people problem
is
in
how
the psychometricians' insistence
rather than their observations of people intelligence.
—
driven by their techniques
a singular and basic
Other approaches to intelligence avoid these and similar
and they do so by studying thinking
complex
be easily absorbed into psychometric
Before fleshing out these points,
made
Yet another
life.
—
that there is
difficulties,
to
to differences
little
they use intelligence in everyday
in science.
we need
itself,
typically in
ways too
tests.
to understand
how
Otherwise, the psychometric school and
its
progress
alternatives
will simply appear to be equally valid points of view. Therefore,
discuss
is
we
first
how one research framework in science gives way to another and we examine psychometrics and demonstrate its fundamental
why. Second, circularity:
Psychometricians have discovered a single, fixed "intelligence"
because they have developed a methodology that works, and ology works because
it
implicitly
assumes 23
that
method-
that there is a single intelli-
CHAPTER gence, essentially fixed native
at birth
ways of understanding
2
or early in
we
Third,
life.
point to alter-
intelligence, looking in particular at the in-
formation-processing framework, which understands intelligence to be changeable.
The Importance of "Paradigms" The
stereotype of the scientist as a lonely inquisitive soul
is false.
communication among scholars, each one would be forced scratch,
Without
to start
from
and accumulating knowledge would be impossible. To communi-
cate, researchers
common
use a
"language" of concepts
any language, paradigms impose
a paradigm. Like
limitations. Crucially,
Some
the questions that researchers can ask.
paradigm and others make no sense
to a
2
in that
the Flat-Earth paradigm, questions about
paradigms limit
research questions are critical
paradigm. For example, in
what happens
if
you
sail
off the
edge of the Earth make sense, while questions about the velocity needed
to
reach orbit are nonsense. In the Spherical-Earth paradigm, the reverse
is
Because the
true.
digms
to delimit
answers
of
list
all
our inquiries. Paradigms also
to those questions
One way paradigms cepts; another is
possible questions
should look
by suggesting how
alert researchers to
para-
what the
like.
and answers
limit questions
we need
is infinite,
to observe, or
is
by defining key con-
measure, those concepts.
Like the questions, empirical observations typically make sense only within a paradigm. For example, precise measurements of
bumps on peo-
heads made sense for the nineteenth-century "science" of phrenology make no sense to twentieth-century psychologists. It is a paradigm that
ple's
but
gives any particular observation
have
scientific
its
meaning.
meaning only within
We will see that IQ test scores
the perspective, or paradigm, of psy-
chometrics. (They have, of course, considerable practical meaning as well.)
New digm
paradigms
loses steam
scholars find
it
when
arise is
old paradigms lose steam.
boring and "tired." Another
way
is
is
that
new
later.)
—
paradigm that
it
the
When in
phenomenal
rise
(We in IQ
will discuss
an ad hoc way. Eventually, so to
third route to
an example
in
psycho-
scores over the twentieth century
apparent anomalies arise, scholars
no longer seems
A
new
research findings appear that cannot be as-
similated by the old paradigm.
metrics
a para-
that scholars raise
questions that cannot be addressed in the old paradigm.
paradigm change
One way
by running out of stimulating questions; younger
first try
many ad hoc
to patch
up the old
patches are applied
be an elegant blueprint for research but instead
24
UNDERSTANDING "INTELLIGENCE" seems
to
be a hodgepodge of poorly understood findings. This
what happened
to the astronomical
motion
that the
perhaps fanciful story
may
Suppose
commonly
hair,
Among
at flat
is
a critical concept.
of a standing
human
question people ask tall
will
because answers to
when planning
you be?"
it tell
It is
at
"How Some
persist in asking the old question,
now
very confusing.
total
with
their
tall
question, as they
know
a small
plaza, she
way
will
to the plaza,
many
you be?" The answers,
not what to report given the chair in
The old concept no longer
Some
Others argue that the concept Still
row of the
people cling to the old concept
fits
the
alter the
the vertical distance that a person takes
it.
this society
height of their body and accoutrements. Others stare
Eventually, arguments ensue.
ing.
A
her reduced "height." Others soon see the gain in sitting,
and eventually chairs abound. Upon making
their hands.
wearing.
young woman fashions
amazes many
and report the
They come
to attend a concert, speech, or
to the gathering, and, sitting in the front
at the
at
people where they should stand.
day, an enterprising (and tired)
however, are
is
an important question in
bench to bring
mean
human
being, plus what-
headgear, high heels, or other footwear the person
"How
blankly
to
community of people who
outdoor plazas, the performers are
these people, "height"
critical
One
suggest, in such a state.
power paradigms have
and the members of the audience stand throughout each per-
play
is
we
deals with the concept of
It
to define height as the vertical length
ever
roughly
Sun and
attend public events such as concerts, plays, and speeches.
level,
formance.
is,
that there exists an isolated
These performances are given ground
based
is
illustrate the
shape questions and observations. height.
is
that the
it
metrics upon which The Bell Curve
A
assumed
that
became so complicated trying to account paradigm fell apart. The version of psycho-
planets revolved around Earth; for planetary
paradigm
is
new
reality.
concept of
up whether
human
height to
sitting or stand-
useless and call for society to dispense
others try to resurrect the concept by correlating the old mea-
sures of height to the vertical distance of people sitting at the plaza. Finding a high correlation
between these two measures, the
call
goes forth to hold
ways of thinking. As we leave this community to its problem, we may draw a few (1) The concept (height) arose from important problems in that
on
to the old
(2) reality (i.e., the
society;
dimensions of space) limited the definitions of the con-
cept that could be devised, but (3) the old
lessons:
it
did not determine a unique definition;
concept became problematic because
it
no longer answered the
questions that society needed answered; and (4) the old concept could be
under a new framework but would likely not have been central to that
25
fit
new
CHAPTER framework. The basic point prior assumptions;
2
that understandings of "height" rested
is
on the
could have different meanings and different measure-
it
ments. So, too, with intelligence.
We will focus on two paradigms for thinking about intelligence: psychometrics and information processing. We will see how each defines "intelhow
ligence,"
only a
fatalistic
intelligence,
and what
toward intelligence. Because
framed perspective on
a narrowly
measure
practitioners try to
definition implies about policy
it is
their
trapped in
intelligence, psychometrics can provide
view of the opportunities for change.
It
should become
apparent that newer approaches, such as information processing, allow us to break out of the dated confines of psychometrics, allowing us to see
we can
raise intelligence
how
and modify inequality.
Intelligence in the Psychometric School The psychometric paradigm assumes critical to
human
functioning
is
to determination, self-discipline,
myriad of human
abilities that
that the
"intelligence."
empathy,
we
fundamental
skill
or talent
gives no or
little
attention
It
creativity,
charm, energy, or a
recognize in people during our daily
Occasionally, a psychometrician suggests (as Herrnstein and Murray
lives.
traits result from intelligence. Famed psychometriEdward Thorndike once wrote that "the abler persons in the world the more clean, decent, just, and kind." 3 But for the most part the talent
do) that other admirable cian are
.
of interest
is
.
.
intelligence.
Psychometricians' descriptions of intelligence arose from their efforts to sort
and rank people.
of psychometrics
is
Assumptions such as chometrics
the foundational, paradigmatic assumptions
this
must rank
in a "bell curve."
one severely limit the approach, rendering psy-
at best irrelevant
inequality. Let us see
In
Among
the premise that people
why
and
at
worst a hindrance to understanding
this is so.
The Bell Curve, Herrnstein and Murray assume
that there is "a general
capacity for inferring and applying relationships drawn from experience,"
which (p. 4;
is
synonymous with
a "person's capacity for
complex mental work"
emphasis added). This basic, singular capacity
cians label g, or general intelligence. Herrnstein and
what psychometriMurray contend that
is
psychometricians have accepted the reality of g, but even a cursory scan of the psychometric literature reveals contentious disagreement
among them
about the existence or importance of g. Psychometric positions range from the idea of a unitary g, through
Raymond 26
Cattell's distinction
between
UNDERSTANDING "INTELLIGENCE" The History of Testing and the Limits of That History Critics of psychometrics often stress the political history of
the United States and the ideological
They
tricians.*
testing in
commitments of many psychome-
IQ
highlight the use of
IQ
mid-twentieth
tests in the early to
century to promote eugenics, restrict immigration, and defend segregation. (Several prominent psychometricians explicitly allied themselves with
coarse racists and Nazi sympathizers.) Based on this history,
many
dismiss
testing.
We do
not take this approach, for
chometric testing in other countries decisions people tests
made about how
two reasons. shows
that
First, the history
it is
to use the tests. In the
United States, IQ
were used for discriminatory purposes, but elsewhere,
Britain, liberal reformers
children and provide
used similar
tests to find
of psy-
really a history of the
Great
like
promising lower-class
them with opportunities.** Second, IQ and similar
widely used by educators and employers as gatekeeping mecha-
tests are
who will or will
nisms to determine
not obtain scarce positions.
It is
highly
unlikely that such tests will be discarded simply because people used
badly in the past. Therefore, ligence testing and * See, for
its
we need
instead to
examine the
them
logic of intel-
limitations.
example, Huston, Testing Testing; Kamin, "The Pioneers of IQ Testing";
and Politics of Racial Research. ** See Wooldridge, "Bell Curve Liberals."
Tucker, The Science
crystallized
and
120 components of intelligence.
No
less
Arthur Jensen, educational psychologist be boiled as
many
down
to
one
entity;
as four independent
We review
way
fluid intelligence, all the
to
Joy Guilford's notion of
prominent a psychometrician than at
Berkeley, disagrees that g can
he suggests that psychometric g
components.
may
reflect
4
the chain of reasoning that leads to the psychometric notion
of intelligence.
We
will see that, in practice, the definition of intelligence
as "a general capacity for inferring and applying relationships" does
not match the procedures psychometricians actually use. fore, are not
The
tests, there-
good measures of what most people would consider
to
be
"intelligence."
We gence
begin with Arthur Jensen's definition: "intelligence tests
5
measure" (emphasis added). This 27
is
is
what
intelli-
an honest and telling state-
CHAPTER
2
ment. In practice, psychometricians have defined intelligence after the fact: after constructing intelligence tests, obtaining the results,
what those
results
mean. 6 So, what then
Measuring Intelligence
is
an intelligence
and interpreting test?
Psychometric Tradition
in the
For psychometricians, virtually any task can be an intelligence intelligence presumably determines success tests are better erties. ability.
first
should have. Specificity
example,
because
test
virtually all tasks.
But some
than others, and the best intelligence tests have certain prop-
These properties
The
on
It is
two
are: (1) specificity; (2) stability;
are properties that any
the third one, differentiability, that
means
we want to
good
that a test
test will test
that also reflected their agility.
7
To be well
(3) differenti-
is
problematic.
only one kind of
we would
people's strength,
and
good measure of anything skill. If,
for
not want to use a test
constructed, therefore, an intel-
ligence test must reflect intelligence only, and
it
must not capture other
potentially important determinants of task success, such as motivation, creativity,
and anxiety. (This will turn out
The Bell Curve.)
means
Stability
to
that if
be a problem for the
we
test
administer the same
used in
test twice,
We
would be puzzled if a football player lifted 500 pounds one day and was unable to do so five days later. We know that the measuring instrument, the 500-pound barbell, has not changed, which is another way of saying that lifting the barbell is a stable test. Simithe measure should not change.
larly,
repeated applications of an intelligence test should yield a consistent
score for each test taker unless that test taker's intelligence changes in the interim. (In reality, scores as the college Scholastic
on many
bution of randomness. That for example, sixty points definitive.)
8
common
tests
Assessment Test [SAT], is
of cognitive
skills,
such
reflect a notable contri-
why modest differences in such test scores
on the verbal
Specificity and stability are
SAT
— should not be considered
good properties
for all
measuring
instruments, including yardsticks, thermometers, altimeters, and more.
The
psychometric
third property of a well-constructed
test, differenti-
good property of a measuring instrument. For example, we would not throw away a tape measure because it cannot help
ability, is not necessarily a
us discern tiny differences in height
Each player has a
specific
if we we must throw away measures
ences are not relevant. Yet, degree, then
among
prospective football players.
and discernible height, but quarter-inch
differ-
require that a test differentiate to the finest that
do not do
measure, and search for more precise ones. Psychometric
28
so, like
our tape
tests drafted to
UNDERSTANDING 'INTELLIGENCE Driving Tests and Psychometric Tests
Many
designed to certify sufficient competence rather than to
tests are
differentiate the population. Driving tests are an important example. certify sufficient
we could
ers to use the highways. Alternatively,
cense
We
competence, which has allowed the vast majority of driv-
test to differentiate the
require the driver's
population of would-be drivers
in the
li-
same
way that psychometric tests differentiate the population by intelligence. If we did so, such important skills as making a left turn in heavy traffic,
— com—would be deemphasized or eventually removed
correctly interpreting street signs, and parallel parking
pose most driving
from the
tests
because they
test
lation. Instead, ability to
way
speeds, to
fail to
make
sharp distinctions
maintain control of the car
make Grand
at
to that quarter-inch
and magnify
practical implications (see
may
damaging
it
take a difference that
—even
the pylons
Differentiating to a fine degree
to
make
is
comparable
such fine distinctions have no
if
box above). is
not a need of testing.
matic commitment of the psychometric
need
the popu-
the driving test.
differentiate the population finely
tricians
among
Indianapolis Speed-
Prix hairpin turns effectively, and to navigate
a slalom course within a given time limit without
would form
skills that
test designers.
fine distinctions in their tests
that fine distinctions in intelligence are real
and
It is
That
is,
a paradig-
psychome-
because they presume
that they matter.
To understand that paradigmatic commitment, let us start with the very name "psychometrics." In scientific disciplines, such as biology and sociology, the suffix "-logy" fers to
their efforts to the
task
may seem
how
it
is
— how much
distributed
son's place on that
from
measurement of
of."
The "-metrics"
suffix re-
this direction?
Why
They
intelligence
among people
assumed
approach
to this
say, in effect, let us
assume
intelligence. Their
quite alien to the outsider.
the fundamentals
and
means "the science
measurement. Since the 1930s most psychometricians have devoted
is
out there in the population
—and
distribution.
Why
set
about finding each per-
do they approach the task
do they not measure intelligence
searchers measure, say, educational attainment or income? the units of measurement. Educational attainment and
29
directly as re-
The answer
income have
is in
clear
CHAPTER and measurable units intelligence are or
—years and
how
2
But nobody knows what units of We do not have "ounces" or "watts"
dollars.
to count them.
of intelligence.
Psychometricians have "solved" the problems caused by the lack of a unit of intelligence rectly) that
by assuming those problems away. They note
/fwe assume
(cor-
(1) that intelligence in the population is distributed
normally, like a "bell curve," and (2) that people can be ranked from the
lowest to the highest on a
test,
then
we can give each
person an intelligence
score by simply converting his or her rank into a score. If the previous two
assumptions are correct, then the units of measurement turn out to be relevant;
we do
intelligence.
not actually need to see or weigh an
Those
are
some big
There
ifs.
amp
or decibel of
a basic circularity here.
is
ir-
The
scores people receive on intelligence tests derive from the measuring tools rather than
What cians
is
from observations of intelligence
at
work.
the distribution of intelligence in the population? Psychometri-
do not know. They assume
that
it is
The
a bell curve.
bell curve is
characterized by the large bulge in the middle representing the preponder-
ance of scores near the average and by long, sloping edges indicating that large departures
from average are infrequent. And the
fectly symmetrical.
But
this distribution is not a
dence about human intelligence. The
first
bell curve is per-
discovery based on evi-
psychometricians chose the bell
curve to represent the distribution of intelligence as a matter of faith that just about everything is distributed this way,
nience. 9 For technical reasons,
it is
and as a
statistical
conve-
easier to convert test scores into
IQ
scores using a "normal," bell-curve distribution than using any other statistical distribution.
If a last
10
psychometrician assumes a bell curve and ranks people from
on a
test,
assigning scores
is
routine
And
are tied at the top or at the bottom.
comes
crucial.
A bell
as long as not too
here
is
where
first
to
many people
differentiation be-
curve, by definition, has very few people at the top
and the bottom. Therefore, a "good" test
—
must produce scores
in a bell
getting top scores and very
test
must also have
this property.
The
curve distribution, with very few people
few people getting bottom
scores. Psychometri-
cians routinely include questions in their tests to produce this very result. If
they are building a vocabulary test (a typical part of intelligence
tests),
make sure to include some words that only one person out of 100 will know and some words that only one person in 100 will know. Whether
they not
—
—
knowing these rare words say, "snood" or "entremets" has anything to do with intelligence as "a capacity for inferring or applying relationships 30
UNDERSTANDING "INTELLIGENCE" drawn from experience"
in effect, irrelevant.
is,
Those very common and
those very rare words define the ends of the scale.
Psychometricians then end up with scores for each
test taker.
But be-
cause the units of measurement are arbitrary, the numbers are not fixed like inches on a carpenter's steel rule.
They
rubber band that can be stretched to
more
are
like pencil
between any two
fit
Give psy-
points.
known
chometricians a word to define (or another problem)
marks on a
to
99 percent
of the population to nail one end of the rubber band to and a word to
only
percent of the population to nail the other end of the band
1
they will be able to
fill
in the rest
In sum, psychometricians
tributed
among people
assume a
single intelligence and that
and so they build
no substantive reason
is
to,
and
it is
dis-
of the distribution, the bell curve.
like a bell curve,
yield a bell curve. There
known
to start
their
method
from these
to
as-
sumptions.
The Bell Curve
By
looking
the
commitment
at
The Bell Curve
to differentiate
call that the basic
Armed in the
itself,
in
The
Bell
we can
see
Curve
how
from
distortions arise
and the commitment
Re-
to the bell curve.
evidence Herrnstein and Murray draw upon
is
a
test,
the
Forces Qualifying Test, given in 1980 to over 12,000 young people
National Longitudinal Survey of Youth. The
AFQT
is
composed of
four sets of questions for a total of 105 test items. (These sets are inter-
spersed
Armed
among
228 questions,
six others, totaling another
Services Vocational Aptitude Battery.) For the
to
form the
moment, we
will
accept Herrnstein and Murray's claim that the 105 questions measure intelligence. Later in the chapter,
measure.
If
we
will look at
many people answered
that
number of questions
shaded shape labeled "Original" in figure to the right,
what those questions actually
one graphs the number of questions answered correctly by
2.
shows how many of the white
1
.
correctly,
how
one obtains the
That shape, leaning heavily
test takers
answered
that
many
questions correctly. (To be consistent with Herrnstein and Murray's analysis
we
of intelligence and class inequality,
in the
NLSY
largely analyze only the whites
sample.) For example, 45 young adults answered 104 or 105
correctly and 126 answered fewer than
represent "smoothed out" data; that
is,
30
right.
slight
and
(Both lines
in the figure
erratic variations in
num-
bers are averaged out for legibility.)
Oops! Despite the best psychometric assumptions, tion is not a bell curve!
this original distribu-
Most white respondents scored near 31
the top end of
CHAPTER
2
Originally, the distribution of test takers by the number of correct answers was bunched upward; it took considerable effort to transform it into a bell curve.
300
Original
10
20
40
30
50
60
70
80
90
100
Number of Questions Answered Correctly
2.
1
.
Distribution of Original Scores on the
AFQT
and Distribution of
Scores as Transformed by Herrnstein and Murray (Source: Authors' anal-
NLSYdata)
ysis of the
the
test.
About 20 percent of the
of the questions correctly. This
key measure
in
test takers is
answered more than 90 percent
the real distribution of scores
from the
The Bell Curve. There are simple reasons why the
did not yield a bell curve. 11
Our concern
here, however,
is
AFQT
with the psy-
chometric insistence that there must be a bell curve. The other line
in fig-
ure 2.1, labeled "Transformed," shows the distribution of test takers after
Herrnstein and Murray recalculated the scores. is
also the source of the
title
It is
roughly bell-shaped;
and the jacket design for The Bell Curve.
it
How
did Herrnstein and Murray get a bell curve from the lopsided distribution
of original scores?
By
a
good deal of
mashing and
statistical
Because they presumed, as psychometricians do,
that intelligence
distributed in a bell curve, they justified transforming the
must be
number of ques-
produced the
bell curve
what Herrnstein and Murray did was give "extra
credit" for
tions each test taker correctly in the figure.
stretching.
answered
until they
12
(Practically,
32
UNDERSTANDING INTELLIGENCE" being
at the
higher or lower ends. For example, a difference of three correct
answers for those
upper end, 105 versus 102, yielded differences
at the
what Herrnstein and Murray labeled "zAFQT" scores of one "standard deviation," from 2 to
3.
full step,
But for respondents near the middle,
— 28 more
in
or it
took far more correct answers to
move one
from -1
took 16 more correct answers to go from
-3
to 0.
to -2.
among If
The
effect of this
exaggerate, to stretch out, slight differences
is to
high and low ends
test takers at the
among people
is
forced into these kinds of contortions.
it
matters It
is
in
order to form a bell curve.)
that an
like height
unseen
matter?
that this transformation creates a bell curve
scheme exaggerates
and lowest ends. And
They "knew from
that is
one
One reason
where there was
paradigm. Another
is
the importance of being at the high-
what Herrnstein and Murray wanted
collateral data that
at the tails" (p. 573).
in a
entity, intelli-
in a bell curve,
is,
Why does this
shows how researchers can be trapped
that this rescoring
occurs
move
it
one builds a science around the axiom
none.
correct to
Near the bottom,
gence, must be distributed
est
step
much
to do.
of the important role of IQ
So they did not use simple
alternatives to that
transformation, like centile score (99th percentile, 98th percentile, etc.) or
raw score (number of questions correctly answered), but instead constructed the bell curve.
expectations that the stein
By
stretching out the
tails are critical.
and Murray because
do so much
they helped affirm their
tails,
the tails are critical to Herrn-
their entire focus is
5 percent (the "cognitive elite") (the "very dull")
And
on demonstrating
better
that the top
and the bottom 5 percent
do so much worse than everyone
else.
they ignore the middle 90 percent of Americans. This
For the most
part,
not good science
is
13 but self-fulfilling prophecy.
In this way, the psychometric testing intelligence. If "intelligence
problem
is
how
to
measure
constructed intelligence
By whether must be
it
is
test.
paradigm prefigures the meaning of
what intelligence
intelligence.
How
tests
measure," then the
The answer
do we know a
test is
is
to use a well-
well-constructed?
sharply differentiates the population. Hence, intelligence
finely distributed (preferably in a bell curve). Herrnstein
ray's analyses of the "cognitive elite" and the "very dull"
impossible without the presumption that intelligence that there are distinctive top
and bottom 5 percents.
is
and Mur-
would
all
be
so finely graded
And
makes sense
it
only within the psychometric paradigm.
Psychometricians' methodological need to differentiate dovetails with other reasons to differentiate. Administrators, employers, and other con-
sumers of psychometric
tests
use them to allocate,
stratify,
and label
dents and employees. Sorting people out into precise ranks
33
may
stu-
not be
CHAPTER
2
necessary for education or for work, but admit, hire, promote, and so forth.
own
it
(It is
is
who
useful for deciding
we allowed
as if
to
only the top
Then we would design driving tests to differentiate so finely. But that seems silly. When more precise distinctions are necessary, say for licensing heavy equipment drivers, then more 5 percent of drivers to
sports cars.
precise tests are used.)
Tests constructed with differentiating as a goal also legitimate the strati-
imposed, make
fication that is
it
seem just. Tests
good devices of
are
legit-
imation because they appear to be objective. The record on the use of intelligence tests to justify ethnic differences
is
a long one.
14
We
will explore
that role of tests in chapter 8.
Sharp differentiations make psychometric
most other
life tasks, like
fundamentally unlike
tests
driving and doing one's job.
A good
test
be like a good task, and neither needs to differentiate in order to about the talent of the person being tested. However, ability
been used primarily
formance, needed for
statistical analysis
enough:
things.
The
desire to
compare
per-
and organizational work, requires
no matter how
that tests discern difference,
scheme of become ingrained in
have not
to tell us about the talent of the person being
tested in relation to the talent of other persons.
the larger
us
tell
to tell us about the talent of the person being tested;
have been used
instead, they
tests
should
trivial that difference
may be
Because we argue against a position
the very idea of testing,
we
cannot
state
it
in
that has
strongly
possible to construct evaluations that point to the talents and
It is
weaknesses of people but
that
do not discover, magnify, and therefore
lidify originally trivial differences.
Such
so-
might incorporate evalua-
tests
tions of diverse skills in general ways.
Most ing, for
skills are useful
example,
ability test that
is
no
even
if
good test
every good
will
everyone succeeds
behave
many
at
them. Walk-
because everyone learns to walk.
most people can pass equally well
to psychometricians or to
have as
if virtually
less useful
is,
however, of
little
An use
users of tests. Therefore, test makers be-
We
test
must
differentiate.
like a
good
task, revealing
argue, however, that a
whatever differences there
are but not magnifying differences for the sake of differentiating people.
The Psychometric Circle In these ways, psychometrics rests Its
on
statistics as
very definition of intelligence, g, derives from
takers' scores
on various kinds of
much
on psychology.
statistical finding that test
ability tests correlate
34
as
with one another;
UNDERSTANDING "INTELLIGENCE" people
who do 15
gence, or
measure a common, underlying property: general
There
g.
The explanation
well on one tend to do well on another.
that the tests all
is
intelli-
nothing wrong with beginning a research program
is
trying to explain a statistical regularity such as the intercorrelation of tests,
nor with making
concerns while one
statistical
what
is
wrong
is
work
limiting the
To demonstrate used
how
it
is
validity has
ways
we need
this circularity
supposed
to only these concerns.
becomes irremediably
to introduce
been evaluated for intelligence
tests.
to establish validity, or three types of validity.
way and
One
type
is
lying
trait
we
measures exactly is
clear in
There are three main
Sometimes
a measure
invalid in others.
A
predictive validity.
can predict an outcome that
closing off
one more concept
it
measure. The psychometric circularity
to
appears valid in one
By
However,
circular.
A test has validity when
to evaluate tests: validity.
what
the middle of this process.
is in
other issues, the reasoning
all
nor with setting aside non-
statistical regularity a goal,
measure has predictive
we assume ought be
validity if
it
predicted by the under-
we assume that intelliwe show that people who do well grades, we can then say that the SAT
are trying to measure. For example, if
gent people get better college grades and
SAT
on the
also get better college
appears to have predictive validity as a measure of intelligence, given that the
outcome of interest
is
college grades. If we were to say that the
what
predictive validity without stipulating
is
SAT has
being predicted, then our
statement would be incomplete. To evaluate a claim of predictive validity,
we need
to
know what
is
being predicted (grades) as well as what
underlying concept being measured
is
the
(intelligence).
Figure 2.2(a) illustrates these assumptions graphically. The double-
headed arrow between grades and covers, that the
two
correlation between lation
are correlated.
score shows what the research un-
(A double-headed arrow
two measures; a single-headed arrow
and an assumption about
the empirical information that
how
SAT
the direction of causality.) This exhausts all
is
available. Figures 2.2(b) and 2.2(c)
analysts invoke additional assumptions to
what the
SAT really
show
draw conclusions about
measures.
Figure 2.2(b) illustrates
how one may
validity as a test of intelligence
assumes
indicates a
indicates a corre-
infer that the
when used
that intelligence determines
SAT
has predictive
to predict college grades.
performance on the
SAT
It
and that
intelligence causes better college grades, so that the observed correlation
between the observed
SAT and
college grades in figure 2.2(a)
entity, intelligence.
35
is
the result of the un-
CHAPTER
2
Predictive validity can imply different causes.
(a)
Correlation
College Grades
v SAT
Scores
(b) Intelligence Interpretation
College Grades
Intelligence
SAT Scores
(c)
Social Class Interpretation
College Grades
v
Parents' Socioeconomic Status
SAT
2.2.
Scores
Different Interpretations of Predictive Validity
36
UNDERSTANDING "INTELLIGENCE" The inference depends crucially on the two assumed causal relationships. They may be accurate, but we have no way of knowing for sure. To know for sure, we would have to have another, better measure of intelligence. Yet,
if
we had such
a better measure,
gate the predictive validity of the the intelligence interpretation
SAT
as a
SAT as
a
not need to investi-
measure of intelligence! In sum,
just that, an interpretation of a finding, not
makes
a finding itself. Figure 2.2(c) pret the
is
we would
that point
more
clearly. Here,
measure of parents' socioeconomic
we
inter-
status, believing that
higher parental socioeconomic status determines both higher grades and
higher scores on the SAT. Juxtapose figures 2.2(b) and 2.2(c) and you realize
why
predictive validity
show
they
that
is
any concept
a very
that
weak form of validity. Taken
observed factors can be said to be captured by the validate.
It
cal music,
is
test really
edly, Herrnstein at
is
a logically
how
and Murray and
their allies
validity.
defend the
tests
16
Repeat-
by saying:
however, exaggerated.) 17 The problem
is,
is
valid test if
we
we
that simpler tests save schools
The
that
is
outcomes;
A test has criterion
already believe
already had one?
newer one may be cheaper, quicker
on the people who take the
that is predicting
criterion validity.
correlates with another test that
would we seek a Note
on predictive
well these tests predict grades, or economic success, or job
A second kind of validity
the
in classi-
weak form of validity. Leaning on more research that would iden-
we do not know whether it is intelligence only know that the tests predict them. it
seeking to
measures. But research psychometricians, after gen-
performance. (That claim
ity if
is
often a substitute for
erations of work, continue to rely heavily
Look
one
what have you.
Thus, predictive validity
what a
test
could have been height, body mass, hair color, interest
predictive validity tify
together,
can be assumed to be related to both of the
valid-
is valid.
typical reply
we
Why that
is
to administer, or otherwise simpler.
and agencies money, but the cost
falls
The correlation between an expensive and more expensive test may be worth it to the person denied a position because the test that was used was second best. The set of assumptions needed to establish criterion validity are shown in figures 2.3(a) and 2.3(b) Figure 2.3(a) shows a statistical correlation a cheap test
is
tests.
not perfect. Thus, a
between an expensive indicator and a cheap
indicator, say ten hours of
personal portfolio evaluation and a three-hour SAT. Figure 2.3(b) shows that, just as in the predictive validity case, there are
two assumptions
for
every observed correlation. The assumptions are that both the expensive indicator and the cheap indicator are caused by
37
some unobserved
third
CHAPTER
Criterion validity also
(a)
makes
2
a causal assumption.
Criterion Correlation
10-hour Portfolio (expensive)
A V SAT (cheap) (b) Intelligence Interpretation
10-hour Portfolio
S
(expensive)
Intelligence
SAT (cheap)
2.3.
entity,
here intelligence.
Interpreting Criterion Validity
We might instead assume that the entity that deter-
mines both the expensive and the cheap indicator
nomic
is
parents' socioeco-
status or almost anything.
Under most circumstances,
criterion validity
the same. 18 In both, assumptions
and predictive validity are
outnumber information two
cause assumptions outnumber information two to one, there
way of
for
faith.
someone But
to
be confident of their validity
faith is not the
to one. is
only one
—one must make
measures the concept
test
maker must specify
simply with
statistics,
a leap
only answer to the validity dilemma; there
third kind of validity: construct validity. Construct validity exists test
it is
supposed
that content
to
measure. To establish
Be-
is
a
when
a
that, the
ahead of time. This can be assessed not
but by substantive analysis.
38
The
inability to evaluate
UNDERSTANDING "INTELLIGENCE" construct validity by statistics has rendered that validity invisible to
many
psychometricians.
commonly argue
Psychometricians, however, relation that is
umn
is
no necessary
being measured. For example, Jensen contends that the falling col-
of mercury in a barometer does not remotely resemble the phenome-
non
that
test
look anything like
is
that there
between the manifest content of a measure and the phenomenon
We need not therefore expect that an intelligence real-world intelligence. We agree that resemblance
predicts, rain.
it
19
not necessary; but connection
between a
tion
barometer
necessary. For example, the connec-
is
column of mercury and
falling
reflects falling local air pressure,
A
rain is physical.
falling
which, given the higher
air
pressure in surrounding locations, creates a wind pattern that leads to a lifting
of
air.
As
the surface air rises
ture, leads to rain.
cools and,
it
if
looks like rain but because there
is
air patterns.
on a
the measure, questions
test,
Hence,
a connection between
it
is
measures
is
it
is
not suffi-
If the
only
the case for the con-
paradigm, however,
really does not address
(Below and
intelligence.
it
predictive validity, then
By attending to construct validity, can be made explicitly. The psychometric weak.
does not attend to construct validity;
what
it
is
and the construct, intelligence.
evidence that something real has been measured is
enough mois-
20
cient for the researcher simply to assert that there
the claim
is
a well-understood causal connection
between the barometer and on-going
nection
there
Thus, the barometer does not predict rain because
we do
in chapter 3
whether
address that
issue.)
A related problem is
the inattention psychometricians have given to the
"population" of intelligence-demanding tasks.
survey research
must be able
statistics is that,
A
fundamental notion
in
before one draws a sample of people, one
to specify the population
from which the sample
is
drawn.
Otherwise, one cannot be sure what the sample represents, and then every inference one
makes from
the sample
analog in psychological testing tasks to be tested before test questions. Testers
metricians
still
is
that
is,
one constructs the
have known
in principle, indefensible.
test, that is,
this for a
draws a sample of
long time, but
many psycho-
contend that one need not specify the population of
before making an intelligence
test.
The
one should specify the population of
Their argument
is
skills
akin to saying one
need not identify the nation under study before conducting an electoral poll.
Many
below,
psychologists
now
reject
many have moved beyond
such faulty reasoning. As
the limits
we
shall see
imposed by psychometric
as-
sumptions explicitly to study "everyday" tasks as measures of intelligence.
39
CHAPTER
When
the psychometric
was probably
construct validity
about
how
paradigm
to define,
2
first
began
to take shape, neglecting
wise. Researchers were very confused
of psychometricians to say more than "intelligence
measure" suggests
tests
now
measure, or shape intelligence. But
that construct validity is
is
still
the inability
what intelligence
not relevant to the
mainstream of psychometric work. In the psychometric paradigm, gence is
a black box, visible only in test results.
is
in that
21
what
it is
intelli-
one can specify what
box (although some psychometricians provide
pretations of out").
No
after-the-fact inter-
measure, such as an ability to "figure
that their tests
Therefore, no one can say what policies the research implies.
cannot embrace the successes of psychometrics that predict well
— without accepting
ignorance about whether and
how
—
One
the construction of tests
the limitations of psychometrics
well the tests really measure "intelli-
gence." In terms of social policy, the limitations are key: Without knowing
what the
tests
measure, policymakers are flying blind. Herrnstein and Mur-
ray try to have
both ways. They assert that the tests measure g as people's
it
capacity for inferring relationships. But there
is
no consensus among psy-
chometricians on the content of g 22 The only way to be true to the psychometric approach is to remain agnostic on what the tests measure. American institutions continue to use such tests, but reliance
sure
is
on them
is
wise.
It is
much more
it is
unclear whether heavy
unclear whether what they mea-
intelligence.
What Does the Test Measure? The problem of is
letting intelligence
demonstrated by The Bell Curve
the
AFQT — the
lent
be whatever intelligence itself.
test that is the basis
of their
measure of intelligence. Indeed, "the
test,"
they write, "but
[as]
group of psychometricians
measure
tests
Herrnstein and Murray claim that statistical
work
AFQT qualifies
—
is
an excel-
not just as an
who defended The
all
measure the same intelligence," implying
that purpose, too.
24
achievement
than an intelligence
test
But we
shall see that the
takers learned and displayed their
measure as well of in taking tests.
By
that the
AFQT is much
test. It is
a measure of
AFQT
their interest, cooperativeness, anxiety,
a school
how
well test
—and
a
and experience
extension, then, other such psychometric tests are
40
tests,
serves
more
knowledge of school subjects
the same.
A
Bell Curve in the Wall Street
Journal wrote that "while there are different types of intelligence they
IQ
one of the better ones psychometrically." 23
much
UNDERSTANDING "INTELLIGENCE"
AFQT
Readers will find no examples of actual Curve, but
we
will provide a
few
here.
amounting
to 105 questions.
(As
we
The
noted
questions in The Bell
test consists
of four subsections,
105 questions Herrn-
earlier, the
stein
and Murray used were part of a longer, 333-question, three-and-a-half
hour
battery.)
1
The four
Section
2,
arithmetic skills 2.
11
Section
AFQT sections
are:
Arithmetic Reasoning, composed of word problems using
— 30 questions with
3,
a 36-minute time limit
Word Knowledge, composed
of 35 vocabulary words
3.
Section 4, Paragraph Comprehension,
ring to short paragraphs 4.
Section
8,
—
Four examples
composed of questions
refer-
15 items in 13 minutes
Mathematics Knowledge, composed of questions testing
algebra and higher mathematical skills illustrate
—25 items
the questions.
in
24 minutes
(These are simulated versions
of what remain confidential questions.) Each of these examples at
in
minutes
about average
difficulty.
Recall that the target population
is
rated
is
high
schoolers. 25
1.
Arithmetic Reasoning: If a cubic foot of water weighs 55
much weight
75 /2-cubic-foot tank
will a
]
trailer
lbs.,
how
be carrying when fully
loaded with water? (a) 1,373 lbs
(b)
3,855 lbs
(c)
4,152.5 lbs
(d) 2,231.5 lbs
2.
Word Knowledge: (a)
"Solitary" most nearly
means
sunny
(b)
being alone
(c)
playing games
(d) soulful
3.
Paragraph Comprehension: People
moting land
remember
in
in resort areas for as little as
the
maxim: You
get
danger of falling for ads pro-
$3,000 or $4,000 per acre should
what you pay
for.
the ultimate purpose in buying resort property. If
sake,
it
was
a
good buy. But
might someday be worth
Land investment
is
far
if it
Pure pleasure should be it
is
was purchased only
more,
it
is
enjoyed for in the
hope
its
own
that land
foolishness.
being touted as an alternative to the stock market.
Real estate dealers around the country report that rich clients are putting
41
CHAPTER money
their
in land instead
2
of stocks. Even the less wealthy are showing
an interest in real estate. But dealers caution that sition
be just so
to
much expensive
a "hit or miss" propo-
desert wilderness.
The author of this passage can (a)
it's
with no guaranteed appreciation. The big investment could turn out
best be described as
convinced
(b) dedicated (c) skeptical
(d) believing
Math Knowledge:
4.
(a)
JK is
In the drawing below,
zoid. All of the following are true
the
median of the
trape-
EXCEPT
LJ = JN
(b)a = b (c)
JL =
(d)
a*c
KM
On
face value, these questions do not measure test takers' intelligence, " 26 their "deeper capability ... for 'catching on.' Mostly, they measure test takers'
exposure to curricula in demanding math and English classes. They
remind us of pop quizzes Elsie
in high school.
Two
scholars, Darrel
Moore, who wrote the authoritative book on
AFQT,
Bock and
this administration
of the
describe the section on paragraph comprehension as "lean[ing]
rather heavily
on general knowledge.
A
well-informed person has a good
chance of answering many of the items correctly without reading the paragraph. This
means
educated
that the better
reading the passages and already knowing
them, should have found
knowledge nal
this test
for scoring well
army "alpha"
tests
on IQ
.
.
.
,
having both the benefit of
many of
the facts contained in
very easy." (The importance of general tests
used in World
goes back a long time. In the origi-
War
I,
10 percent of the questions
required familiarity with national advertising campaigns, as
in: "
'There's
27
Bock and a reason' is an 'ad' for a: drink, revolver, flour, or cleaner?") Moore say of the section on math knowledge that the answers would be "known only to persons who had some exposure to high school algebra and geometry, or far cry
who had
28
This seems a
"ability to learn
from experi-
studied textbooks on these subjects."
from measuring intelligence as the
ence." 29 After looking at similar items, political scientist
Andrew Hacker
concluded, "At best, The Bell Curve authors have identified not a genetic
42
UNDERSTANDING "INTELLIGENCE" meritocracy, but what might be called a testocracy: individuals possessed
of a specialized
skill
which, on further examination, has
little
or no rele-
vance to most human endeavors." 30 not necessary to use school curricula questions for measuring "intel-
It is
ligence." For example, social psychologists
Schooler have measured what they
"Suppose you wanted
like:
locations available.
What
to
Melvin Kohn and Carmi
call "intellectual flexibility"
with items
open a hamburger stand and there were two
questions would you consider in deciding which
of two locations offers a better business opportunity?" 31 Other scholars
have also realized
that the kinds of questions
used in the
AFQT and in most
other similar aptitude tests totally miss what they call "everyday" or "practical" intelligence, the kinds
after
of ability that mature people develop, long
high school, as they deal with the complexities of real
life.
These
ways of testing people's abilities to 32 figure out solutions to problems, ways not wedded to school curricula. Such tests may be less finely differentiating, more qualitative, and more
researchers have also formulated other
labor-intensive than psychometric tests such as the sults are different
—but they
would seem
also
to
AFQT— and
their re-
be more valid measures of
intelligence broadly understood.
Researchers have also found that such paper-and-pencil
tests as the
AFQT are limited in predicting how people apply their knowledge in pracContext matters. Children, for example, given an abstract
tical situations.
logical
video
problem
have great trouble; embed
to solve
game and
that
same problem
in a
they often do brilliantly. In another example, child street
vendors in Brazil have trouble solving abstract math problems, but they do fine
when
those
same problems
are presented as
commercial transactions.
Middle-class Brazilian children react in the opposite way; they do better with the task phrased as a school
test.
Perhaps most discouraging to us as
university faculty are findings that students
who do
well in classes on re-
when presented with analogous logical probclassroom. The point is that there is not much transfer
search design do not do well
lems outside the
between academic intelligence and everyday
Given the
limits of psychometrics,
why
intelligence.
33
did Herrnstein and Murray not
choose another perspective on intelligence? The psychometric paradigm not only scientifically limited, the resources It is
needed
possible, as
we
to
make
it is
also limiting because
shall see in the next section, to identify policy prescrip-
adopted an alternative perspective
—
that
probably
Had
—information
Herrnstein and Murray processing, for
does allow one to speak about action, the
much
does not provide
policy prescriptions to improve intelligence.
tions using other approaches to intelligence.
ple
it
is
else in
it
would have changed. 43
title
exam-
of their book and
CHAPTER
2
Psychometric reasoning has become both circular and self-reinforcing. Psychometricians found that different hypothesized that they tests,
making
all
reflected a
tests intercorrelated
common
component,
this statistical
and sharply differentiating in the
tests.
g,
and from that
They refined the more and more distinct, stable, factor, g.
Educational institutions and em-
ployers increasingly used highly "g-loaded" tests for admission and place-
ment
decisions. Precisely for that reason, the predictive validity of such
has probably increased over the years. (Yet
tests
SAT
scores, for example,
do not predict college grades any better than high school grades do.) 34 The
more
institutions sort people
by
test scores, the better the test scores predict
sorting. This predictive validity is then taken as a sign that the tests
measuring intelligence and legitimates further refinements is,
making
the tests even
ing. All this occurs,
actually
more narrow, more
stable,
must be
in the tests, that
and more differentiat-
however, without any clear evidence that the
measure what they purport
to
refined to magnify differences, there
ferences are as large as they
measure.
is little
now seem to
And because
evidence that the original
be. Finally,
even
all
that the tests are designed to
mountainous
if
make minute
dif-
of this would
not be of such concern were psychometrics put into perspective.
would care
tests
the tests are
Who
differences appear
the discussion and use of the tests kept their limits firmly in
the foreground?
Most college ity
applicants and administrators probably ignore the ambigu-
of the tests and the mountains-out-of-molehills strategy of test construc-
tion as they
open
their mail
from Educational Testing Service. But
construction in the psychometric paradigm magnifies differences
people and so impedes our understanding of inequality.
psychometric tradition the reasoning stable,
and highly differentiating
measuring something
one
real.
No
to believe in the fatedness
may
From
outcomes, so they must be
wonder, then, that psychometrics
anyone might lose sight of other
inside the
appear impeccable: Specific,
tests predict
of
test
among
may
lead
Trapped
in that circle of reasoning,
possibilities
and become convinced of
life.
the hopelessness of ever ameliorating inequality. Fortunately, there are
other ways,
changing
ways
that identify intelligence
and are more hopeful about
it.
The Political Arithmetic of Testing Tests appeal to decision makers and the public alike because they appear to
be objective. The
SAT is
a prime
example
in
many
respects.
It is
said to be
35 and appears to be heavily "g-loaded," predicts college grades modestly,
44
UNDERSTANDING "INTELLIGENCE" neutral.
can be useful when applied properly, but too often the
It
test is
An important case in point is the recurring controversy about the relative SAT scores of whites and members of minority groups at some universities. Many people who care about college admissions point out that minority freshmen have average SAT scores far below those of white freshabused.
men. The gaps
that they point to are often quite large.
36
The
minorities,
Herrnstein and Murray argue, must have gotten an edge in the admissions
how
process. Otherwise, they ask,
could freshmen with scores so far below
those of the average white student get into this university? But there statistical fallacy here. It is ity
impossible to
tell
how much
is
a
of an edge minor-
students have been given simply by looking at the difference in average
test scores.
This
preference
so even
is
The groups
sion.
scores were the only criterion for admis-
if test
among freshmen,
will differ in scores
absent any racial
because minority applicants score lower on average than
at all,
do whites. (Why minority applicants have lower average scores
is
the sub-
ject of chapter 8.)
As
a simple example, imagine a very small college that will admit ten
freshmen. The college receives twenty applications from eighteen white
high school graduates and two black ones (roughly the proportions nationwide).
The twenty
students have the following
scores:
1300
1225
1200
1175
1150
1125
1100
1075
1050*
1025
1000
975
950
925
900
875
850
825
800*
750
The two black students twenty
SAT
scores
is
students
is
925.
1014
we
sidering that If the
1018; the average for the two black
is
difference
(These twenty
are
SAT
at this little
admit students, then the ten students
are admitted.
The
process,
if
students were given no edge. Yet
man's score
mean is
— leaving
sion process simply passed
a small
number of "observations.")
college uses nothing but test scores
row
in the top
a
would
like
it
when we examine
for the nine white
1050
22 points smaller than the na-
will
be the ones
who
followed, would be completely race-neutral,
just as the critics of current practices
see that the
is
scores are surprisingly realistic con-
working with such
dean of admissions
these
14 above the national norm). 37 The aver-
(just
The 93-point
tional difference.
The average of
are identified with asterisks.
age for the eighteen white students
to
SAT
freshmen
gap of 103
is
1
points.
to be.
The two black
the entering class,
we
153, and the black fresh-
The
race-neutral admis-
on the preexisting difference
in test scores
from
Any race-neutral process will. much larger and more realistic num-
the applicant pool to the freshman class.
We have bers.
38
repeated the exercise with
The conclusion
is
the same. Race-neutral admission practices that
45
CHAPTER rely solely
SAT
on
scores
would
2
yield an entering class that
shows a
dis-
parity in scores. Race-neutral selection processes pass disparities in the
applicant pool through to the freshman class. Therefore,
gap
in test scores as if
some
to
it
reflected an
edge
that the
students at the expense of others. In part,
tages that suppress the measured achievement of
blacks and Latinos.
It
might also
reflect the
cannot read a
it
reflects the disadvan-
some groups,
Suppose the system does give some edge an affirmative-action system?
Is that
urally occurring" difference that 39
Although the
to minority applicants
we just saw
race-neutral admissions pass
SAT gap between SAT
under race-neutral admissions, the
blacks and whites in the
would be
to black students' scores.
American descent
is
higher than that
flects the distribution
among freshmen of Asian
among white
that universities are discriminating against
may
it
gap would not increase by the
Similarly, the fact that the average test score
process
through
edge simply added on top of the "nat-
freshman class would be larger under affirmative action than
amount added
especially
tendency of minority applicants
on other criteria of admission, such as grades and class rank.
to score higher
through? No.
we
admission process gives
students does not prove
Asian Americans.
It,
too, re-
of test scores in the applicant pool. The admission
simply reflect the higher average scores that Asian American
applicants bring to the freshman class. In short,
we
cannot actually
tell
much
about the race-sensitivity of the
admission process from the racial disparity of the outcome.
A race-neutral
admission process does not result in equal means for students from groups, even
use other criteria as well). The only conclusion that holds in general the profile of
SAT
all
scores are the only criteria for admission (most schools
if test
is
scores in the freshman class reflects the profile of
that
SAT
scores in the applicant pool.
We draw testing.
The
two lessons from first is
this exploration
of the political arithmetic of
the substantive point that the discussion of racial and
ethnic preferences in college admission
is
based on a
statistical fallacy.
The
average score of freshmen from one group might be lower than the average for
freshman from some other group for a variety of reasons;
it
does not
imply that applicants from the lower-scoring group were given preferential
Our second lesson is that the apparent objectivity of tests invites The claim that affirmative action is reverse discrimination would be much harder to make if those who make it could not cite group differences in test scores. Ironically, affirmative action's critics make themselves seem authoritative and objective in the act of committing a fundatreatment.
their abuse.
mental
error. In fact they are neither authoritative
of the numbers
is
nor objective. Their use
selective and calculated to persuade, not inform.
46
UNDERSTANDING "INTELLIGENCE"
The Information-Processing View of Intelligence Despite the inclination of Herrnstein and Murray and
make psychometrics appear to be it
is not.
the only perspective
much of the media to on human cognition,
There are several alternative approaches, ones
that posit multiple
intelligences or that stress the relativism of intelligence, for example.
we
illustrate the alternatives,
look closely
at
To
one: information processing.
In contrast to the psychometric school, scholars within the information-
A
processing tradition begin with a definition of intelligence.
ment of
this
suggest
some of
paradigm
is
beyond our scope, but a
ways
the
in
full treat-
brief discussion should
which the information-processing school
breaks through the psychometric circle. The psychometric and information-processing traditions have very different implications for think about inequality and what
An
we might do
about
how we
it.
Information-Processing Definition of Intelligence
One noted information-processing
adherent, Robert Sternberg, defines in-
telligence as "purposive adaptation to and selection and shaping of real-
world environments relevant
management." 40 Note testing.
Depending upon the
tests
may
to an
environment. The
tell
sterile one.
ing about
to one's life. Stated simply,
that this definition
test takers' familiarity
us something about
Even
how
how
test itself is
at best,
with
tests, intelligence
however, standard psychometric
tests tell us noth-
proficient test takers are at shaping or selecting their envi-
environment
(i.e.,
who shapes differently) or who re-
us that the test taker
tell
interprets the questions
a failure. Thus, a psychometric test
tell
self-
proficient test takers are at adapting
fuses to spend any time on the questions is
mental
an environment, although a particularly
ronments. At worst, intelligence tests the test
it is
does not dismiss psychometric
(i.e.,
may
selects another environment)
tell
us
some
things but cannot
us other important things about the test taker's intelligence.
If intelligence is
defined as mental self-management, then intelligence
can be taught. The very existence of business schools
dence that management
skills
attests to
can be taught. To teach intelligence,
searchers must discover what mental processes are invoked
solve problems. sources.
As
They must
learn
how
test
re-
when people
people manage their mental re-
these processes and strategies
can devise ways to
our confi-
become understood,
researchers
people's use of them and to teach people to use
An example will suggest what may be learned from such research and how that learning can influence intelligence testing and training.
them
better.
47
CHAPTER Many
psychometric intelligence
tests
2
contain analogies.
On
their face,
these types of questions should directly test what Herrnstein and
Murray
have defined as intelligence, the "general capacity for inferring and applying relationships" the inference
may have used ear syllogism;
(p. 4).
Sternberg argues, however, that
also important;
is
two people who
how people
arrive at the
reach
same answer
different processes to get there. Consider the following lin-
simpler than an analogy, but
it is
still
requires that the test
taker grasp a relationship:
Mark
Who
than
is taller
Adam
is taller
is
Adam.
than Jerry.
shortest?
One can use
either a verbal strategy or a spatial strategy to
answer
this
question. In the verbal strategy, the test taker attends to the literal relation-
ship between the objects. In the spatial strategy, the test taker might visualize or
draw a map of
appear on a logic
test
the relationships.
might appear on a
(which answer, a to Part
1
But
this
question
is
unlikely to
of spatial intelligence. However, a question with a similar
| 1
Part 3
spatial test.
Consider the following spatial analogy
belongs in part 4?):
d,
Part 2
I--
Part 4
7
(a)0 (b)
0-0
-
(c)« (d)
•
The question
requires one to infer
part 2 to the relationship
problem presented spatial terms, but
An
from the relationship between
between part 3 and part
in spatial terms.
some
translate
it
Many
This
is
part
1
and
an analogy
answer the question
in
problem and then solve
it.
test takers
into a verbal
4.
information-processing analyst would point out that three distinguish-
able skills are being tested in these examples: (1) the ability to recognize relationships; (2) the ability to apply such relationships to another domain;
and
(3) the ability to translate a given
problem
into the terms that are per-
sonally easier to solve.
Two
key inferences may be drawn from the analogy examples:
First,
intelligence can be taught. Success in such tests depends in part
on recog-
And something can be
has been
nizing relationships.
seen before
Sternberg
—
lists
that
is, if
recognized only
if
it
one has been educated about such relationships.
thirteen different kinds of relationships that can be
48
found on
UNDERSTANDING "INTELLIGENCE"
common
one
test,
the Miller Analogies Test. If these relationships can be
so identified, then students can certainly be taught them.
set
of signs, the twenty-six
difficult
it
would be
letters
had one been exposed
chance of answering correctly, one must know one's
Some may
find
it
regularly
how
to only half of
maximizes one's
own
strengths and
helpful to translate a problem into another
Good mental self-management can
format before proceeding.
a small
of the English language. (Imagine
to read or write
the letters.) Second, in order to select a strategy that
weaknesses.
We
among
teach small children to recognize thousands of relationships
identify
one's strengths and weaknesses. Also, weaknesses can be strengthened, just as strengths
can atrophy, so that training can shape success on such
intellectual tasks.
In sum, intelligence in the information-processing
framework
is
mental
self-management, and mental self-management involves selecting, adapting
to,
and shaping real-world environments. These intelligence
skills
can
be taught and trained; they are neither fixed nor singular as the psychometric
view assumes. The implication for the entire debate on inequality
clear:
To
the extent that intelligence does affect
who
is
gets ahead, so does
teaching.
Measuring Intelligence Critics often
complain
earlier discussion
psychometric intelligence
that
should suggest
why
that criticism
metricians: the tests are supposed to be narrow. But
ligence must take into account the
many
tests are
narrow.
Our
does not faze psycho-
good measures of intel-
aspects of intelligence that are
neglected, deemphasized, or even negatively evaluated in psychometric tests (e.g., creativity).
We
can see the multiplicity of intelligence
existence of "idiots savants"
who
are mentally deficient in others. that points to
many
in
An example
psychometrics but
critical to
a problem. For example, defining a intelligence. In
normal
explicit questions tion,
one of which
ability to define a
Good
it
in
some ways and
in recent brain research
life, it is
of a dimension of intelligence uninformation processing
problem
the rare
is
is
defining
probably a task that requires
problem
that arrives neatly, with
and "the" four alternatives placed just below each quesis
surely correct.
A good intelligence test should test the
problem.
tests in the
tual process.
also see
different centers for information processing rather than
a single central processor.
explored
can perform brilliantly
We can
in the
information-processing tradition also
Not content
test the intellec-
to leave intelligence as a black box, information-
49
CHAPTER
2
how people process information. One line of research that may reveal much about how people process information, as well as shed light on how to improve human performance, is the comparaprocessing researchers study
of experts and novices.
tive study
Experts
know more
facts about a subject than novices do, but that dis-
tinction does not fully explain the differences in
perts and novices.
performance between ex-
The two apparently process information and
think about
a problem in qualitatively different ways. Yet because every expert
once a novice, somehow novices must learn ences in
how
was
to think like experts. Differ-
novices and experts process information could therefore be
taught, as a part of mental self-management.
In one well-designed study, Jan
Maarten Schraagen investigated how
novices and experts differ in their use of problem-solving strategies. 41
Schraagen studied four groups: beginners, intermediates, "design experts," and "domain experts." The specific research question.
was to design an experiment to answer a The domain experts were experts in the specific test
research topics; the design experts were expert in
how
to design experi-
ments. (Thus, Schraagen was able to determine whether knowledge of content alone explains the difference
between novice and expert.) Schraagen
found that both kinds of experts brought to bear problem-solving strategies that beginners
more
and intermediates did
proficient than design experts,
not.
Although domain experts were
owing
knowledge
to their vast store of
about the domain, design experts resembled the domain experts in
how
they reasoned. In short, there were qualitative differences in the structure of
reasoning between experts and nonexperts.
For example, experts tended linking information so that
to use a
procedure
known
remembering one element
as chunking
in a
chunk
forth the full
chunk of information. The opposite of chunking
member each
piece of information as an isolated element.
is
—
calls
to re-
By chunking
information, experts not only sped up retrieval but also assured that re-
any one item
trieval of
in a
chunk would
call forth other
needed items.
Another example comes from chess: A. D. de Groot compared how quickly master players and amateurs could recall the locations of chess pieces on a board.
The masters performed much
the pieces
domly on
were placed not
the board, there
better than the amateurs.
in a recognizable
chess-game pattern but ran-
was no difference between
Masters did better than amateurs not because
But when
the
two groups.
their brains
Why?
were inherently
superior but because they had learned expectable patterns that emerge in a
game
(e.g.,
the positions of a
which kind of defense
is
few key pieces
will trigger the
being used). Remembering
50
memory
of
one piece of the pat-
UNDERSTANDING "INTELLIGENCE"
When
tern stimulates recall of the entire pattern.
placed, learning matters
As
much
less.
the pieces are
randomly
42
these and other expert tactics are studied further, researchers are
learning
how
For example, Jon-
to teach cognitive strategies explicitly.
athan Baron found that children labeled as retarded could perform
most as well as "normal" children on memory tasks trained to use the
al-
after they
had been
same kinds of memorizing and monitoring
strategies
other children use. Lest one think these strategies can only aid students
lower range, explicitly teaching monitoring strategies
at the
(i.e.,
mental
self-management) also increases the success of college-level math dents.
43
We
teach such strategies whenever
we
teach novices to
stu-
become
experts.
There are many more aspects
to the information-processing perspective
on intelligence testing than we can tified
show how
ric tests
treat here.
44
The elements we have
iden-
aspects of intelligence are beyond the scope of psychomet-
but flow easily from the information-processing tradition. Psycho-
metric tests are limited because they tap only a small part of intelligence
and then magnify small differences. Moreover, the information-processing
paradigm does not discard the findings of the psychometricians but puts
them
into perspective
and then adds additional
insights. This is the familiar
process by which one paradigm supplants another.
Implications Particular policy prescriptions that psychometricians might dismiss
a lot of sense from an information-processing tradition,
among
make
other alter-
Murray dismiss educational intervention a waste because they believe that how intelligent people become is
native paradigms. Herrnstein and as
largely fixed at birth or shortly thereafter. Thus,
serious educational proposal
is
to invest
"dull" ones. But because research learned,
money
is
it
—
can be taught and
—
time, talent, attention, and
We can
improve Americans' cogni-
even more than we already have (see next box).
intellectual
self-management
even long
We
in school, tailoring instruction to
The information-processing
tradition, recognizes that lives,
The Bell Curve's only
in "gifted" students, not
that intelligence
sensible to invest resources
into educating all our people.
tive skills
dent.
shows
more
tradition, in contrast to the
can teach
each
stu-
psychometric
people can grow intellectually throughout their
after their formal education has ended. In chapter 7,
review several specific, concrete ways
Americans' intelligence. 51
in
which our
we
social policies shape
CHAPTER
2
The Brilliance of Americans Today: A Puzzle Psychometricians have discovered a dramatic and encouraging but, for them, embarrassing finding: According to their
world are
far smarter
now
people around the
tests,
than were their parents and grandparents. Little
psychometricians' genetic paradigm allows them comfortably to
in the
explain this change.
The "Flynn the evidence,
named
Effect,"
shows a
after
rise in
James Flynn who brought together
Western countries of roughly
fifteen
all
IQ
points in one generation. Another scholar estimates that Americans of the
1970s were twenty-two points smarter than Americans of the 1890s. This is
an amazing change, implying that the average American of about
would be considered "dull"
in
1
895
1975 and that a large proportion of Ameri-
cans today would have been near-geniuses in their great-grandparents' day.
Psychometricians
who
hold that intelligence
have great trouble accepting
this
is
genetically determined
happy conclusion, however. Genetic
change cannot account for such a rapid increase, no matter how optimal breeding patterns
on claim
may be. (Most of the
sources Herrnstein and Murray rely
have been driving intelligence
that breeding patterns actually
downward. But
the historical data are clear and imply that intelligence has
been sharply elevated by changing social environments.) Psychometricians have taken different routes to escape tal
contradiction. Flynn argues that the
IQ
this
fundamen-
tests are faulty; they
do not
measure intelligence but something loosely related
to intelligence,
"abstract problem-solving ability." Herrnstein and
Murray
tentatively sug-
gest that the change
and partly the
really
is
people having learned
change
is
real
partly real (but temporary)
how
to take
and concludes
wrong: Lamarckism
is
IQ
tests.
result of
Miles Storfer accepts that the
that evolutionary theory
must therefore be
possible and people can pass on learned skills bio-
logically!* If "intelligence is
what intelligence
tests
measure," but the results seem
implausible, what can a psychometrician do? These are the sorts of contradictions that unsettle and eventually break * Flynn,
down
a paradigm.
"Massive IQ Gains"; Herrnstein and Murray, The Bell Curve, pp. 307-9,
passim; and Storfer, Intelligence and Giftedness, chaps. that social classes are
converging
and Murray (Weakliem
et al.,
in intelligence
"Toward Meritocracy?").
52
5, 18.
There
is
—another contradiction
also evidence to Herrnstein
UNDERSTANDING "INTELLIGENCE
Concluding Reflections Herrnstein and Murray discuss information processing and other alterna-
psychometrics
tives to
by the end of
in their introductory chapter, but
it
they claim to have good reason for using only the psychometric tradition.
They argue
on the "relationship of human
that their focus is
abilities to
public policy" (p. 19). In doing so, they deny an interest in the development
of
human
abilities.
Indeed, Herrnstein and Murray admit that, had they an
interest in cognitive
development, they would have spent more time using
So we agree with them on this The information-processing approach is far more applicable than is psychometrics when cognitive development is the issue. To the extent that Herrnstein and Murray address the (im)mutability of intelligence, howthe information-processing tradition (p. 20). point:
development
ever, cognitive
tradition
With grams
is
precisely the issue and the psychometric
not appropriate.
is
the exception of criticizing affirmative action and supporting pro-
for the "gifted,"
The Bell Curve gives
little
attention to
educational system might be altered to improve cognitive intelligence
is
skills.
how
the
A study of
remiss in neglecting the inner workings of the only institu-
tion in this country
whose very raison
to solve intellectual
and
life
d'etre
is
improving people's
abilities
problems. The only justification for ignoring
schooling appears to be that Herrnstein and Murray believe intelligence to
be immutable. Yet they use a theory of intelligence that they themselves
admit
We,
is
of limited value in assessing mutability.
the authors of this book, are in a bind.
intelligence
is
analyzed in The Bell Curve
single, essentially
the rest of
the
immutable
entity.
is
We
have argued
wrong. Intelligence
"most benign interpretation" of
is
way
not a
That should be sufficient to dismiss
The Bell Curve. For some psychologists, all
authors were simply operating with
that the
it
is.
the errors in that
One wrote
book
that
"is that the
outmoded psychological notions
.
.
.
an old-fashioned psychometrics and almost equally outdated behavior genetics."
45
But
respond fully to Herrnstein and Murray's arguments,
to
will in the next
two chapters
psychometric stance.
We
set aside these reservations
will look closely at exactly
we
and accept their
how
they applied
psychometrics to the topic of inequality.
When Americans cial policy,
they need not be bound by the limits of the outdated psycho-
metric paradigm.
accompany
discuss the connection between intelligence and so-
They can transcend
the fear-laden politics that
the zero-sum result of psychometric reasoning.
53
seem
Newer
to
under-
CHAPTER
2
standings of human psychology offer better
the psy-
chological factors that advantage
race for
success.
They
ways of understanding some people over others in the
also offer a realistic
hope of building a smarter and
society.
54
better
CHAPTER But
3
+
Is It Intelligence?
M.
any American parents have had the experience of trying to help a teenager with his or her homework and discovering that they have forgotten how to calculate the volume of a sphere or could not recall "lowest
common to
them,
factors." at
Even
age forty or
grammar seem embarrassingly vague teenager often whizzes through the same
subjects like
Yet the
so.
problems. According to psychometric Bell Curve, the teenager
same teenager who
how
is
more
always
tests,
including the one used in The
intelligent than the parents. This
without being driven,
socially
is
ward, and cannot remember to turn out the lights in his or her
How can the
tests that
doing well on other
forms the evidence
The proposition
measure learning
life tasks. It is this
at the
in school
own room.
and
that
do not measure
limited notion of intelligence that
that natural inequality in intelligence explains
among people
"intelligence." If
is
economic
based on the assumption, which the previous
one nevertheless proceeds on
is
a single, fixed trait of
that assumption, the re-
searcher must then find an instrument that accurately measures such in order to test the proposition. In this chapter,
instrument used in The Bell Curve, the
AFQT is cial
a
we
AFQT. We
will
will
poor measure of innate intelligence and instead
environment
that
environment matters most
further implication
is
intelli-
examine the
show
that the
reflects the so-
shapes people's academic performance, largely their
schooling. Because schooling helps people get ahead, this that the social
mance on
awk-
foundation of The Bell Curve.
chapter demonstrated to be dubious, that there
gence
the
teenager be more intelligent? Only by defining intelligence as
doing well on
inequality
is
cannot figure out
late getting off to school,
A to point B
from point
to get
is
that
we
is
confirmation
in explaining inequality.
And
the
can, through training, raise people's perfor-
the sorts of tests that are admission tickets to
upward mobility
in
our society.
The reader
will recall that the central evidence in
The Bell Curve comes
from the National Longitudinal Survey of Youth's repeated interviews with over 12,000 young respondents. In 1980,
were between the ages of
Richard
Arum
coauthored
fifteen
when
these respondents
and twenty-three, they took the AFQT.
this chapter.
55
CHAPTER
3
Herrnstein and Murray use scores on this test as their measure
operational definition
—of
The Bell Curve they argue
low
on the AFQT, determined which
came
poor,
became
as their
Through a few hundred pages of by low scores
intelligence. that
—
intelligence, as indicated
NLSY
youth dropped out of school, be-
criminals, and otherwise had problematic
young
adult-
hoods.
The AFQT, however, was not designed to measure intelligence or cognitive ability. It was designed to predict how well high school graduates would do in the armed forces. It was written to measure school achievement more precisely, to measure how well teenagers attained high school-level math and reading skills. (As many critics have noted, the
—
1
original Binet tests
were designed
school and specifically not as a chologists turned
it
into a
test
problems
to identify children with
in
of inherent intelligence. American psy-
measure of "IQ.") Furthermore, the
AFQT
problematic even as a gauge of a person's school learning, because
is
test
takers' scores reflect other factors as well, such as the instruction they re-
ceived outside of school, their social backgrounds, and their motivations.
The Content of the AFQT In chapter 2,
we
discussed
how
the psychometric tradition that Herrnstein
and Murray follow works from the
test
backward
to the concept of intelli-
gence. Psychometricians did not identify g, the general factor for
gence, by observing people behaving intelligently; they derived statistical
analyses of test questions, from the tendency of people
swer one question accurately built
from the
later forget
it:
test
to
answer others accurately.
It is
upward. (Herrnstein and Murray admit
"The evidence
on
statistical analysis rather
an-
a concept
was perva-
than direct ob-
servation. Its reality therefore was, and remains, arguable"
does the
from
who
this point but
for a general factor in intelligence
sive but circumstantial, based
intelliit
[p. 3].)
What
AFQT measure?
In chapter 2,
we
looked
at a
clearly tested an examinee's
few questions from the
command
AFQT
itself.
They
of school curricula. Here are a few
more examples: 2
Two
X and Y, agree to divide their profits in the ratio of their If X invested $3,000 and Y invested $8,000, what will be Y's
partners,
investments.
share of a $22,000 profit?
56
BUT
INTELLIGENCE?
IS IT
$8,250
(a)
(b)
$16,000
(c)
$6,000
(d)
$5,864
Reveal most nearly means cover again
(a)
(b) turn over (c) take
open
(d)
The
away to
greatest
view
common
factor of
1
6, 24,
and 96
is
(a) 8
(b)2 16
(c)
(d) 12
If j
and k are positive whole numbers andy + k =
12,
what
is
the greatest
possible value of jkl (a)
6
(b)36 (c)32 (d) 11
As
before,
we
see that the
AFQT questions
are manifestly about school
tasks. Still,
intelligence as
most people generally understand
probably learn more in school and so
pay attention, and care more. At this rationale for the
AFQT
test better
least
test
cal arts,
—and thus
—
as
do youths who
better
sit still,
as a test of intelligence. First, school subjects
ignores those other realms
and so on
probably con-
two problems, however, cloud even
comprise but one of many realms of life This
it
on the AFQT. Youths who process information
tributes to doing well
is, at
—
in
which intelligence might
social relations, business,
best, a partial test.
matter.
mechani-
Second, for youths to
display intelligence on this test requires having been at least exposed to the
school subject matter. material
—and
it
If intelligent
youths have not been instructed
schools, lousy homes, lousy attitudes, or simply being too
taken these subjects yet ple, the case
in the
matters not whether the deficiency arises because of lousy
—then they
young
to
will not score well. (See, for
have
exam-
of the Founding Fathers on page 58.) Moreover, of the four
subtests in the
AFQT — arithmetic
reasoning,
57
word knowledge, paragraph
CHAPTER
3
Were the founding Fathers "Dull"? In the course of our history, Americans'
bers and numerical operations
—has
Revolution, most Americans had calculations.
"numeracy"
—command of numAt the time of the
greatly increased.
familiarity with, or interest in, exact
little
The few who formally learned
arithmetic struggled with un-
sophisticated teachers and crude textbooks that focused on rules for converting currencies If the
mnemonic
and measures.*
AFQT had been administered to Revolutionary-era
Americans, to
Founding Fathers perhaps, most probably would have scored
the
lower range. Would
we then conclude
that they
today's high school students? Probably not.
today's students
know more mathematics
were
We
would conclude
that
than did George Washington and
his fellows because today's students are taught
were the Founding Fathers. And
in the
less "intelligent" than
that is just
more mathematics than
what the
AFQT
measures:
what students have been taught. *
A
Cohen,
Calculating People.
comprehension, and mathematics knowledge
—
the
the greatest difference in the final scores. 3
make
two math components
These are subjects
that
probably most require having enrolled in the right classes, having had good teachers, and having paid attention.
The
AFQT
questions themselves testify that what
is
being assessed
is,
foremost, environmental, not "natural." Vocabulary words, profits and ratios, greatest
common
usually in school. the sense that
it
The
factors
AFQT
— is
these are topics that children are taught, also a test of environmental influences in
has a very specific substance, curricula taught in
twentieth-century American high schools. Similarly,
and concepts specific
lary
to middle-class
it
late-
also taps vocabu-
American homes
in the late
4
twentieth century. In another place or time, an armed forces screening test
might consist
of, for
example, reading signs of an animal
how
to find water, staying
rests
on questioning the AFQT's content
upwind
of g by simply reading the the
AFQT
is
a better
test.
of prey,
There
trail,
knowing
and so on. 5 Our critique here
validity (see chapter 2) as a test is
other evidence, as well, that
measure of social background than of "native"
intelligence.
58
BUT
INTELLIGENCE?
IS IT
Schooling, Age, and the Statistical
evidence supports reading the
AFQT
AFQT as essentially a test of maswe
tering school curricula. Following Herrnstein and Murray,
the
NLSY survey,
focus on white respondents only, and use the
they calculated from the 105 scores
"zAFQT,"
bell curve.
6
AFQT questions.
They
call those
referring to standardized scores designed to
how
(See chapter 2 for
reanalyze test
scores
transformed
conform
to a
Herrnstein and Murray constructed a
bell-curve distribution from original test scores that were not distributed in a bell curve.)
Cognitive
people grow test
ability, if
it is
like other
developmental
(at least until the debilitation
traits,
should grow as
of advanced age).
When
doctors
children on, say, hand-eye coordination, they take into account the
children's ages.. Similarly, psychometricians
dren of different ages on raw IQ
do not simply compare
test scores,
because they presume that
children score higher as they age. Psychometricians "standardize"
scores for age.
cause the
Age needs
NLSY
gave the
be taken into account
to
test to
in the
AFQT,
most
a roundabout fashion. In their statistical analyses
test
too, be-
part, dealt
showing
with age in
that the
AFQT
outcomes such as poverty, they introduced the respondent's age
the time he or she took the test as a "statistical control variable." point, all that
sumed
IQ
youths ranging from fifteen to twenty-
three years old. Herrnstein and Murray, for the
predicts
chil-
means
for us
that test takers'
they wanted to hold
it
that Herrnstein
is
AFQT is
at
this
in effect, as-
scores increase as the test takers age, and so
constant.
perspective: Intelligence
and Murray,
At
7
This makes sense from the psychometric
inborn or nearly so and matures in a develop-
mental manner, just as eye-hand coordination does.
But
if it is
instruction in curricula rather than biological
that is critical, then a different attribute
takers' ages, but
taught
when
how much math and
they took the
crude measures of the
test.
The
NLSY
test
provides researchers with only
of instruction: the number of
years of school the test takers had completed in
— not
vocabulary they had already been
test takers' histories
whether or not they had been
development
needs to be controlled
at the
an academic track
time of the in school.
test
and
These are
crude measures, in part because they do not reflect the quality of instruction the test takers
had received. Graduates of Andover, the Bronx High School
of Science, a
weak
troubled children
inner-city high school, or a remedial high school for
all
score "12" for years of schooling completed. These
59
CHAPTER
3
measures are crude also because they do not informal instruction
Children
who
at
reflect students'
home, during the summer,
home
are read to at
exposure to
and so on.
in travels,
whose parents encourage them
or
to
pay
the bill in a store will be ahead of the curve in vocabulary and math. Yet
turns out that even these
it
two crude measures of
instruction
—
schooling and having been in an academic track closely with test takers'
AFQT scores
correlate
—years of
much more
than do their ages.
(Readers will recall that a positive correlation coefficient, between 1
,
means
that people
A negative
who
correlation,
are high
on one
to -1, means that people high on one trait The higher the absolute value of the correlation,
from
tend to be low on the other.
which are
the stronger the connection. Practically speaking, correlations,
symbolized as that there is a
above
r,
r
=
.20 to .30
how
or below r
= -.20
to
how
— suggest
-.30
people scored on
they scored on another.)
The table below shows how ages,
—
noteworthy correspondence between
one measure and
test takers'
AFQT scores correlate with their
on the one hand, and with two indicators of instruction, on the
Clearly,
one can predict how high someone scored on the
by knowing how
had been
in;
and
tend to be high on the other.
trait
far the
other.
AFQT pretty well
person had gone in school and what track he or she
one can hardly predict
AFQT
score at
all
by knowing the
test
taker's age.
Correlations of test taker's
age
at the
AFQT
time the
test
score with.
years of education completed at time test
whether the
test taker
had been
in
.
.
.
was taken
.16
was taken
.54
an academic track
.45
AFQT strongly
Scores on the
correlate with how many years of schoolwhen they took the test (.54). Another way to see who were among the highest and the lowest scorers. The
ing test takers had had that is to look at
high scorers, the top 5 percent (Herrnstein and Murray's "cognitive
were those who answered 101 or more of the 105 questions
elite"),
correctly.
Over
two-thirds of these high scorers had had at least one year of education
beyond high school when they took
the test.
Of
those
who
scored in the
bottom 5 percent (the "very dull"), almost half had already dropped out of high school before taking the
test.
Age, on the other hand, weakly correlates with fact,
age correlates negatively with
constant.
8
That
is to say,
AFQT
if
AFQT
score (.16). In
one holds years of education
among NLSY respondents
with grossly similar
exposure to instruction, the older ones scored below the younger ones. So, for example, older respondents with twelve years of schooling scored
60
BUT
INTELLIGENCE?
IS IT
below younger respondents with twelve years of schooling. The explanation for this result
logical
had been out of
that the older respondents
is
school longer and had had more time to forget what they had learned there. 9 It
makes sense because
to accept Herrnstein
the
intelligence, these results
hood buy
AFQT measures
and Murray's premise
would imply
that as
young adulthood they became
into
instruction. If
we were
instead
AFQT measures
that the
native
people aged from teenage-
Only teenagers might
stupider.
that explanation.
Herrnstein and Murray are not
much concerned
because they assume that intelligence
and purposes even before age
AFQT
were 15
to
with
immutable
—
test takers'
fixed for
ages
intents
all
Herrnstein and Murray wrote, "The
fifteen.
NLSY
test scores for the
is
sample were obtained when the subjects
23 years of age, and the IQ scores were already as deeply rooted
them
a fact about
as their height" (p. 130). That
is
a bold but untenable
statement.
Think about the heights of teenagers. Take a group of
them up from
three-year-olds and line
fifteen to
shortest to tallest. Gather
twenty-
them
to-
gether again in eight years and line them up again. In general, the order of
people from shortest to
tallest will
heights will have changed a fifteenstill
lot.
We
have changed some, but
cannot validly compare the heights of
and twenty-three-year-olds on any one day. The fifteen-year-olds
have growing
have reached
to do, but nearly all of the twenty-three-year-olds will
their full height.
Now think
about the
AFQT scores
of fifteen- to twenty-three-year-olds.
These comparisons are even more problematic
mere .16 shows. score.
their actual
The
It
implies that there
correlation
is
is little
—
as the correlation of a
connection between age and
weak because youths
differ greatly in
how much
education they get between age fifteen and age twenty-three. People leave school
all
along the
by maturation
would be
AFQT
—
way during
as height
is
AFQT score was determined correlation between age and AFQT
those years. If
—then
the
as high or higher than the correlation
score.
It is
far lower,
however. Those
who
between education and receive
more or
better
schooling between age fifteen and age twenty-three pass ahead of those
who
receive less or worse schooling. Consider two
young women
in the
sample, one fifteen and the other twenty-three, and suppose that the fifteenyear-old has exactly the gotten
if
AFQT score the twenty-three-year-old would have
she had been tested at fifteen. Suppose also that the fifteen-year-
old will ultimately attend the identical schools for exactly as long as the
twenty-three-year-old already has done. to obtain the
same
Can we expect
these
two women
AFQT score on the same day? (If intelligence were 61
fixed
CHAPTER by age
fifteen
would be
and
yes.)
if
the
3
AFQT measured this
The answer
is,
intelligence, then the
answer
of course, no. The twenty-three-year-old has
already finished high school and gone on to college; the fifteen-year-old in the ninth or tenth grade.
still
of the instruction she received since age
test all
old
still
this
lem
in
—
but the fifteen-year-
between teenager and parents with
the comparison
chapter opened
AFQT to be The
fifteen,
is
to the
has that instruction ahead of her. The apparent paradox about aging
and intelligence
which
The twenty-three-year-old can bring
—underlines
the fallacy of taking tests like the
measures of intelligence broadly understood.
issue of aging and testing does not merely reveal a technical prob-
The Bell Curve's use of the AFQT.
we mean by
It
speaks to the essence of what
what we can do about
"intelligence" and to
Until the 1960s
it.
and 1970s, psychometricians had convinced most specialists of aging
from about age twenty onward, people became
less
and
that
less intelligent,
because researchers had found that older people scored considerably worse
on the psychometric
AFQT,
tests
than younger people did. But those
largely tapped school
tests, like the
and school-like learning. In recent years, psy-
chologists have realized that such academic tests measure only one special
kind of cognitive ability; the tests ignore other psychological capacities,
such as practical
and wisdom.
skills, social acuity,
New research
uses tests
measure a wider variety of "everyday" problem solving, for example,
that
how
would advise a friend who had a financial or medical probThe research shows that people sustain or even expand their mental 10 The inability of the psychometric paradigm capacities well into old age. to handle this most elementary point, that in many ways mature adults are a person
lem.
"smarter" than teenagers, casts doubt on that entire enterprise.
The age
of schooling.
been
in
AFQT is really a test So does the high correlation between AFQT score and having
pattern in the
NLSY data
an academic track
Classes in academic tracks teach more
(.45).
about the topics covered in the
shows how the
AFQT
than do classes in general or voca-
tional tracks. This should not matter if the
formance the
is
some kind of raw
main determinant of
The In
latter is
main determinant of AFQT
intelligence.
AFQT
score
is
It
per-
should matter a great deal
instruction in academic
if
material.
what the evidence shows.
some ways,
the
AFQT might be a good measure of instruction, but not
one of native intelligence. What
it
people encountered and absorbed.
captures best
It
is
how much
instruction
does that better than does the conven-
tional "years of education" measure,
because the
AFQT
seems
to assess
educational quality and informal instruction as well as simply time in school.
It
taps the difference
between those who spent time 62
in classes with
BUT
IS IT
INTELLIGENCE?
rich curricula, energetic teachers, motivated students,
sources and those the difference
and those
who
who
and plentiful
spent time in classes without those qualities.
It
re-
taps
between those who are "instructed" outside the classroom are not. (See, for example, the discussion of
tions in chapter 7.)
summer
vaca-
These differences favor rich children over poor ones. Of
AFQT also taps the difference between those who concentrated
course, the
who did not. Yet even these we saw in the discussion of infor-
and remember what they were taught and those cognitive skills can be and are taught, as
mation processing
One response
in chapter 2.
to
ally
do measure
is
that
we have
mere technical debate;
we
instruction
a test of instruction,
AFQT
scores re-
intelligence; and, as proxies for intelligence, they cause,
and smart kids go further
test
is
the order reversed:
rather than are caused by, schooling. That
on the
AFQT
our argument that the
especially schooling,
it
is
to say, smart kids score high
in school.
This controversy
is
not a
has profound policy implications. Whether the
give children molds their intelligence or their intelligence
molds the instruction they receive says a
lot
about whether and
how
to
invest in schooling.
To support
their
claim that the amount of schooling youths complete
only reflects the intelligence they brought to school, the authors of The Bell
Curve conducted a complex
NLSY researchers
statistical analysis.
IQ
cess to scores from earlier school-administered
tests for
had ac-
about one-fifth
of the respondents. Herrnstein and Murray report that the higher respondents had scored on the earlier
How many all
IQ
There ercise.
are,
tests
to
little
know was
by education
to the ability to predict
—
measures
however, several problems with Herrnstein and Murray's ex-
errors, which,
when
to
measure the
corrected, double their
intelligence.
12
A
more than
made
own
more sophisticated
same data by two economists showed
AFQT score
AFQT
supposedly measure, intelligence (pp. 589-92).
For one, using years of education
on
AFQT
1980 were caused almost solely by
and, therefore, the
ignores quality of education. 11 For another, they
effects
on the AFQT.
the earlier test score. Herrnstein and
AFQT scores in
that
scores, not
what other IQ
contributed
tests
one needed
Murray concluded earlier
the higher they scored
test,
years of school the respondents had completed in the period
between the two score;
IQ
three times as
effect of schooling
a couple of technical
estimate of education's
statistical analysis
much
as Herrnstein and
mated. 13 Furthermore, Herrnstein and Murray
fail to
Murray
63
esti-
report an important
piece of evidence about the predictive validity (see chapter 2) of the Test takers' scores correlated with the
of the
that years of schooling affected
AFQT.
number of years of school they had
CHAPTER
3
finished at the time they took the test at r
=
.54. Test takers' scores corre-
number of years of school they completed after taking the = .33. The AFQT score better "predicted" past schooling than
lated with the test at it
only r
did future schooling. That
is,
AFQT
the
measured what
test takers
had
already learned, not their ability for future learning.
A
more fundamental problem with Herrnstein and Murray's effort to separate the AFQT score from schooling is that other IQ tests also reflect instruction. Take, for
example, the
children of the respondents. the
NLSY
test
used by the
the late 1980s,
many
NLSY
to assess the
of the young
women
had been tracking since 1978 had themselves borne children.
Interviewers gave these the
By
women's
Peabody Picture Vocabulary
three- to six-year-old sons and daughters
Test. In this test,
an adult reads a word and
then asks the child to pick out from four choices the correctly corresponding picture. This most surely tests substantive learning of vocabulary learning that
may come from
talking with parents, listening to parents read
books, watching educational television, and the
like.
14
Typical IQ tests
share the property of measuring school-like learning. Finally,
we must
not forget that schools often use such tests to track
students. This, too, will create correlations
exposure to curricula, and scores on dents
who
later tests,
scores on early
IQ
tests,
because schools place
stu-
score well in the early tests into enriched classes and place low
scorers into remedial classes.
much
among
The
classes students are placed in affect
instruction they receive and, subsequently,
how
how
well they do on fu-
ture tests. (See chapter 7.)
That instruction in curricula
is
the cause, not the consequence, of test
scores can be illustrated with another sports analogy. Suppose to
measure people's
people to
Some
lift
"lifting quotient"
—
call
it
"LQ."
We
we wished
ask a sample of
weights and use their performance to measure their LQs.
of these people have been working out in
ened themselves and received instruction
gyms where
they strength-
in weight-lifting techniques;
others have barely exercised since their physical education classes in high
how much time people had spent in a gym would correhow much poundage they lifted in the test. Maybe high LQ causes more gym time (perhaps people who are strong choose to work out more), but it is much likelier that the causality runs largely the other way: More gym time causes high LQ scores. Similarly, more instructional school. Obviously,
late highly
time
—
in
with
and out of class
—causes high IQ
scores.
The
AFQT is a measure
15 of instruction received, absorbed, and displayed. And, as such,
the environment
—
it
reflects
schools mostly, but also families and communities
the test takers.
64
—of
BUT
IS IT
INTELLIGENCE?
This understanding of the connection between education and is
not only
more accurate than Herrnstein and Murray's.
we can
us to believe that
raise test
scores
test
also encourages
It
performance through more and better
teaching.
What AFQT
is
gence," but
it
The
a better is
AFQT Measure?
Else Does the
measure of instruction than
it is
of "natural
intelli-
not a perfect measure of the former either. Scoring well or
poorly depends on other factors as well, most notably, on
test takers'
moti-
vation to display that learning.
To see
this, let
us look closely at the
and lowest-scoring
the highest-
of the bell curve distribution,
tails
We
test takers.
showed
in chapter 2 that
Herrnstein and Murray squeezed a bell curve out of what was not bell-shaped distribution of original scores.
They did
that
convinced not only that intelligence had to be distributed but also that the dramatic differences in the extremes.
most
ute the
It is
in a bell curve,
outcomes are found between
life
the highest 5 percent, the "cognitive elite,"
to society
at all a
because they were
and the bottom 5 percent, the "very
who
dull,"
contrib-
who
are
responsible for most social problems.
Who
were
in the
upper end, the 5 percent "cognitive elite"? As noted
they were overwhelmingly people
earlier,
beyond high school before taking the disproportionately
men
math than have American
is
girls. In addition, the
— lucky because had they
empha-
further in
top 5 percent were also
answered just one or two more
down
to being
merely "bright."
a result of standard
psychometric insistence on differentiating, de-
When
small differences are exaggerated, chance be-
scribed in chapter
comes
the test
American boys have gone
questions wrong, they would have dropped (This
who had had some schooling 16 And the high scorers were
—68 percent—probably because
sized mathematics. Traditionally,
lucky people
test.
all
the
2.
more important.)
Who were in the bottom, the disreputable 5
percent?
We pointed out
that
they were uneducated. In fact, 27 percent of the "very dull" had dropped out of high school at least three years before taking the
some of
the
test.
17
Furthermore,
bottom 5 percent had mental problems. Interviewers
in later
years rated forty-two of the white respondents as "mentally handicapped,"
and those respondents were cent. If
we
believe that the
likelier than others to
AFQT
is
an IQ
test,
be in the lowest 5 per-
then
all
of the bottom 5
percent were mentally "retarded" by the conventional standard that IQ
65
CHAPTER scores below 75 one-fifth of the
mark
the retarded.
18
If
3
we
use a line of 70, then about
bottom 5 percent were retarded. (Unfortunately, the two
—
ways of assessing mental handicaps interviewer rating and AFQT score hardly agree, casting more doubt on the AFQT as an intelligence measure.) 19 Including these two groups in the analysis distorts the picture
—
of how intelligence operates in the general population, the huge majority of
which
We
neither mentally retarded nor handicapped.
is
many
also suspect that
of these low scorers were neither mentally
disabled nor unintelligent but were instead discouraged test takers
gave up or were, to use the colloquial, "screw-ups" jected the
test,
fooled around, or just did not take
who
—respondents who
it
seriously.
Why
re-
do we
think that? Consider that sixty-nine of the whites (1.5 percent of them)
scored below chance. Test takers
who
simply answered the questions ran-
domly would have, on average, gotten twenty-six correct. 20 How do we interpret someone scoring below chance? Many of the questions were designed to be answered correctly by virtually everyone. Yet these did worse than throwing darts at the answer boxes.
who
of the respondents
because
it
was hard
test seriously, either
We
test takers
suspect that
many
scored below chance simply lost interest, perhaps
them or
for
for other reasons; others did not take the
answering randomly or not answering many questions
at all.
We
believe that
many
many who
of the lowest 5 percent, including
scored below chance, were "screw-ups" for three major reasons.
First,
some of these low scorers had scored at or even above average in previous, school-given IQ tests. 21 Second, the below-chance scoring is greatest in the last section
test
of the
test:
form (sections
Three of the
2, 3,
and
4).
AFQT
From
subtests appeared early in the
1.5 to nearly 5 percent
of whites
scored below chance on those. In the last subtest Herrnstein and Murray used, section 8 (which appeared after respondents had already faced 264
multiple-choice questions), 10 percent scored below chance tion to us that, as the test
went on, more
test takers
"dropped out."
sections 4 and 8, five respondents actually scored zero sible result
on a lengthy multiple-choice
test,
—
is
This pattern of taking
The seems
setting in to
our tests
which the
own
were tested
roughly ten in hotels,
On
both
suggesting that they did not
many test among youth.
that
observations of test-taking
has wider implications.
NLSY respondents took the AFQT during
have been optimal for valid
(6 percent)
indica-
a virtually impos-
even attempt the questions. The third reason we believe takers dropped out
—an
individually.
testing.
The
rest
About 700 of were tested
libraries, or similar locations.
66
Two
1980
the subjects
in
groups of
interviewers ad-
BUT
IS IT
INTELLIGENCE'
ministered and proctored each group administration. Respondents were
show
paid $50 (worth over $90 in 1995 currency) to
on the
test
how
they scored
did not matter. Respondents were also enticed to participate by
being offered the
were
up;
test results
mature as
relatively
and vocational information. The
test takers
go
and had already demonstrated their commitment to the interviewed a couple of years
earlier.
test takers
(fifteen to twenty-three years old)
NLSY
by being
22
Contrast this situation to the typical setting for standardized tests in schools: restless, tries to
A
teacher hands out booklets to a class of twenty to thirty-five
perhaps resentful, adolescents or preadolescents. The teacher then
maintain order during the long
at least as
test. It is
reasonable to assume that
high a proportion, but probably a higher one, of students become
distracted or discouraged or just
NLSY. Even one-on-one
"blow off
testing in schools
that test as did the is
problematic. School psychol-
do so many evaluations
ogists in inner-city schools, at least, are pressed to that their testing tends to
Students likely to
who
AFQT in the
be rushed and superficial. 23
ignore or resist teachers' instructions for taking a test are
do poorly
in school,
and they are likely to do poorly outside of
school as well. Employers do not appreciate such rebellious attitudes any more than teachers do. Consequently, low scorers will often become low achievers. Perhaps this outcome is due, in part, to poor intelligence. But it is
probably due to poor attitudes
even
if
—one reason
that
such
tests predict well
they do not measure intelligence well.
Rebelliousness, apathy, and other attitudes reduce test scores. So do
emotional conditions such as anxiety. 24 Where do such attitudes and
feel-
come from? Perhaps they are to some degree "natural," inborn temperaments. But many studies of youths point to social conditions such as
ings
—
poor schooling, disorganized neighborhoods, stressed parents, high unemployment, and being in a stigmatized minority tudes.
Once
—
that stimulate
such
atti-
again, then, the evidence points to the environment as the basic
explanation for
test scores.
We
will revisit these points
explaining racial differences in test scores in chapter
when we
turn to
8.
Conclusion
We have made
several important critiques of using the
AFQT as a measure
of "natural intelligence." To these, Herrnstein and Murray can provide one rebuttal:
The
AFQT
has predictive validity. That
correlated with other, later outcomes that
67
we
is,
scores on the test are
care about.
High scorers
are
CHAPTER likelier to
go on
do well
to college, to
so on; low scorers are likelier to
how
(although
weak
strong or
whether a
in the military, to get
same
the
fail in
the connection
we
will see in the next chapter). But, as
validity tells us
3
is
good jobs, and
We
tasks.
grant
pointed out in chapter
test correlates
all this
with such outcomes
we
2, predictive
with outcomes, but not
why
it
does. Cognitive skills are surely part of the explanation for the correlation
between
tests
numbers
who
skills are
read faster,
who can juggle more who think more
recognize more words, or
on the AFQT. These
clearly will generally score higher
them well
who
and outcomes. Youths
in their heads,
One
in school and, to a lesser degree, in life.
learned in a variety of settings,
to successful learning than others.
some of which
The other point is
skills will serve
point
are
is
that such
more conducive
that the
AFQT in large
degree reflects other characteristics besides some basic intelligence.
were just "noise," just random influences on
If those other characteristics
the
AFQT
would not undermine The Bell Curve argu-
scores, then they
— such how — not mere
ment. But those other characteristics school and self-defeating attitudes
how
people end up in
life.
attribute all the effect they
as
The problem
is
have found of
native intelligence; they should attribute
that
AFQT
much,
if
recently people were in
They influence Herrnstein and Murray
are
noise.
scores on outcomes to
not most, of the effect of
AFQT to these other characteristics, such as instruction and motivation. Another way to understand what we have shown is that test takers' AFQT the
scores are
good summaries of
a host of prior experiences (mostly instruc-
someone do well in adult life. The error is to attribute of the AFQT to some unseen native intelligence. 25
tion) that enable
significance
the
The AFQT basically measures how much formal and informal instrucsomeone has received and absorbed. It is no surprise that youths who
tion
do well
in school will usually
continue to do well in school, in addition to
doing well on the standardized
tests.
(Doing well on early
tests also
opens
up more opportunities for advanced schooling, for example, by moving into a higher track.) to
do well
in life.
society today
is
And it is no surprise that doing well in
As every
parent of an adolescent
is
structured so that the college diploma
imagine success. But
this points to the direct
school helps one
acutely aware, our
is
necessary even to
consequences of schooling
(And cognition, unlike genes, can be see in chapter 7.) The AFQT also reflects
rather than of cognitive ability.
changed by mental
how
policy, as
we
shall
disabilities, motivation,
well
someone does
and
attitude.
in postschool
deavors requiring school-like
skills
And these traits,
endeavors
too, influence
—
especially those en-
and school-like discipline. For such
reasons, the tests do predict outcomes well.
68
BUT
As predictive
devices, then, the
a general way, although they
it
AFQT and
similar tests work, at least in
do not predict very well the performance of
AFQT
any given individual. But the
whether
INTELLIGENCE?
IS IT
predicts
outcomes irrespective of
reflects intelligence or, instead, reflects other characteristics.
practical purposes
— such
as admission to college
—
predictive validity
be sufficient. But for understanding, predictive validity
must be concerned with content sure, with
why
gauge an
ability
formed early
then where people end up in
and inequality
is
predetermined.
If,
life
not enough.
AFQT
and similar
in life, stable thereafter
The evidence strongly
The Bell Curve, the AFQT,
is
really a
and
largely reflects natural differences
however, these
tests largely assess in-
struction and other environmental conditions, then inequality
natural nor fixed.
We
with what the tests actually mea-
the tests correlate with outcomes. If the
tests essentially
fateful,
validity,
is
For
may
is
neither
measure
indicates that this key
measure of instruction and
in
attitude,
of things that are changeable, not of innate intelligence.
The policy implications of tion
this analysis are straightforward:
—both formal and informal —
matters.
So do
Educa-
efforts to affect youths'
motivations and attitudes. Instead of shrugging our shoulders and dismissing efforts to raise academic achievement,
cognitive training (see chapter If the analysis
of this chapter
fails at this point. If the
tion
is
correct, then
AFQT heavily
and social environment
we can
successfully invest in
7).
The Bell Curve argument of instruc-
reflects the contribution
to individual
achievement, then Herrnstein
and Murray have not shown the dominance of native intelligence over environment, but the opposite. In the next chapter, however,
We
we will assume how well the
AFQT does measure intelligence. AFQT predicts outcomes compared with measures of the social environment. We will see that, even were we to scrap this chapter and grant all that
that the
will evaluate
Herrnstein and Murray claim about the validity of the text
still
better accounts for
who gets ahead and who
for success.
69
AFQT,
falls
social con-
behind
in the race
CHAPTER
*
+
4
Who Wins? Who Loses? I
n the old blues refrain, Albert King asks, "If you're so smart, how come you're not rich?" Herrnstein and Murray issue the academic equivalent of that challenge. Their evidence shows, they claim, that being intelli-
gent leads people to be successful in school, wealthy, and stable in marriage; being unintelligent leads people to
even prone
to injury. Herrnstein
be poor, divorced, criminal, and
and Murray also claim
that differences in
people's social backgrounds are of minor importance in determining their lives turn out. This is the crux of
The Bell Curve's empirical
more strongly shapes
tions, that intelligence
environment. The challenge to the book's 1995,
the
is
"way IQ dominates
in explaining life
outcomes than does
critics,
this thing
we
outcomes. The contrary 1
we showed
In chapter 2 as
life
call
'socioeconomic status'"
is true.
that intelligence is not single, unitary,
measured by the AFQT. In
very same evidence used in The Bell Curve,
were a good measure of
intelligence, the
We
—
specifically,
who becomes
With
errors.
statistics
we demonstrate
this chapter,
and fixed if
it
were,
using the
we show that even if the AFQT more strongly determine
life
poor.
reflects intelligence. at
poverty and
is
a
way
step, to correct their technical
no more complex than
that social contexts
theirs
and conceptual
—just more accurate
shape individual outcomes more than
we grant them the dubious assumption that AFQT Later in the chapter, we step back to take a broader
does intelligence, even
Curve
even
follow Herrnstein and Murray's data analysis as closely as possible
and then go on, step by
look
that,
socioeconomic status of people's
parents and their broader social environment
outcomes
asser-
social
wrote Charles Murray in
The Bell Curve assumes. In chapter 3 we showed
intelligence is not well
how
if
its
causes.
to see the real
Redoing the survey analysis
dynamics
that sort out
in
The Bell
winners and losers in
America. Other scholars have also reexamined the NLSY, using somewhat different procedures than
we
did,
Herrnstein and Murray are wrong.
and have come
to the
same conclusion:
2
Herrnstein and Murray exaggerated the importance of intelligence, not
only by using a measure that largely reflects instruction, but also by defin-
Richard
Arum
coauthored
this chapter.
70
WHO WINS? WHO LOSES? ing social environment narrowly and incorrectly. For them, the educational attainment, income, and occupation of people's parents comprise the total-
of their family environment. We, however, recognize and demonstrate
ity
that the schools, live also
communities, regions, and social circles
shape their
which people
in
lives.
Herrnstein and Murray also exaggerated the relative importance of intel-
making several technical
ligence by
most
of these. Ironically,
tistically the
them
one of
importance of the
AFQT score.
bulk of our
By
statistical
work
intelligence
the years.
we accomplish
that people's ultimate fortunes
—
Which
that social context is
NLSY
predicts
Poor?
we
not
We show
their social envi-
more important than
tested
in the social sciences
An Overview
over
of the Analysis
respondents became poor in 1990? Herrnstein and Murray
who was poor
in 1990: the respondents'
and Murray claim that the
the rate of poverty for people
scored low on the
poverty between those
which one most strongly
attributes to see
the respondents' parents' socioeconomic status
this
have sim-
a larger end:
depend on
—has been repeatedly demonstrated
up a contest between two
who
We
2.
3
Who Becomes
stein
of their errors led
analyses will be found in appendix
ronments. This finding
set
to underestimate starest
taking these steps to reanalyze The Bell Curve's evidence,
many ways
correct
for presentation in this chapter.)
only refute Herrnstein and Murray, the
them
But the
We
importance of social environment. (The
to underestimate greatly the
plified the
errors in their analysis.
their errors led
who
AFQT who
AFQT
is
wins
AFQT
scores in 1980 or
(SES) around 1980. Herrn-
this
match. The difference
scored high on the
AFQT
in
versus those
greater than the difference in the rate of
scored high versus low on parental SES.
evidence they conclude that "natural intelligence"
is
the
On
key determi-
nant of inequality. This result, the contrast between the strong association
of
AFQT with
erty,
has
poverty and the
become
weak
association of parental
the central claim in Murray's defense of
How could sociologists
SES
with pov-
The Bell Curve. 4
and economists have been so blind
to the
impor-
tance of intelligence? Perhaps they have not been blind but have been too
scared to reveal this "truth." to
it
nor scared of
devoted
itself to
it.
The answer
Since the
late
is:
They have been
explaining economic outcomes.
by Christopher Jencks and
neither blind
1950s an entire school of research has
his associates is a
71
The 1972 book
Inequality
well-known example of such
CHAPTER
have incorporated intelligence
studies. Social scientists
AFQT— in
cluding specifically the
typically found that conventional
AFQT,
test scores
—
in-
work. 5 These researchers have
measures of intelligence, including the
large; intelligence as
measured
in
such
does not have an overriding influence on economic outcomes; and
scores are but one factor sive
their
are correlated with important outcomes, especially with education
and earnings. But the effects are not tests
4
among many
and well-known research
that
literature
test
shape inequality. 6 This exten-
made
us very skeptical of the
claims that were emanating from press coverage of The Bell Curve. Years of accumulated research, not ideology, nity's chilly reception of itself,
lies
behind the academic
commu-
On close examination of the book
The Bell Curve.
our skepticism turned out to be well founded.
We
redo the
analysis of the
statistical
NLSY
data in the following
sequence: •
We
review a few of the
explain •
how we
critical errors
Herrnstein and Murray
We replicate Herrnstein and Murray's basic finding — that differences in AFQT scores matter more than differences in parental SES in predicting the chances that NLSY respondents were poor in 1990 — and then show what the finding looks
like
once a few technical errors are corrected: The
structure of respondents' families of origin
•
made and
corrected them.
is
then about as important as
AFQT score in predicting poverty. We expand the notion of social background
to include aspects of the
communities respondents lived
show
in at the
time of the testing and then
that these factors affected poverty, too.
Together with
home
envi-
ronment they are equally important as AFQT. •
We
add into the explanation the formal educational experiences of the
respondents, and
becomes
at this
significantly
point social environment broadly understood
AFQT
more important than
as a predictor of
poverty. •
We
•
We close the reanalysis by
include two attributes of the respondents' communities in 1990 to
show
the importance of local conditions.
showing the
critical
of three factors that Herrnstein and Murray
importance to being poor
essentially ignored: being
female, being unmarried, and being a parent. In the end,
become one of
AFQT
scores
the less important explanations of respondents' poverty.
In defining social background, Herrnstein and
Murray overlooked some
aspects of families, such as their size, and completely ignored other ele-
ments of the environment outside the family ples' life chances,
that nonetheless shape peo-
such as the kind of schools they attended and local job
72
WHO WINS? WHO LOSES Parental
SES
Parents' educations Parents' occupational status Parents' income
Inequality Poverty
Unemployment
Social Problems Out-of-wedlock births Incarceration
Welfare dependence Divorce Injury "Idleness"
Herrnstein and Murray's
4.1.
Problems {Note: Solid
Model of
the
Causes of Inequality and Social
lines indicate strong effects;
or insignificant effects. Herrnstein and
reasons but do not examine
its
Murray
dashed
lines indicate
weak
also control for age for technical
effects. Source:
Authors' interpretation of The
Bell Curve)
opportunities.
Even our
analysis does not capture nearly
environment, because data are unavailable for of the environment
—
for example,
many
all
of the social
important dimensions
on people's social contacts
what you know, but who you know").
("it's
not
7
The authors of The Bell Curve seem blind to this broader meaning of the 8 They, and some of their critics, too, seemingly cannot raise their sights beyond the household to see obvious social influences on individuals. When, for example, a region enters a recession, many workers social.
plunge into poverty
—
individual and family have
little to
do with
it.
For
another example, residential segregation by race and class increase the risk of poverty for those
who
are isolated.
9
Yet Herrnstein and Murray's mea-
sures of social origin take none of such realities into account.
show below, their
the kinds of
communities people
chances regardless of their individual
traits.
We
As we
abilities
or other personal
10
can display the differences between the "natural inequality" expla-
nation of individual outcomes and the environmental one the following graphs. Figure 4.1
we propose
by the AFQT, strongly determines their inequality.
It
with
shows what Herrnstein and Murray claim
happens. People's genetically given intelligence, as measured in the
mines
will
live in substantially alter
their
economic outcomes and thus
also determines the chances that they
NLSY deter-
become
involved in problematic behavior. Herrnstein and Murray acknowledge
73
CHAPTER Parental
4
Home
Environment Parents' educations Parents' occupational status
Schooling
income Two-parent family Parents'
•
Years of school
•
Academic
track
Number of siblings Farm
Social Problems
Adolescent Community
Environment • •
Out-of-wedlock births
Cognitive skills
School composition Region of the country
•
Incarceration
AFQT
Welfare dependence Divorce Injury
Race Most
"Idleness"
relationships
vary in strength from
group
group
to
Inequality
Gender
• •
Our Model of
4.2.
have drawn parental
in
Poverty
Unemployment
the Causes of Inequality and Social Problems (Note:
only the strong effects to simplify the diagram.
We
assume
We that
home environment and adolescent community are correlated and adwe do not specifically analyze that correlation)
just for that in our analysis, but
that people's family origins, specifically their parents'
SES, has some
ef-
but those, they say, are weak.
fects,
Figure 4.2 displays our alternative. figure 4.
1
because the reality
is
It is
considerably more complex than
considerably more complex. Original con-
ditions include aspects of individuals' families (labeled here "Parental
Home
Environment") but also include aspects of the wider social milieux,
such as the region people grew up
in.
and gender because these are genetic
(We do
Origins also include people's race
traits that
take on social significance.
Herrnstein and Murray's procedure of examining inequality only.
Chapter 8
is
devoted to the subject of race and
the record in figure 4.2 that racial identity influences ple's
we among
not in this chapter examine racial differences because
home and community
origins affect the
ethnicity.
follow
whites
We note
amount and kind of schooling
they receive and, together with schooling, affect their cognitive skills
which we use the Figure 4.2 individuals'
is
AFQT as
for
most processes.) Peo-
—
for
a measure.
simplified for easier legibility by not displaying aspects of
contemporary
situations: their adult
community environment,
including where they live and the local unemployment rate; and their con-
temporary family situations, including whether they are married and have
74
WHO WINS? WHO LOSES children.
These conditions also
affect the
We
do include them
become model. They would appear
odds
poor, independent of the other factors in the
roughly in the center of figure 4.2,
at the risk
1
that people will
of considerable visual
clutter.
in the analyses that follow.
Adolescent environments, schooling, and contemporary environments shape people's economic circumstances. Finally, those circum-
in turn
stances affect the odds that people will encounter social problems.
ognize that figure 4.2
is still
current environment,
it
NLSY. For
rec-
a simplification. Aside from not displaying the
leaves out other social factors such as people's peer
groups. Also, elements listed in each box are limited by what in the
We
instance,
we have no
direct
is
available
measure of school
quality.
Figure 4.2 also simplifies the complex causality. For example, getting injured or divorced often leads people to to contrast
become
poor.
Still,
the figure serves
our perspective with the vastly oversimplified and mistaken one
presented in The Bell Curve.
It
also provides a road
map
to the analysis that
follows.
Fixing the Errors
we
NLSY
we needed to correct several errors in Herrnstein and Murray's analysis. One technical error actually led Herrnstein and Murray to wrc^restimate how much AFQT score affected the risks of being poor. We corrected it in Before
could use the
to explore the sources of inequality,
our analysis. 11 Yet the bulk of their errors led them to exaggerate the relative
importance of
AFQT
score by greatly underestimating the impor-
tance of social environment. erred in
how
importantly, Herrnstein and
Murray
they constructed the measure they call "parental SES," the
measure they use as the key uals' social
Most
environments.
—
essentially the only
—
indicator of individ-
We quickly review four kinds of errors that they
made. MISSING INFORMATION
Key information was missing 17 percent did not
no information on
list
for
many
of the respondents. For example,
their father's occupation.
their parents'
income
And
21 percent provided
for either 1978 or 1979, about
two-thirds of those because they were not living with their parents then,
so they were not even asked their parents' income. 12 Herrnstein and Murray
still
used those respondents
in their analysis, in effect
assigning each
of them the average parental income reported by the other respondents. M
But these respondents with missing information were not average respcn75
CHAPTER dents.
14
By
4
assigning them the average income, Herrnstein and Murray
rendered their measure of parental income less accurate and thus less likely
show up
to
this error
as an important influence
on becoming poor. 15
by using a more appropriate procedure
We
corrected
for handling the cases with
missing information. 16 RELIABILITY
The term
"reliability" refers, roughly, to
surement
is.
A metal
time after time;
it is
how
ing measures of the
same
and trustworthy a mea-
same measurement of an object
ruler will give us the
reliable.
stable
A ruler made of soft rubber will give us vary-
object;
it
is
not reliable. In this research, the
AFQT is more reliable than the measure of social environment. Whatever AFQT measures —intelligence, years of school, motivation, and so
the
on
—
same score
same person time after time. That is because many years of psychometric work have gone into making it reliable (see chapter 2). The components of the parental
SES
it
does so
reliably.
It
gives roughly the
For example, youth are only approxi-
scale tend to be less reliable.
mately accurate in estimating their parents' income. 17
The implication of
this difference in reliability
components of parental SES
is
this:
because
will
its
have a higher
between the
When two
equally associated with a third, the one of the
measured
And only 4 questions
measure SES, compared with the 105 for the AFQT.
are used to
the
for the
first
two
statistical correlation
measurement contains
Herrnstein and Murray set up, the
less
random
AFQT
and
attributes are both that is
more
reliably
than the other. That
is
error. So, in the face-off
AFQT has the critical advantage of much
greater reliability over parental SES. University of Minnesota public policy
professor Sanders
Korenman and Harvard
sociologist Christopher
Winship
repeated Herrnstein and Murray's analysis, using a statistical adjustment to correct for unreliability in the
AFQT and the parental SES
index.
By doing
number of Herrnstein and Murray's conclusions: Corunreliability, parental SES was often more important than the
so they reversed a rected for
AFQT. 18 This
is
one correction we did not do
others had done
it,
however, our
our analysis, partly because
partly because not all scholars agree
ness, and partly because so,
in
own
we have more
its
substantive concerns.
appropriate-
Had we done
reversal of Herrnstein and Murray's conclusions
would have been even stronger than sion, as well as others,
on
we
are
the
one we present below. In
more cautious
justifiably be.
76
in
our critique than
this deci-
we might
WHO WINS? WHO LOSES? WEIGHTING THE COMPONENTS OF PARENTAL SES Herrnstein and Murray discussed in detail their construction of the parental
SES
index, proudly announcing that they bent over
good measure of social
class
background
to
They added together mother's education,
did not.
backward
to build a
AFQT. But
run against the
they
father's education, the
head of household's occupation, and parental income, averaged for 1978 and 1979,
to create
an index. In principle, each of these four counted
equally in building the scale. In practice, they
made
errors that
let
the
two
education measures count for more than occupation and income. 19 But their real error
Here
is
was
to
even
try to construct
why: In creating
their
such an index.
four-component index of parental SES,
Herrnstein and Murray were in effect insisting that each component deter-
mined outcomes with equal weight. For example, having poor education, they
tacitly
poverty as having parents with a low income. But what is
wrong? What
if
having a father or a mother with
assumption
education does not
little
income substantially increases those chances? Lumping sures together in a scale
would wash out
precisely what happened. In the
parents earned
but
if this
chances of becoming poor, but having parents with a low
alter people's
is
mother with a
a
assumed, should be as much of a risk factor for
how much
the different
the effect of parental income. This
NLSY, how much money
makes a big difference
mea-
in his or
a respondent's
her chances of being poor,
education their parents had makes no difference. Herrnstein
SES index leads them to underestimate home environment, specifically income, for
and Murray's parental
the impor-
tance of parental
the risk of
poverty.
We our
corrected this error by simply using each component separately in
statistics.
rately.
There
is
no
statistical necessity for
combining measures
and there are often good reasons for looking
indices,
That
plaining
is
how we
who becomes
learned that parental income poor.
at the is
so important in ex-
(Even here our reanalysis leans
and Murray's favor. Income earned over two years, even ported, to
look
is
at
into
measures sepa-
if
in
Herrnstein
accurately re-
an unreliable measure of a family's affluence. Economists prefer
measures such as accumulated wealth or income received over
several years.)
OMITTED VARIABLE BIAS
The fearsome
label "omitted variable bias" refers to leaving important
causal factors out of an explanation. Such omissions
77
mean
that
one both
CHAPTER
4
misses the whole story and distorts the analysis of the causes that are in the
The bulk of our corrections to The Bell Curve analysis conproblem: Herrnstein and Murray left out many important features
explanation.
cern this
who
of the social environment that affect
them back
is at
risk of being poor.
We added
in.
For example, in trying to estimate the social class background of each respondent, Herrnstein and Murray included income (sort of; see above),
but they did not include the studies
show
that the
effective wealth
who grow up
more
number of siblings each respondent
had.
Many
siblings people have, the lower their families'
and the lower
own
their
chances of getting ahead. People
with no or one sibling get more space, resources, and atten-
tion than those with several siblings.
20
We included the
number of siblings
We
each respondent had had in 1979 in our reanalysis.
also included
whether or not the respondent had been reared on a farm, because calculations of
income and needs
farm families (see next box), and
differ for
whether the respondent had grown up
in a two-parent family,
because that
strongly affects a family's long-term wealth.
The Bell Curve also ignored the community context within which people
were raised and within which they currently tion
the
live.
from which people come, the communities
communities
for example,
in
which they currently
live also
by providing schools of varying
Whatever the home
in
situa-
which they grow up and
shape their
quality,
fates, the first,
and the second, for
example, by providing jobs of varying quality. Herrnstein and Murray nored these sorts of environmental conditions;
made
we
ig-
included them, and they
a difference.
Researchers do
make
errors; in
any project as large as The Bell Curve,
they are bound to happen. Unfortunately, the great bulk of the errors Herrnstein and Murray cial
made
led
them
to underestimate
how much
the so-
environment changes the odds that someone will become poor. And
made a low AFQT score seem more important who becomes poor. We corrected these (and other) errors in our reanalysis; consequently we came to more accurate conclusions. Compounding the problems of measuring parental home environment and adolescent community environment are the problems of how to interpret the AFQT scores. As we argued in chapter 3, the AFQT captures a broad array that underestimate, in turn, in
determining
of instructional experiences that are related closely to both periods of youths' environments. In other words,
outcomes
are indirect via their effects
that is true, Herrnstein
the
some of
the social influences on
on the AFQT. To the extent
to
which
and Murray also overstate the role of intelligence
AFQT effects. 78
in
WHO WINS WHO LOSES Defining Poverty Deciding
who
government
should be counted as poor
ties the definition
were understood
complex
a
is
task.
The
federal
of poverty to "nutritional needs" as they
in the late 1960s. In
1968 a group headed by Dr. Mollie
Orshansky of the U.S. Social Security Administration took the Depart-
ment of Agriculture's calculations of how much tional
it
number
for other combinations of adults
and children and also took into
account that farm families could contribute their surveys
comes
cost to meet the nutri-
needs of a family of two adults and two children. They adjusted that
for food, Orshansky's
estimate families' total needs. erty lines" for
own
produce. Because
time showed that poor people used one-third of their
at the
group multiplied the cost of food by 3
The
result
was
farm and nonfarm families of varying
sizes.
Census Bureau has adjusted each of these poverty
account.
Since 1968 the
lines for
ratio
of
in nutrition, tastes, or other eating patterns
Nor has
total
changes
in
Nor
been taken into
there been any attempt to check the assumption that the
needs to nutritional needs
is 3.
Most researchers
the poverty line understates national rates of poverty; sify families
to
a table that defined the "pov-
overall prices, but not for changes in the prices of the specific foods.
have changes
in-
with incomes 25 percent over the
original poverty line for a
some
realize that
prefer to clas-
official line as
"poor."*
nonfarm family with two adults and two
The chil-
dren was an annual income of $3,477 (in 1968 currency); adjusting for inflation yields an official poverty line for a similar family in
1994 of
$15,029. * See, e.g., Jencks, "Is the
American Underclass Growing?"
Poverty, Test Scores,
and Parental Home Environment
how AFQT NLSY respondent was
Herrnstein and Murray's key statistical analysis compares scores and parental
poor
in 1990.
We
Murray got from
zAFQT
SES
influenced the odds that an
have already shown the
number of
score of intelligence.
in a
siblings,
two-parent family
at
chapter 2)
And we have
with their measures of parental SES.
ronment include
(in
how
correct answers on the
Herrnstein and
AFQT
just discussed the
Our measures of
parental
to their
problems
home
envi-
farm residence, and whether the respondent lived age fourteen. These together assess family back-
79
CHAPTER
4
ground more accurately and more powerfully than does Herrnstein and
shows how much more complex peothan Herrnstein and Murray acknowledge.
Murray's flawed index. This ple's family situations are
(Korenman and Winship,
home
including
attributes
set also
in their reanalysis,
went
farther than
we
do,
such as whether the respondent's mother
worked, what age she was when the respondent was born, and whether the family regularly received newspapers. The reader can with think of several other features of lives but
home environments
were not even asked about
work schedules, family wealth,
in the
that
NLSY—for
effort
little
shape children's
example, parents'
parents' physical and mental health, and
21 the involvement of other relatives.)
We begin our reanalysis by reproducing Herrnstein and Murray's results NLSY respondent was poor in 1990 (see
concerning the chances that an details in
appendix
Then we
2).
contrast that pattern with what
the corrected measures of parental stein
as a
home
environment.
and Murray's practice of graphing the results of the
We
we get using
follow Herrn-
statistical analysis
method of making comparisons.
Part (a) of figure 4.3 reproduces The Bell Curve's key finding (p. 134) as closely as possible.
22
Here
the probability that an
is
how to read the
graph:
The
NLSY respondent was poor in
axis displays variation in any particular causal variable,
deviations" below the
shows The horizontal from -2 "standard
vertical axis
1990.
mean (which is about where the bottom 5 percent of mean to +2 standard deviations above the mean
people are) through the
(where one finds the top 5 percent or
The
so).
solid, higher line in figure 4.3(a)
AFQT
respondents'
shows the association between
scores in 1980 and the probability of being poor in
1990, for respondents with average parental
measured
who
it
SES
as Herrnstein
and average ages (within the range of 25
scored -2 standard deviations below the
to 34).
mean on
the
and Murray
Respondents
AFQT
estimated probability of being poor of about 28 percent. That dents of average parental
On
SES and
average age
is,
respon-
scored that low on the
opposite page:
4.3.
Probability That an
Family Background:
(a)
NLSY
Respondent Was Poor
AFQT
score
—
solid line
number of siblings, farm
in
1990 by
AFQT
Score and
Herrnstein and Murray's Calculations; and (b) After correc-
tions {Note: All variables except either parental
to
who
had an
— were
set at their
home environment
mean
—dashed
line
—
or
values. Correction involves adding
residence, and two-parent household to the parental
SES
index
form the parental home environment index, and also replacing uniform weights for
index items with "effect proportional" weights. Source: Authors' analysis of data)
80
NLSY
WHO
WI NS
'
WHO LOSES
effects of AFQT scores and parental home environment on the probability that young white adults are poor: Herrnstein and Murray's original results show AFQT scores to be much more important, but corrections show that
The comparative
they are not.
(a)
Herrnstein and Murray's Calculations
30% o o As
£
AFQT score goes
from low
to
high
20%
o>
o
«
10%
As parental SES goes from low
0%
I
-2
to
high
I
-1
1
Standard Deviations from
(b)
Mean
Corrected Measures of the Effects
30% o As
So
AFQT score goes
from low
to
high
20%
o
~§
10% h
o
As parental home environment goes from low to high
0% -1
1
Standard Deviations from
81
Mean
CHAPTER
AFQT had a
.28 chance, better than
one
4
in four, of being poor.
respondents with average
AFQT
scores and then drops to about 2 percent,
a one-in-fifty chance, for very high scorers the
AFQT
parental
AFQT
mean. The lower, dashed
SES
—how
line
+2 standard
for
the risk of being poor varied for persons with average
who had
score and age. Respondents
parents of very low
who had
SES had very high
parents had about a 5 percent chance of being poor. Note the contrast
in the lines:
SES low
deviations above
shows the same information
a 14 percent chance of being poor in 1990, while those
SES
That proba-
drops sharply to between 7 and 8 percent, about one in thirteen, for
bility
line.
The
There
solid is
to very high
to very
AFQT line
steeper than the dashed parental
a twenty-one-point drop in the risk of poverty from very
AFQT score
versus only a nine-point drop from very low
high parental SES. This
intelligence "dominates"
(These
much
is
statistics are
SES
the basis of the trumpeted claim that
is
as an explanation of poverty.
derived from "logistic regression analysis." For a
brief overview of regression analysis, see appendix 2.)
But now
let's
tions. Part (b)
making
see what happens after
few elementary correc-
a
of figure 4.3 shows the consequence of correcting the mea-
sure of parental SES and also expanding it to be a fuller measure of parental home environment. Separating the SES components into the two parents'
education, occupation, and income;
and adding
data;
siblings,
ents in the respondents'
much
23
correcting the treatment of missing
farm residence, and whether there were two par-
homes
—
substantially increase our estimate of
family origin influenced the chances of becoming poor.
the risks of poverty about as
much
AFQT
as
we
how
influenced
scores did. In quantitative
terms, the corrected estimate of the family effect the corrected
It
AFQT effect (the AFQT effect also
is
86 percent as large as
increased a
because
little
corrected a technical error Herrnstein and Murray made). Comparing
extremes,
we see that,
all
else constant,
1
8 percent of people
from the poor-
est families grew up to be poor while only 4 percent of people from the
most advantaged families became poor. For technical reasons, even this home environment compared
correction underestimates the importance of
with
AFQT score. 24
Murray challenged ion, parental
critics
SES can be
of The Bell Curve to show how, in any fash-
as important as
that simple corrections of errors suffice to ited
measure of home environment
the
AFQT
measure
is.
is
AFQT meet
score.
25
We
have shown
that challenge.
Even
a lim-
as important a predictor of poverty as
Korenman and Winship
reanalysis of the data: "Estimates based on a variety of methods
82
from
also conclude, .
.
.
their
suggest
WHO WINS? WHO LOSES? that parental family
background
more important than [AFQT
is at least
as important,
and may be much
score] in determining social and
success in adulthood" (see also work by Dickens et
al.).
economic
26
Poverty and Communities People's social environments involve more than simply the financial, educational,
and demographic assets of their own families. 27 The
nity matters, too.
getting married at
commu-
local
Research shows, for example, that women's chances of
depend on the number of men
in the area
good wages. 28 The immediate neighborhood
who are employed
also affects people's
ways
life, whatever the family's own resources. This simple fact is one reason why Americans try so hard to find and afford "good neighborhoods." 29 It is one thing to come from a low-income family but live in a pleasant suburb
of
with parks, low crime, and quality schools, and another thing altogether to live in
an inner-city neighborhood that lacks those supports. The child
in a
well-endowed community gets the benefits of the locale regardless of his or her family's particular situation, just as the inner-city child bears the bur-
dens of a low-income community even
if his
or her family might have a
moderate income. Social scientists have increasingly taken
and have found compelling evidence
community context
seriously
that residential segregation
and the
concentration of the disadvantaged exacerbates the consequences of poverty,
family break up, crime, and deterioration. 30 (In chapter
the role of residential segregation
8,
we
will see
on minority achievement.) The concencommunities and particular
tration of the disadvantaged in particular
schools undermines the fortunes of otherwise able youth. Schools in low-
income and minority neighborhoods tend
to lack resources
struction; the concentration of children with
learning; and separation
and quality
in-
problems can distract from
by social class cuts poor children off from friend-
ships with advantaged ones. In the local neighborhoods, similar effects occur.
Low-income
areas have fewer jobs, fewer resources, and poorer-
quality services than
do
affluent ones. Local culture can clash with high
aspirations, public disorder creates insecurity, leave. 31
Other youth with similar personal
taged families but benefit
who
find
from the obverse of
From
the
NLSY,
themselves all
and similar disadvan-
good schools and neighborhoods
32 these conditions.
unfortunately,
the respondents' adolescent
in
and remaining opportunities
liabilities
we could
extract only
two measures of
community environments. One, 83
their region of
CHAPTER residence in the
some
first
differences in
specific.
year of the survey (1979),
economic and
is
quite global but captures
cultural conditions.
The other
is
quite
an index of school composition for the high school the
It is
spondent
4
re-
The index is composed of three indicators: the perbody that was nonwhite, the percent that was economi-
last attended.
cent of the student
cally disadvantaged (that qualified for free lunches),
and the dropout
rate.
33
(Direct measures of instructional resources and quality were unavailable.)
This measure
tells
us about the immediate context in which the respondents
spent their days as teenagers and, because schooling tells
us something about the local
score
—
is
locally organized,
as well.
The higher
the
the fewer dropouts, the fewer poor students, and the fewer black
students in.
community
—
the
(We do
more advantaged
mean
not
the
communities the respondents grew up
to suggest that
per se drags on achievement.
The
low-income and black students are
fact is that in the
United States schools
and neighborhoods where poor and minority youths are concentrated tend to lack all sorts of "social capital."
34
lems they bring from home, and
that contributes to a social climate that
may
Also, poor students tend to have prob-
interfere with attainment.)
These two
partial
measures of adolescent environments make a
statisti-
cally important difference in the respondents' probabilities of being poor.
Since uals
we have
—notably
"controlled" or "held constant" attributes of the individ-
their
AFQT
scores
—
these results
show
that the
wider com-
munity, as well as the specific family, shape individuals' economic fortunes.
The combined,
community 105 -question AFQT score. As
distinct effects of family of origin plus
of origin equal the distinct effect of the
figure 4.4 shows, the poverty rates predicted
cent community environments, on the other, are similar.
The
effect of
by parental home and adoles-
the one hand, and
by
AFQT
scores,
on
35
communities
is
even greater than appears
in figure 4.4,
because local conditions can temper or aggravate the consequences of getting low scores or
average
AFQT
poverty rate
if
coming from poor
families.
scores and average parental
For respondents with
home environments,
they had attended disadvantaged schools was
over ten points higher than for identical respondents
the
14 percent,
who had come from
advantaged schools, 4 percent. 36 But the contrast by school composition is
even greater for those
who had low
scores or poor parents, about
teen to twenty points difference in the chances of ing on whether the school
fif-
becoming poor depend-
was above or below average
in
composition. 37
Disadvantaged communities put even relatively advantaged individuals at risk;
advantaged communities help even the unfortunate. Disadvan-
84
WHO WINS? WHO LOSES
Adding adolescent community environment - school composition and region - to home environment shows social
the probability that
that
AFQT scores have equal effects on
environment and
young white adults
are poor.
30%
O o
As
M 20%
to
social environment - home and community - goes from low
g
high
(P
=
-.679)
/
M-H
o
%
10%
/
•§
As AFQT score goes from low to high ((3 =
0%
-.668)
-2-10
1
Standard Deviations from
Probability That an
4.4.
NLSY Respondent Was
Mean
Poor
in
1990 by
AFQT
Score and Social Background (Note: Social background includes parental
home environment and ing the plot, or
AFQT
adolescent community environment. For comput-
variables except either social background
all
—
score
thors' analysis of
solid line
NLSY
— were
set at their
mean
activity
line
Au-
data)
taged communities have schools where learning
economic
—dashed
values. Source:
is
hindered, they lack the
necessary to provide jobs, and they even limit the
chances of marrying out of poverty. Advantaged schools help residents in these
and other ways. 38 These calculations make the fundamentally
sociological point that individuals' fates are not theirs alone. Their
chances depend on their social surroundings
own
much
as
on
life
their
intelligence.
We sis,
at least as
could stop right here because, by repeating The Bell Curve's analy-
we have shown
that social
environment during childhood matters more
as a risk factor for poverty than Herrnstein and
85
Murray
report and that
it
CHAPTER matters statistically at least as
measure
much as do the test scores that purportedly we have been conservative in our re-
intelligence. (Recall that
analysis.) its
The key
finding of The Bell Curve turns out to be an artifact of
method. Although we could
go on, because our
will
4
rest
our case against The Bell Curve,
larger purpose
is
we
to explore the social sources of
inequality.
and Education
Poverty, Test Scores, In chapter 3
we showed
that the
AFQT
largely
motivation), but for the purposes of this chapter stein
and Murray's assumption that the
what
is
measured schooling (and
we have adopted
Herrn-
AFQT measures intelligence.
If so,
the role of schooling itself in determining the chances of being
poor? Social scientists have long established that people's educational
at-
tainments are the strongest immediate determinants of their economic fortunes. tion is
39
The
works
to say,
further
in part
you go
in school, the
by "transmitting" the
how much
more money you make. Educa-
effects of earlier experiences.
education people obtain
affluence and education of their parents. But attain is not the result only of
and what happens in them
and so on
—
—
also determine
itself
depends
how much
That
on the
in part
education students
such personal factors. Schools themselves
the instruction, the teachers, fellow students,
how much
students learn and
how
far they
go
(see chapter 7). Independent of talents or personal background, then,
schooling
is
We now
part of the social environment.
add formal schooling into our analysis. In estimating the
of education on poverty
we
distinguish between
the years of school respondents had completed before taking the the years completed after taking the
effect
two phases of education
AFQT. We
AFQT and
also distinguish between
an academic high school education and a general or vocational one.
how
Figure 4.5 shows
including education further accounts for the
chances that someone became poor. At
this stage in
our reanalysis, formal
schooling and childhood environment are each more important than score in predicting the risks of poverty. If one takes the
AFQT
AFQT as a measure
of intelligence, then figure 4.5 shows that intelligence does affect the odds of being poor. But
it
also
shows
that formal schooling
and childhood envi-
ronment matter more.
As we discussed in chapter 3, Herrnstein and Murray contend that how much formal education people obtain largely reflects their intelligence. One should, therefore, assign the effects of formal schooling in figure 4.5 86
^ WHO WINS? WHO LOSES
Adding education to the analysis gives a more complete description of what determines the probability that young white adults are poor: education and environment more than AFQT. ^U7o
Years of education and track (P = -.634)
Home and community
o
environment
£ m>
20%
AFQT (P
s
•9
X
s
\^ s^
O
^ 2
N
VV
=
-.628)
-.423)
v
^\^ -v
10%
^****'«««,.^^.
^"^.^V
no/
"^
~—
1 1
-2
1
-1
1
Standard Deviations from
4.5.
=
V
pa M-|
3a £o
(p
V
Probability That an
NLSY
Mean
Respondent Was Poor
in
1990 by
AFQT
Score, Social Environment, and Formal Education (Note: Social envi-
ronment here includes only parental home and adolescent community environments, not the contemporary environment; education includes years of schooling completed before taking the
and whether the respondent had been the plot, for,
each
all
in
AFQT,
years completed
variables in the equation were set at their
in separate estimates: social
after,
an academic track. For computing
mean
values, except
environment (dashed
line),
AFQT
score (black line), and education (gray line). Source: Authors' analysis of
NLSY
data)
to intelligence.
But we showed
likely that the opposite
premise
that, for a
is true,
not the cause, of formal education. the
AFQT
few reasons,
it
is
that test scores are the
We
much more
consequence,
should then interpret the effect of
in figure 4.5 as largely representing the effects
of schooling not captured by years or track
—
for
of those aspects
example, the quality of
teaching, the content of courses, and extracurricular instruction. servatively,
we could claim
instead that the
87
AFQT line
More con-
in figure 4.5
shows
CHAPTER
4
the effects of variation in respondents' verbal and
Whatever
able to their formal schooling. less important than
fecting the risk of
math
skills
not attribut-
the interpretation, such skills are
formal schooling and childhood environments in
becoming
af-
poor.
Adult Community Environment
To
this point,
we have
largely looked at
ronment in adolescence influences
their
how young
people's social envi-
chances of becoming poor in adult-
hood. But the contemporary community context surely matters, too. Most important, perhaps,
is
the local
economic
situation.
poor are higher for residents of communities
The only
measure we have of
direct
indicator of the
unemployment
The chances of being
in the
that in the
economic doldrums.
NLSY
is
an approximate
rate in the respondents' labor markets.
40
We
also can distinguish whether respondents lived in inner cities, suburbs, or rural areas. Together, these
plain
who became poor
in
two
rough as they
indicators,
are, further ex-
1990 (data not shown; see appendix
2).
41
Finer
geographical distinctions, by neighborhood perhaps, and further descriptions of
contemporary communities would no doubt have shown yet
stronger effects. But the key point for us tions
—
far
comes poor and who does alike, become poor.
not. If jobs depart,
Poverty, Gender,
that local
is
any recognition
economic condi-
—help determine who be-
more people, smart and dumb
and Adult Family Environment
Perhaps the most surprising omission of of poverty
is
beyond the control of any individual
that
women
all in
The Bell Curve's discussion
are far likelier to be
men. Figure 4.6 shows how great the gap was
in the
poor than are
NLSY.
It
compares
AFQT scores, holding constant age, education, and social environments. A young woman would have had to score fortyone points higher on the AFQT than a young man of the same age, formal men and women by
their
schooling, and background in order for her risk of being poor to have been as
low as
his.
Put another way, just being a
woman
raised a respondent's
AFQT points (the difference similar point can be made "bright"). A being being "dull" and between concerning parental income. Holding the AFQT and the other factors conrisk of poverty
stant, a
by the equivalent of forty-one
woman's
parents
would have had 88
to
have earned $63,000 more than
WHO WINS? WHO LOSES
7
A white woman needs to score 41 AFQT points higher than man from the same environment and with the same education in order to reduce her probability of being poor to his lower level. a white
30%
o o bp
20%
.S
Women
c
l-H
£
Jo
.
1 1
1
1800
1850
1
1900
1
2000
1950
Year
5.1.
Estimates of Inequality from
1
800
to
Today (Note: Lines
are authors'
estimates of trends described in sources listed in note 6)
mum-wage
legislation,
ing subsidies, and
expansion of public universities, the GI
many
growth and middle-class
Bill,
hous-
other programs that effectively brought economic life to
more Americans.
Inequality since the 1970s During the 1980s
politically
American middle
class
into rich ing,
and poor.
engaged academics argued over whether the was "disappearing" and the nation was splitting Defenders of freer markets denied that this was happen-
blaming misinterpreted data or the business cycle. 16 But the answer
clear now.
The
trend toward a
more equal
is
society that developed in the
twentieth century stopped in the early 1970s. Income divisions widened
and continued expanding through the 1980s. 107
CHAPTER
A Case of
5
Historical Amnesia: Caring for the Needy
While The Bell Curve's treatment of American history lematic, one of
America
its
greatest historical errors
dealt with the unfortunate.*
when neighbors provided
a time
is
generally prob-
the discussion of
is
The authors of The
how
a past
Bell Curve recall
the "safety net" that the poor needed.
all
But such memories are notoriously unreliable. There really was no such time
in
America.
If the
unfortunate were long-
time residents of the town, victims of bad luck, and considered morally upright, then neighbors might pass the hat, allot funds treasury, or
was
typically small
especially
(One aided
if
from the municipal
board the destitute with a local family. Even
and grudging.
If the
so, the
support
needy did not pass these
historian wrote of a colonial Puritan village that "poor persons if
tests,
they were newcomers, then usually nothing was forthcoming.
they were
members of
were
a townsman's family, otherwise they were
no matter how hungry they might be.")** From colonial
sent packing
when towns "warned out" newcomers whom residents feared might become public dependents, to the Great Depression, communities' major response was to move the needy out. days,
(We have
forgotten
cans used to be
how many Americans were on
much more mobile
than they are now.
off or losing a farm, hundreds of thousands
while and then
move
again.
the
moved on
move. Ameri-
Upon
to find
These people do not appear
in
being laid
work
for a
our
Andy
Hardy-like memories of small-town America, because, being poor and transient, they
But
were anonymous residents on the wrong side of the
their stories belie the nostalgic
image of the small town
that
tracks.
succored
the needy.)
The
local structure for helping the needy,
most American voters well enough sion destroyed that structure. residents also
came
became
Many
destitute.
pinched as
for generations.
it
was, satisfied
But the Great Depres-
long-time, middle-class, and "worthy"
Would-be charity-givers themselves be-
needy. Local philanthropies, even where supplemented by town and
county government contributions, could not keep up. In response,
New
Deal programs funnelled massive amounts of money into local communities directly
actually
and through work projects. Whether or not these programs
ended the Depression, federal intervention certainly supported
* See, especially, pp.
** Lockridge,
536-40.
A New England Town,
p. 15.
108
THE REWARDS OF THE GAME many of the needy through
the hard times.***
sustaining the poor, sick, and elderly has
By now,
become
the federal role in
even for
essential,
pri-
vate charities.****
Some have proposed
that
America today deal with
the destitute by re-
turning to private philanthropies and local communities the responsibility to
do
so, returning to
an earlier "safety net." Advocates of such policies
can appeal to nostalgia but they cannot appeal to
community shows
local
that
its
"safety net"
history.
The
history of the
was composed mostly of
holes.
*** For general overviews, see Katz, In the Shadow of the Poor House; Trattner,
From Poor Law the
to Welfare State; Keyssar,
Unemployed. For the history of the
Out of Work; and
New
of Reform; and Brock. Welfare, Democracy, and the ies
Sautter, Three
Cheers for
Deal's role, see also Chambers. Seedtime
New
Deal. For monographic stud-
of the needy in earlier eras, see, for example, Althschuler and Saltzgaber, "The
Limits of Responsibility": Vandal, "The Nineteenth-Century Municipal Responses to the
Problem of Poverty"; and Monkkonen (ed.). Walking to Work. **** Estimates are that replacing a proposed $400 billion cutback
grams
for the
in private
giving to charity (Steinfels,
"As Government Aid
In the first roughly twenty-five years after ilies
of
all
in federal
classes shared in
Evaporates*').
World War
II,
American fam-
economic growth; since 1970 only the
families have seen a significant rise in their standard of living. figure 1.1 the trend since 1959. In the ten years
incomes
pro-
needy by the year 2002 would require immense and improbable increases
for the richest
$36,000 per person
We
richest
saw
20 percent of households grew from $29,000
(in real
in
between 1959 and 1969, to
spending power); incomes for the middle 60
percent of households grew from $1 1,500 to $16,000 per person; incomes for the poorest
20 percent grew from $2,900
to
$5,400 per person. The
percentage increase for the poorest households was larger,
at 6.5
percent
growth, than the others. In the twenty years after 1969, the incomes of the top 20 percent have risen an additional $28,000 per person, the incomes of the middle
60 percent have gone up by a modest $4,600 per person, and the
incomes of the poorest 20 percent of households have actually
$200 per person. For several reasons discussed below, played
in figure 1.1
may
fallen
by
the pattern dis-
underestimate the declining conditions of the mid-
dle group.
These converging and then diverging fortunes are further indicated the data
on the shares of the nation's 109
total
income
that
went
in
to different
i
CHAPTER
5
From 1930 to 1970, the lowest-income 40% of American households began receiving almost as much of American income as the highest-income 5%, but that trend reversed after 1970.
30%
a,
25%
£ o u C »—
Share received by top
5%
-d
o 20% QJ en
O
X %
—*
15%
- --.
**
Share received by lowest 40%
C
'+—*:
aj
u jjj
10%
5%
0% 1940
1930
1950
1960
1970
1980
1990
Year
5.2.
Percentage of All Household Income Received by Highest-Income 5
Percent and Lowest-Income 40 Percent of Households, (Sources:
U.S.
Ryscavage,
Bureau of the Census, Historical
"A Surge
in
Growing Income
p.
301;
Inequality?"; U.S. Bureau of the
Census, "Income and Poverty: 1994, Highlights")
110
1930-1994
Statistics,
—
THE REWARDS OF THE GAME income groups. Figure 5.2 shows
1930, the highest-income 5 per-
that, in
cent of American households
income
that the
combined received over twice the total bottom 40 percent of American households combined re-
The highest-income 5 of every 100 families received about oneof all the income received by American households that year, while third the lowest-income 40 households out of 100 received one-eighth of naceived.
tional
income. By 1968 equalization had developed to the point that the
bottom two-fifths
finally
brought
home almost
as
much
as the top one-
twentieth (15.3 percent versus 16.6 percent). Households between the 40th
and 5th percentiles had increased their share from 58
68 percent (not
to
shown). After the 1960s, however, the income share of the top 5 percent rose sharply and that of the lower
40 percent of families dropped
reversing the earlier trend toward equality.
tinued into the 1990s.
The unequalizing
sharply,
trend has con-
17
Economists and sociologists, working from different perspectives, have analyzed and reanalyzed the available data and have Inequality in
income expanded greatly
made more money and people the chances of
in the
after
come
to a consensus.
about 1970. People
bottom half made
low-income Americans moving
less. In
into the
at the
18
top
recent years,
middle class have
dropped and the chances of middle-class Americans moving into poverty have grown. 19
The growing
inequality in annual
income displayed
understates the magnitude of the change.
One
reason
in the figure actually
is
that
income accumulate as some families invest and others
gaps
in
pile
up debts.
annual
Wealth has become even more unequal than has annual income. From about 1975 to 1992, the wealthiest tional
1
percent of families' share of the na-
household wealth rose from about 22 percent to about 42 percent;
in
the 1980s the wealth held
by the poorest 40 percent of families actually
dropped
The middle
in absolute value.
class has also felt the strain. In the
1960s the middle one-third of Americans saved about 5 percent of their annual income; in the 1980s they saved virtually nothing. 20
Another reason the annual income changes shown state the
change
is
that the increase in inequality has
in figure 5.2 under-
been most acute
in the
younger generations. The proportion of young men earning a "family
wage"
—
that
is,
enough income
to
keep a family of four out of poverty
has fallen sharply since the 1970s. Figure 5.3 shows the proportion of time, year-round least a
family
employed men, divided
wage from 1964
to 1994.
into a
We
full-
few types, who earned
at
can see that the proportion that
earned enough went up substantially from 1964 to 1974. But then the proportion declined.
It
declined dramatically for high school dropouts and for
111
CHAPTER
5
The proportion of full-time employed men whose earnings could keep a family out of poverty rose until 1974 and then dropped.
100%
Age
25- -34 "---.
""•
/ /
—
'
^
^
/
/
90%
N
m
-^^^^^
// /All men
• if
80% -
^^^
\High school dropout
„.•_
^^^. »»
-^ \
•
*'
\
/
V_
x /
A
\
/
/Age 18-24
\
/
\ / \ /
70% _
\
/ /
\
/ /
60%
i
i
1964
1969
1974
1
1
1979
1984
1
1
1989
1994
Year
5.3.
Percentage of Full-Time Workers
Who
Earned Enough
to
Keep
Family of Four Out of Poverty, 1964-1994 (Note: The measure
is
a
the
percentage of full-time workers whose annual earnings exceed the poverty line for a family
with
Low
of four. Sources: U.S. Bureau of the Census, "Workers
Earnings"; and Jack McNeil, U.S. Bureau of the Census, unpub-
lished tables,
November 1995)
112
THE REWARDS OF THE GAME 18-to-24-year olds, but larly,
men
even declined for workers aged 25
it
to 34.
21
Simi-
turning 30 around 1990 were notably less likely to have already
attained a middle-class
income than were men who had turned 30 around
1980. 22 These figures underestimate the sinking fortunes of workers be-
cause they count only fully employed men;
been
employed.
fully
ladder,
in recent years,
fewer
men have
Less able to save, less able to move up a career
younger men are seeing differences
widening yet in
23
further. All these
earnings and wealth
in lifetime
compounding changes mean
economic security and independence has accelerated
portion of Americans, aside from the elderly,
who
that inequality
The pro-
sharply.
are poor has increased.
Unlike the leveling that accompanied the economic growth of the 1950s, the
economic growth of the
late
1980s failed to stop or seriously slow
this
unequalizing trend.
Although young workers have been the most vulnerable
to the
economic
dislocations since 1973, older male workers have also been affected. This
can be seen most clearly in figure annual income (cost-adjusted) of
5.4.
It
men
shows what has happened
as they
grew
to the
older, contrasting the
1950s and 1960s with the 1970s and 1980s. The gray arrows show what
happened
income they with,
started
started the
decade with and the income they ended the decade
now aged
started
each decade aged between 25 and 35
the
men who
35-45. The black arrows show that pattern for
each decade aged between 35 and 45. In the period before 1973,
men could
expect to earn more, on average,
they did at the beginning. For example, in
—
men who
to
1950 made about $12,000
at the
end of the decade than
men who were between 25 and
in constant dollars,
35
and by 1960 they averaged
men who had started the men who had had stagnant incomes. And
about $17,000. However, from 1973 to 1983,
decade aged
35^5
(black arrow) saw drops in income;
started the
decade aged 25-35 (gray arrow)
from 1983
to 1993,
men
in the
35^5
cohort had
25-35 cohort had growing incomes, but
flat
incomes.
that cohort of 30-ish
Men men
in the
started
down that even the gains they had made by 1993 left them behind men who had been the same age in 1973. The assumption that men will move up the salary ladder as they mature now seems untenable. M.I.T. economist Frank Levy, who developed the original version of figure 5.4, offers the image of "yesterday's $25,000 steelworker who now so far
clerks in a
K-Mart
at
24 $4.25 an hour." But
declining industries like steel had to sectors of the
economy were
move on
displaced.
pared job mobility
in the late
found that workers
in the
it
A
is
to
not only that workers in
new
jobs.
Men
1980s to mobility
in the early
1990s were more likely to lose jobs. 113
in
many
1995 Census Bureau report com1990s and
And
full-
CHAPTER
5
Before the 1970s, 30- and 40-year-old men made major gains in income over a decade, but that has not been true since 1970.
$30
5.4. ties,
r
Income Changes Over
a
Decade
for
25-35
35-40 years old
35-45
45-50 years old
Men
in
Their Thirties and For-
1950-1993 (Note: Modeled on Levy, Dollars and Dreams,
p. 81.
Sources: U.S. Bureau of the Census, Current Population Reports, various series;
and
Statistical Abstract
1994)
114
THE REWARDS OF THE GAME
who
time workers
saw
a
jobs more frequently ended up as part-time
lost their
workers; even those
who found
20 percent drop
in their
a full-time job after a period of joblessness
average weekly earnings. 25
There has been some controversy about the claim of growing inequality that
we
made. Few observers deny
just
have gotten much
that the rich
richer faster than everyone else in the United States. But the dissenting
voices claim that the economic situations of average and poor Americans
have also improved
decade or
in the last
so, if not as rapidly as those
of the
wealthy. If that were true, one might be able to claim that the fortunes of the
down
wealthy have trickled
Some
to the rest of
argue, for example, that the
Americans.
way
the
Bureau of Labor
Statistics
calculates the cost of living exaggerates the year-to-year cost increases,
much
perhaps by as
as 1.5 points. All the
numbers we have used
here,
of living, would therefore underpower and overestimate the growth in poverty. Instead of concluding that the median earnings of fully employed men dropped 12 percent in buying power from 1979 to 1994, we would
which adjust incomes for increases
in cost
estimate the growth in real earning
conclude that they rose 14 percent. Recalculating the price index would not
change our conclusion
that the
gap between
classes has increased, "but
instead of the usual story of the rich getting richer and the poor getting
new
poorer, the
exactly
how
basket" and
would be that the rich got a The debate over this claim can
story
held their own." 26
weigh apples and oranges
to
same
products.
Is,
seem
clear,
however: One
only to those
who
visit to
for example, a 1995
Is
a visit to an
a private practitioner?
that the debate over earning
is
are earning.
incomes today than before. 27
consumers' "market
as a 1975 television set but has stereo
sound and a cable socket a "cheaper" television set? doctor the same "value" as a
while the poor
get arcane, dealing with
in the typical
how to account for changes in
television set that costs the
lot richer
As noted
A second
is
earlier,
fewer
A
HMO
few points
power applies
men
are earning
that the "big-ticket" items of the
home and college education for the chilmuch more rapidly than the little items that
middle-class life-style, such as a dren, have risen in "real" cost fill
up the
market basket. Americans have had
statisticians'
to
harder and go into more debt to attain those pieces of the
Dream." 28
A third
with children thirds of
at
them 29
point
home
—a
is
that
between 1975 and 1993 many more wives
started working,
great
many because
from fewer than half
much
to over
two-
they believed, rightly or wrongly,
that maintaining a middle-class life-style required
ing jobs,
work a lot "American
it.
As mothers took pay-
of the free labor they had contributed to the household in 115
CHAPTER
5
cleaning, cooking, child care, and the like had to be purchased or done
without. If one takes these points into account, the decline in living stan-
dards has probably been greater, not
Another
among
than the crude numbers indicate.
less,
line of defense charges that the apparent increase in poverty
the nonelderly
is
Some commentators
also a statistical anomaly.
point out that often government aid to the poor such as food stamps
is
not
counted as income. But add into the calculations government assistance
and the trends are
still
the same. (Recall that the
much a problem of the poor,
not so
although
it is
problem of inequality
that, too, as
it is
a
is
problem
of most American families falling behind the wealthy, as well as behind their
own
expectations.) Others have argued that the increase in poverty
the product of family
women
change
—
the increase in divorce and in the
bearing children out of wedlock. This argument implies that the
structural situation is fine, the
problem
chosen to be poor by making poor
is
that
many women
decisions. This
life
divorce and single-parenthood are often
First,
is
number of
economic
of a
strain, typically the inability
Therefore,
much
of the family dislocation
fails.
not usually) the results of
(if
man
have, in effect,
argument also
to help support his family.
itself is the
consequence, not the
we found in chapter 4 that having been poor was the key determinant of whether a woman would later have a child out of wedlock.) Second, even if we were to assume that every cause, of economic dislocation. (Recall that
divorce and out-of-wedlock birth were not the result of economic distress,
such family changes account for only about one-third of the rise in inequality.
And,
at the
same time, other family changes have
the inequality trend.
More lower-income women
and they are having fewer children. trend toward greater inequality the
growing inequality
The
is real, it is
post- 1970 increase in
are
not for these
substantial,
and
it
down
working than before, last
two changes,
would have been more severe
of the rich outpacing an advancing the rest have been retreating.
If
actually slowed
is
still.
the
In sum,
not simply a matter
have been advancing and
field; the rich
30
income inequality
largely arose
from increas-
ing inequality in workers' earnings. (Increasing wealth inequality was, in great measure, the result of the escalation in the value of financial instru-
ments
relative to
owner-occupied homes during
especially those of younger men, have ticular,
this period.)
31
Earnings,
become much more unequal.
earning differences by education have widened. During
In par-
the 1980s
had graduated college but declined 8 percent for
men who men who had dropped out
men who had
only graduated from high
the hourly
wage
(adjusted for inflation) went up 13 percent for
of college, dropped 13 percent for school, and
plummeted 18 percent
for
116
male high school dropouts. 32
In-
THE REWARDS OF THE GAME equality in earnings has also expanded cation,
and workers
in all sorts
why wages have
explaining
among workers of comparable
diverged
more
is
difficult.
University of Massachusetts economist Barry Bluestone
why
ent theories for explaining tion
and
skill
more than
used
it
edu-
of industries have been affected. In turn,
the market
to.
33
One
seems
lists
ten differ-
rewarding educa-
to be
part of the explanation appears to
be technological change, especially computerization, which would make
more highly educated workers more valuable points to this as a factor, although
it
is still
Some evidence controversial claim. 34 Many
to
a
employers.
making them more complex
technological changes simplify jobs instead of
and therefore encourage employers to hire less-skilled rather than moreskilled workers.
turing jobs
35
move
"Deindustrialization"
the process
by which manufac-
—
America explains what happened in terms of reduced dework of the less-educated. Generally, competition from other workers has depressed middle Americans' wages. When, for ex-
service jobs in
mand
—
overseas or are automated, leaving more poorly paid
for the
nations'
ample, software programming can be done in India cost here,
it is
pressure seems not, however, to have depressed earlier.
at a fraction
of the
hard for American workers to demand higher pay. 36 (This
CEO
salaries, as
we saw
Indeed, Disney executive Michael Eisner banked a record $203
million the year his
company
suffered a 63 percent drop in profits.)
37
Also
contributing to the free-fall in less-skilled workers' wages are the weakening of unions, the stagnation in the
education (see chapter It is
minimum wage, and
cutbacks in higher
6).
important to understand that this widening inequality in earnings has
occurred, not only in the United States, but in most affluent Western nations during recent years. that
However, as we
compares the American experience
inequality has been greater and
anywhere
its
shall see in the section
to that
below
of other nations, expanding
consequences more severe here than
else (except perhaps for the United
Kingdom). 38
Analysts often treat inequality in earnings as simply the result of "market" operations and separate those effects from the effects of governments'
after-market interventions too.
But we
misleading.
By
its
first
—
taxes and transfer payments.
will
do
that,
note that the distinction between market and policy
The market
taxes and by
is
permeated with policy. 39 (See box,
its
"transfers"
—
income created by
—government can blunt
inequalities in earnings
American government today does reduce largely for older people.
117
is
p. 118.)
social security, Medicare, food
stamps, unemployment insurance, and the like equalities in
We
in-
from the market.
inequalities of income, but
CHAPTER
How the What people
5
"Free" Market Rests on Government Policy
earn in the labor market cannot be separated from policies
that structure the market.
We
discuss such policies in detail in the next
chapter, but pause here to consider just a
few examples of how government
policies shape difference in earnings:
—Licensing practice a
The
laws:
Governments
stipulate the requirements necessary to
wide range of professions, from hair-cutting
Imagine what would happen
their earnings.
earning 5 percent of Americans fessorate essarily
if
to neurosurgery.
fewer the practitioners, and the higher
tighter those requirements, the
to the
incomes of the top-
entry into medicine, or law, or the pro-
were made considerably easier? (As professors, we are not nec-
recommending
—Direct and urban
this
move.)
subsidies: Direct subsidies, such as agricultural
transit construction, divert
American economic development pended heavily on federal
gifts
farm supports
income from some people
to others.
in the early nineteenth century de-
of land to the states and on state borrowing
Democrats largely opposed
for infrastructure. (Jeffersonian-Jacksonian
such spending, and Hamiltonian Federalists, predecessors of today's
"Wall Street" Republicans, supported
it.)
Later, sizable subsidies to rail-
road companies spurred the interconnection of American towns.
—Laws governing property and
finance:
Fundamental
to the
economic
system are the laws that govern property. In the early nineteenth century,
American courts provided
critical rulings that
ited-liability corporations
and that protected businesses from paying for
the incidental
damages they caused. Different
different earnings. Later, the
Amendment's
enabled the creation of lim-
rulings
Supreme Court ruled
would have meant
that the Fourteenth
protections for people extended as well to corporations, fur-
ther aiding the expansion of the large corporate sector.
Other examples of rules, tax deductions,
even these few
income
is
far
policy shapes the market, such as unionization
and regulations,
illustrations
from a
show us
pristinely
will
be discussed
that the
way
in chapter 6.
But
the "market" apportions
economic process.
many of which may be
of policies, to
how
It is
embedded
so longstanding that
in a set
we assume them
be "natural." Nevertheless, they are policy choices. "There
is
no such
thing as a free lunch," the laissez-faire economists remind us. True, and there
is
also
no such thing as a "free" market.
118
THE REWARDS OE THE GAME
Since 1966, the poverty rate among the elderly has dropped it has increased among other Americans.
by half;
30%
r-
-
V
65+ years old ""^
£ 20%
o
/
^ /
/ Under 18
"' -'"
*«.-
"*^
\'~ c QJ
u 10%
18-64
.
PL.
0%
i i
1966
1970
i
1974
l
i
1978
1982
i
i
1986
1994
1990
Year
5.5.
Rates of Poverty by
Age Group, 1966-1994
(Sources: U.S. Bureau of
the Census, Current Population Reports, series P-60, no. 178, and series
P-23. no. 188; and "Income and Poverty: 1994, Highlights")
Not so long ago, older people were poorer than
the rest of Americans,
and their poverty was a social problem of wide concern. But, as
among By 1994 under 12
we
first
pointed out in chapter 4, poverty
older people has dropped dramati-
cally (see figure 5.5).
percent of older people were poor,
compared with 15 percent of
the nonelderly.
this reversal is social security.
For decades,
retire early.
In
1940, 42 percent of older
17 percent did. Then, in
more recent
The
it
single major reason for
has allowed older people to
men worked;
in
1994, only
years, cost-of-living adjustments ac-
celerated the increase in social security payments. In the 1980s, while the
median income of
all
American households increased
just 5 percent, the
median income of older households increased 20 percent. 40 By 1993 the wealth of the typical older household
age.
41
was over twice
the national aver-
America's social policy has successfully fought the "war on
poverty"
— but mostly on behalf of older people. 119
CHAPTER
5
Inequality has widened since 1970 not just in money, but also in the quality of life
money can
buy.
something Americans value
homeownership fell
We
see
it,
for example, in
homeownership,
Between 1983 and 1994 rates of for every age group under 60 while rising for every age greatly.
group over 60. For example, the percentage of heads of household 40-44 but the
who owned
homes dropped from 73 percent to 68 percent, percentage of those aged 70-74 who owned their homes grew from
years old
their
75 percent to 80 percent. 42 People's sense of security
relies in part
on hav-
ing health insurance, to take another example, but the proportion of Ameri-
cans covered by any health insurance (increasingly caid) for an entire year has fallen recently,
1993 and
to 85 percent in
is still
this
falling (though not
among
covered as they are by Medicare). 43 More dramatically of men
has meant Medi-
from about 87 percent
in
1987
older people,
about one-third
yet,
who changed jobs recently found that they had lost health insurance
in the process.
a widening
44
gap
denned broadly
Also paralleling the widening income and wealth gaps
and mortality among social
in health
in
such ways
classes.
45
is
Inequality
also growing.
is
Conclusion
What have we history?
We
learned about inequality from this quick look at American
see that systems of inequality are changeable. In
little
more
than a generation, Americans became notably more equal and then notably
more unequal. These changes cannot be explained by changes in individuals' "natural" talents, be it IQ or other inherent traits. (Changes in Americans' acquired traits probably played a role. Far more Americans were college graduates in the 1980s than in the 1940s.) Even if, as some testing data suggest, Americans
became "smarter" during
(see chapter 2), inequality
the changes
was
changed
just too great.
the twentieth century
in different directions,
and the scale of
Market forces also cannot explain these
variations in inequality. Certainly, technological developments and global
trade have helped shape inequality in the United States, but
under the control of policy
and policy that corrects
when we compare
—
how
policy that structures
how
much
has been
the market
works
the market works. This point emerges again
the United States with other nations.
Inequality Here and There
A glance
behind us to American history shows that our pattern of inequal-
ity is far
from fixed or naturally determined. 120
A
glance sideways to other
THE REWARDS OF THE GAME wealthy nations makes the same point. The United States has the greatest degree of economic inequality of any developed country. inequality that
nomic conditions but
we
will
and
capitalist
compare
have ways
to
is
is
are our competitors in the global market
this chapter,
CEOs
numbers
The
best and latest evidence on
comes from
—and
yet they
noted
how much wider
the
in the is
the
gap
United States than
in it
greater even than these
indicate. In general, our high earners earn relatively
more and our low earners earn
is
we
and average workers
elsewhere. America's distinctive inequality
illustrative
The nations with which
the result of policy choices.
reduce inequality and remain competitive.
between
is
a level of
the United States are also modern, affluent, democratic,
—they
At the beginning of earnings
It is
not fated by Americans' talents nor necessitated by eco-
is
do workers elsewhere.
relatively less than
how
nations
compare
Luxembourg Income Study
(so
in levels
of inequality
named because
the project
headquartered in Luxembourg). Social scientists affiliated with the study
have collected detailed, comparable data on earnings and income from over a dozen nations.
Our
first
use of their research appears in figure 5.6, which
speaks to the question of inequality in earnings, specifically earnings of
men, aged 25-54, who worked
full-time, all year during the
mid-
to late
1980s. (Comparable data on earnings were available for only five nations.
We
are looking just at
labor force
was
in
men
here, because the situation of
such flux and varied so
vertical line in the figure serves as
each nation.
It
men
at the
median
90th percentile in earnings
—
line display the ratio
to the earnings of
men
median worker earned. The bars stretching
same comparison between
the
The
at inequality in
The
of the earnings
near, but not at the top of, the earnings ladder received
1.8
—those
median. In
at the
1986 the 90th percentile American male worker earned the
nations.)
in the 46
represents the earnings of the average (median) worker.
horizontal bars to the left of the that
much among
an anchor for looking
women
times what
to the right represent the
median earner and a low-paid worker, one at median worker
the 10th percentile of earnings. In the United States, the
brought
home
2.8 times the
amount
the 10th percentile
left-hand bars, therefore, display inequality of earnings
worker
did.
The
between the high-
earners and the average; the right-hand bars display inequality between the
average and the low-earners. Together, they display
total inequality. In the
United States, the 90th percentile worker earned five times that of the 10th percentile worker.
These numbers are highest
in the
United States. That
ings between the rich and the average worker
where, as
is
the
is
is,
the
gap
in earn-
greater here than else-
gap between the average and the low-paid worker. The
contrast between the United States and Europe sharpens further
121
when non-
CHAPTER
5
The gap between the highest- and the average-earning men was widest in the United States — as was the gap between the average- and the lowest-earning men. Gap between high-earner and average:
ratio of 90
percentile to
low-earner: ratio of median th to 10 percentile
median
Median
9Qth
United States
Gap between average and
th
jQth
\Z
1
Canada
I
Australia
Vest
1
Germany
Netherlands
Sweden I
i
i
l
l
i
i
Ratio
5.6.
Ratios of Earnings for High-, Median-, and Low-Earners in Six Na-
tions (Source:
Adapted from Gottschalk and Smeeding, "Crossnational
Comparisons," table
1)
monetary compensation
is
added
to the picture. In
national law requires that virtually
all
most European
nations,
workers have the kinds of benefits
such as strong job security and four-week vacations that in the United States only workers with seniority in major firms have.
These national differences expanded
in the 1980s,
47
when
inequality in-
creased globally. International economic forces widened the gaps be-
tween what the
better-
ized nations, but this
and the worse-educated earned
in
most
industrial-
chasm opened up farthest and fastest in the United Kingdom. (These were the years of Thatcherite
States and the United
reforms that reduced the role of government in the United Kingdom.) Elsewhere, the gap in earnings between the better- and worse-educated
widened
less,
barely at
all,
or even narrowed. There seems no clear con-
nection between these differences and other economic trends such as
122
— THE REWARDS OF THE GAME growth
The reasons
rates.
power of unions and ern countries.
lie in
government
policies, notably the relative
the expansions of higher education in the other West-
48
The biggest
income inequality between
contrast in
the United States
of the developed world, however, appears after taking into
and the
rest
account
how government
means
deals with the results of the market. That
accounting for taxes, tax deductions, transfer payments, housing subsidies,
and the
like.
we
(Again,
note that this before- and after-government distinc-
tion underestimates the role of government.
ments require employers
Where,
for
example, govern-
more market
to provide certain benefits, there is
equality.)
to
To look at international differences in household income, we turn again the Luxembourg Income Study. Peter Gottschalk and Timothy Smeed-
ing compiled comparable data on households' disposable
income
after taxes
seventeen nations.
in
incomes
and government support, adjusted for household 49
we
In Figure 5.7,
size
use just the figures for nations
with over ten million residents in 1980; our conclusions about the United States
As
would be
same
virtually the
if
in figure 5.6, the bars to the left
household's income
(at the
we showed
the smaller nations, too.
50
of the median display the ratio of a rich
90th percentile) to an average one's income,
while the right-hand bars show the ratio of an average household's income to that of a
in the
poor one
( 1
United States,
0th percentile). 5
unequal nations (4.0 for
Italy,
The rich-to-average
average-to-poor
2.1, as is the
rich-to-poor ratio, 5.9 (not shown)
1
is
much
and so the
higher than that of the next most
Canada, and Australia). 52 In
States has the greatest degree of
ratio is greatest
ratio, 2.9,
income inequality
in the
short, the
United
West whether one
focuses on the gap between the poor and the middle or the gap between the
middle and the
rich.
Even
tiveness, because they
these
numbers underestimate America's
do not count the
and lower-income families receive
in
distinc-
sorts of "in-kind" help that middle-
most other nations, such
as free
health care, child care, and subsidized housing and transportation. also underestimate inequality in tion of
income
at the
America by not displaying
very top of the income ladder.
Western nations generally take two routes discuss in chapter 6,
They
the concentra-
some
to
reducing inequality.
As we
intervene in the market to ensure relatively equal
distributions of earnings by, for example, brokering nationwide
wage
agreements, assisting unions, or providing free child care. Others use taxes
and government benefits
few do both
to
reduce inequality of income after the market.
seriously, such as the Scandinavian countries.
States does the least of either. If one sets aside older people,
123
A
The United
who
benefit a
CHAPTER
5
The income gap between the richest and the average household and the gap between the average household and the poorest are both wider in the United States than elsewhere.
Gap between ratio of 90
th
rich
and average: median
percentile to
90 th
Gap between average and poor: ratio of
median
to 10
th
percentile
Median
10 tb
United States \Z
i
Canada
i
Australia
U.
i
Kingdom Italv
France
i
i
1
[
i
W. Germany
i
Netherlands i
i
i
i
i
i
i
Ratio
Ratios of Incomes for High-, Median-, and
5.7. in
Low-Income Households
Eight Nations (Source: Adapted from Gottschalk and Smeeding,
"Crossnational Comparisons," table 3)
great deal
from government action
taxes and transfers here
changed from the way
it
is
in the
to leave the
United States, the net effect of
degree of inequality virtually un-
was determined by market
earnings.
53
(See also
chapter 6.)
When
everything
income inequality
is
is
accounted
for, the
Western nation with the most
the United States. But the United States
is
also ex-
ceptionally unequal in terms of wealth. At the end of the 1980s, the richest 1
percent of families
owned about 40 124
percent of household wealth here,
THE REWARDS OF THE GAME more than
in
any other advanced nation; the richest
1
percent
owned only
25 percent of the wealth in Canada and 18 percent of the wealth Britain, for
example. 54
Add
in
Great
the less tangible features of "wealth," such as
vacations and security of medical care, and the conclusion
reinforced that
is
Americans are remarkably unequal.
(Some critics of crossnational comparisons contend
one ought not
to
contrast the United States to other nations, because the United States
is
distinct in certain ple.
ways.
We have so many single-parent families,
But even looking only
unusually unequal.
55
that
at
for
exam-
two-parent families, the United States
is still
America
also
seems exceptionally diverse
racially
and ethnically. But other Western nations also have ethnic diversity, the racial caste still
system we do.
exceeds that
And
among white
poverty
among American
if
not
whites only
or majority populations elsewhere. 56 Such
reservations do not challenge the conclusion that the United States
is
un-
usually unequal.)
America's level of inequality
is
by the distribution of its people's
by design.
talents,
It is
not given by nature, nor
nor by the demands of a "natural"
market. Other Western nations face the same global competition that
and are about as affluent as
we
are and yet
have managed
terns of inequality less divisive than ours. Ironically, that
Americans were proud of comparing
to the class-riven, hierarchical,
it
to
we do
develop pat-
was not so long ago
their relatively egalitarian society
decadent societies of Europe. In the
last
couple of decades, America has become the more class-riven and hierarchical society.
57
The United
States
is
unusually unequal and Americans are unusually
show that Americans back moves toward expanding opportunity but oppose moves toward equalizing outcomes. They endorse wage differences among jobs that are pretty similar to the wage differences that they believe exist today (although the real differsupportive of this inequality. Surveys
ences are greater than Americans imagine), and they do not approve of
government programs six nations,
to
narrow those differences.
In a survey of people in
only 28 percent of Americans agreed that government should
reduce income differences. The next lowest percentage was 42 percent (Australians), while in the other countries majorities supported reducing
income differences. 58 Whether we have as much opportunity as Americans want
is
debatable (see chapters 4 and
8),
between the desired and the perceived
may be because Americans
but
we seem to have
level of
outcome
a rough
think that considerable inequality
stimulating productivity and a high standard of living.
125
match
inequality. That
Is it?
is
needed
for
CHAPTER
Is
Inequality the Price of Growth?
Some commentators equality.
A
5
straightforwardly defend our current level of in-
congressional report in 1995 conceded that the recent trends
toward inequality were real but argued, "All societies have unequal wealth
and income dispersion, and there degree of market determined
is
[sic]
no positive basis for inequality."
59
criticizing
Disparities in
any
income
and wealth, some analysts argue, encourage hard work and saving. The rich, in particular,
jobs for
all.
60
can invest their capital in production and thus create
This was the argument of "supply-side" economics in the
1980s, that rewarding the wealthy
—
for example,
taxes on returns from their investments benefit of
The 1980s did not work out
all.
the theory
is
still
We
influential.
these analysts say, but doing so
by reducing income
—would stimulate growth that way, as
we have
to the
seen, but
could force more equal outcomes,
would reduce
living standards for all
Americans.
Must we have
so
much
nomic research concludes
inequality for overall growth? not,
it
even suggests
economic growth. In a detailed
statistical analysis,
Persson and Guido Tabellini reported finding
had more inequality of earnings tended
that
to
The
that inequality
latest eco-
may
retard
economists Torsten
that, historically, societies
have lower, not higher, sub-
sequent economic growth. Replications by other scholars substantiated the finding: eties.
61
More unequal That
fits
nations grew less quickly than did
more casual observations
as well:
We
more equal socisaw that, in the
United States, our era of greatest recent growth was also an era of greater equalization. ica's
62
economic
give us
stiff
And we rivals
saw, at the beginning of this chapter, that
do not need
to
pay
their
CEOs
Amer-
exorbitant salaries to
competition. In fact, during the 1970s and 1980s, America's
national wealth did not
pean nations.
grow
as fast as that of the
more
egalitarian Euro-
63
Close examination of detailed policies also suggests that greater equality helps, or at least does not harm, productivity. Researchers affiliated with
the National
economic
Bureau of Economic Research closely examined the
flexibility (that is, the ability to shift resources to
tive uses) of several redistributive policies
effects
on
more produc-
used by Western nations
—job
homeowner subsidies, health plans, public child care, and so They found that such programs did not inhibit the functioning of those
security laws,
on.
economies. 64 Indeed, a study of over one hundred U.S. businesses found
126
— THE REWARDS OF THE GAME that the smaller the
wage gap between managers and workers,
the business's product quality.
how
This recent research has not demonstrated precisely helps economic growth,
66
we can
but
the higher
65
greater equality
consider a few possibilities. Increas-
ing resources for those of lower income might, by raising health, educational attainment,
and hope, increase people's
abilities to
Reducing the income of those
entrepreneurial.
at the
be productive and
top might reduce un-
productive and speculative spending. Take, as a concrete example, the
American corporations ones.
are run
The American companies
whose main
responsibility
prices to shareholders and to
by largely autonomous managers
are run
is to
way
compared with German and Japanese return short-term profits and high stock
—because they
are often paid in stock options
themselves as well. Japanese and German managers are more
like top
employees whose goals largely focus on keeping the company a thriving enterprise.
The
latter is
more conducive
to reinvesting profits
long-term growth, 67 Whatever the mechanisms to
may
undermine growth. Americans certainly need not
accept the high levels of inequality
we
and thus to
be, inequality appears feel that they
must
currently endure in order to have a
robust economy.
A related concern for Americans is whether "leveling" stifles the drive to get ahead.
Americans prefer
encourage Horatio Alger striving and to
to
provide opportunities for everyone. Lincoln once said "that
shows
rich
that others
may become
believe that inequality if so,
how much
is
needed
inequality
is
to
rich."
68
Many,
if
some would be
not most, Americans
encourage people to work hard. 69 But,
needed?
For decades, sociologists have been comparing the patterns of social mobility across societies, asking: In which countries are people most likely to
overcome
lar,
the disadvantages of birth and
move up
the ladder? In particu-
does more or less equality encourage such an "open" society? The an-
swer
is
that
Western societies vary
little in
economic successes are constrained by
the degree to their
which children's
parents' class positions.
America, the most unequal Western society, has somewhat more intergenerational mobility than
most equal Western in this
In
society.
70
do other
There
is
nations, but so does
no case for encouraging inequality
evidence, either.
sum, the assumption that considerable inequality
even encourages, economic growth appears to
fluid
Sweden, the
make
equality;
a morally wrenching choice
we can have
both.
But even
if
127
to
be
false.
is
needed
We
for,
or
do not need
between more affluence and more such a choice were necessary, both
CHAPTER sides of the debate, the "altruists"
and the supposed "realists"
who
who resist
5
favor intervention for equalizing it,
agree that inequality can be
shaped by policy decisions: wittingly or unwittingly,
we choose our
level
of inequality.
Conclusion Either
we Americans have come to desire the inequality we live with or the we live with reflects our desires (or both). Certainly, there are
inequality
values and beliefs
—individualism,
more unequal
creating a
society.
capitalism,
freedom
—
that
can justify
But many Americans also believe
that
such levels of inequality are inevitable, the result of inequality in natural ability or
come
of the market or both. This belief has no basis in evidence. In-
and,
more
generally, wealth, standard of living, and quality of life
have been more equal in other times and are more equal
in other places,
without any evidence that talents were more equal earlier or are more equal elsewhere, or that market pressures are different there.
Perhaps today's trend toward inequality will reverse; fore in
American
But the keys equality
we
history, as
economic and
with
is
nomic conditions strongly influence the shape of its
changed.
The degree of
not a "natural" result of either inherent
talents or a "free" market. Certainly, people's skills
acting through
has reversed be-
political conditions
to understanding inequality will remain.
live
it
and
in-
human
societies' eco-
inequality.
But a people,
government, can contract or expand that inequality.
The
policy changes enacted by the 1995-96 Congress will, certainly in the short
run and most probably in the long run, widen inequality. choices
we have
in
more
detail in the next chapter.
128
We
explore the
CHAPTER
*
How
*
6
Unequal?
America's Invisible Policy Choices
A,.mericans
can significantly
them. In chapter
much
it
varies
5,
we showed how
from nation
how much
inequality
to nation.
from changes and variations specific
alter
Such
that
chapter
more
measure
focus on several
as welfare spending, are not the
even the most important ones,
"invisible" practices are
we
is
shape inequality.
Obvious redistributive programs, such only policies, or
fluidity results in large
in policy. In this
American policy choices
among changes over time and how inequality there
that affect inequality.
Many
For example, American housing
significant.
and road-building programs have largely subsidized the expansion of suburban homeownership for the middle class. Other largely unnoticed policies set the ball,
ground rules for the competition
batters
have the advantage (see box,
regulations favor last
to get ahead. Just as in base-
where the height of the pitcher's mound p. 130),
affects
whether pitchers or
so in the marketplace laws and
some competitors and disadvantage
We saw in the
others.
chapter that the United States has the greatest inequality in earnings
among full-time workers and that that inequality has increased since 1970. Some policies narrow inequality and some widen it. Again and again we will see that the basic dimensions of social inequality how rich the rich are a are and how poor the poor are, and even who becomes rich or poor
—
result
of our social and political choices.
directly,
and hence
invisibly.
Many
The programs
—
of our policies operate
that help the
in-
poor are glaringly
obvious, but those that aid the rich and middle class tend to be invisible.
Obscured even more are the policies
we
labor market. In this chapter,
game" for the many ways that
that set the rules of "the
will reveal
some of
the
social policy shapes inequality.
We icy,
will begin
which
is
by looking
at
to provide, with
one general pattern of American social pol-
one hand, limited direct help
to
some of
the
poor and indirectly to subsidize, with another, the middle class and the wealthy. Next,
we
will
uncover one of the most hidden arenas of social
policy, the regulation of the labor market,
and show
how
the
ground
rules
shape inequality. Finally, through an examination of higher education, we 129
CHAPTER Baseball: and
In sports, talent
determine ceed.
We
who
How
6
Rules Help Pick the Winners
effort should
who
determine
wins. But the rules also
has the advantage and even the kinds of players
who
suc-
see that in the history of baseball. In 1893 the team owners
lengthened the pitching distance to sixty feet and pitchers
lost
an edge;
game the next year. Then, game plummeted (to seven
runs soared to an average of fifteen per
number of runs per
early 1900s, the
National League) because
counted the
first
two
partly determined
new
rules created a wider
foul balls as strikes.
who became
home
in the in the
plate
These new regulations,
and
in turn,
successful players. Baseball historian Ben-
jamin Rader writes:
With the lengthening of the pitching distance
in
mound"), the sheer
pounds tall,
the
5' 11")
size of pitchers
were
the pitchers
same height
and stood
By
as the hitters.
averaged an inch and a half
average of 9 pounds heavier (180
taller
of
dirt ("the
1894
to increase sharply. In
relatively small; they averaged
lighter than the hitters)
pitchers of 1908
began
1893 and the grow-
mound
ing practice of placing the pitching slab on a
at five feet
168 pounds (4
and ten inches
1908, however, pitchers
(at
than the hitters and were an
lbs. vs.
171
lbs.).
Notice that the
weighed a whopping 12 pounds more than the aver-
age of their counterparts in 1894.
Following another hitting drought
in the
mid-1960s, baseball owners
again attempted to right the balance by lowering the pitching fifteen to ten inches, ordering
home
fences closer to
plate,
umpires to tighten the
and
in the
ignated hitter" to bat for the pitcher.*
changing rules
molds
why is
some of
American League allowing a "des-
As we
write,
owners once again are
is
114-16, 169.
the diverse
inequality. In the end,
inequality
moving
in the pitcher's favor.
* Rader, Baseball, pp. 87, 89,
will look at
mound from
strike zone,
we
ways
in
which public investment also
will better understand the
historically so inconstant
so high.
130
major reasons
and why inequality
in
America
HOW UNEQUAL
1
Reducing Poverty through Redistribution
Visible Policy:
Over
American government has done much
the last century,
to help those
poor by the market. Public health programs, school lunches, food
left
stamps, Aid to Families with Dependent Children benefits
have reduced the inequality
Americans have chosen not
to pursue
left
(AFDC), and
survivors'
by earnings differences. Yet
such programs as far as citizens
in
other affluent nations have (and the programs are being sharply cut back as
we
Most
write).
ilies
industrial societies provide "family allowances" to all
fam-
with children and some form of universal health care or health insur-
ance to
residents. In such ways, the
all
numbers and problems of the very
poor are sharply reduced by government policies that are directed toward
everyone and that do not single out the poor. Most American welfare pro-
—
who
can
prove that they are poor and that they are otherwise deserving. These
tar-
grams, in contrast, are "means-tested"
available only to those
geted programs consequently lack wide political support and are vulnerable to budget-cutting.
Only
social security
and Medicare, nearly universal
entitlements for the elderly, have largely survived cutbacks in recent years.
Most other
nations, unlike the United States, also substantially subsidize
housing for
many moderate-income
who make
it
into higher education,
citizens, provide stipends for students
and support the long-term unemployed.
Recent American antipoverty programs have had some success, but mostly in reducing poverty rity
and Medicare, and
among
the elderly, largely through social secu-
in taking the
edge off misery (see figure
can see the emphasis on the elderly by looking cans
who
are pulled
above the poverty
line
at the
by
all
government
programs (taxation, unemployment support, welfare, social put together. In 1992, 22 percent of Americans
below the poverty
line if all that
families' earnings.
Government taxes and
cent, a
drop of ten points
elderly, taxes
5.5).
We
percentage of Amerifinancial
security, etc.)
would have had incomes
had been available
them were
to
their
transfers reduced that to 12 per-
in the proportion
of poor Americans. For the
and transfers reduced the proportion by forty points, from the
50 percent who would have been poor based on nongovernmental income alone to the 10 percent
come and was
to
taxes.
who were poor
after including
governmental
For children, however, the net effect of taxes and transfers
reduce poverty rates by only seven points, from 24 percent to
percent. For
in-
young
adults, the
drop was merely 131
five points,
1
from 21 percent
CHAPTER to 16 percent.
1
6
This generational imbalance
is, in part, the outcome of polweakened the equalizing effects of taxes we have no data on the effects of the 1993 Clin-
icy changes during the 1980s that
and transfers. 2 (As of
yet,
ton tax changes that raised the earned income tax credit for low-income families and raised the
income tax
rates for the very wealthiest households.
Presumably, these laws shifted net incomes toward equality a
changes enacted by the Republican Congress elected
incomes away from If
we
income trition etc.
But the shift
equality.)
the programs that helped nonelderly
list all
little.
1994 will
in
— food stamps, AFDC, Women,
Americans with low
Infants and Children
(WIC
—
a nu-
program), Medicaid, SSI
—they sound
like a lot.
disability, the earned income tax credit, Adding together these programs and adding in
as well a variety of federal, state, and local spending directed not just at the
poor but also
many people who
at
are
above the poverty
line,
such as col-
lege loans, job retraining, and energy assistance, the total expenditures for
"persons with limited income" in 1992 amounted to almost $290 billion.
As
sizable as that figure
ment expenditures
is, it
represents less than 12 percent of
all
govern-
comes to about $5,900 per lowincome person. Almost half of this total, $134 billion, was for medical care, largely Medicaid. Nonmedical spending came to about $3,200 per limitedincome person, of which about $2,100 was in the form of cash or food stamps. That $2,100 is roughly what the typical American family spent on eating out in 1992; it is within a few hundred dollars of what typical homeowners saved on their federal income taxes by being able to deduct mortgage interest. Even after this government spending which is probably a high-end estimate of what America spent to aid low-income people in 1992 over 14 percent of Americans, 21 percent of American children, at all levels that year. It
—
—
remained poor. 3
We
can best evaluate the effort to redress poverty comparatively. Low-
income American children
are
worse off than low-income children
in
any other industrial nation. In the 1980s, for example, about 20 percent of American children lived in poverty, while 9 percent of Canadian children and of Australian children were poor, 7 percent of children in the
United Kingdom, and even fewer respectively.
Why
in France,
West Germany, and Sweden,
4
are so
many American
children poor? Charles
Murray claimed
in
an earlier book that American children are poor because welfare policies
encourage poor
women to have more children. He is wrong.
by demographers demonstrate minimal bearing. Rather,
young parents
are
effects, if any,
more 132
of
Careful studies
AFDC
on
child-
susceptible to poverty, and their
HOW UNEQUAL? poverty makes their children poor. 5 American children are more often poor, first,
because American adults are more unequal
in
both wealth and income
than people in any other industrial society. Second, children suffer especially
because the incomes of young
mid-1970s. More young of poverty, and
many
men
men have
fallen so sharply since the
cannot earn enough to keep their children out
then refuse to take on the responsibilities of marriage,
young mothers and children even poorer. We can see how American government compares with others in dealing with poverty by turning again to the Luxembourg Income Study. Lee Rainwater and Timothy Smeeding calculated, for eighteen nations, the percentleaving
age of children
who were
poor. (To be able to
compare across countries,
"poor" was defined as being in a household with real purchasing power less than half that of the median in the nation. Half the median the poverty line in the United States calculated.)
6
was
in the
roughly what
is
1960s when
Figure 6.1 shows the percentage of children
it
was
first
who were poor
before and then after including taxes and government transfer payments in the calculations. Again,
we
look only
at the
populous nations. Before gov-
ernment intervention, a relatively high percentage of American children
were poor, but not as high as
in
France and the United Kingdom. After
counting taxes and government payments, however, the poverty rate for
American children was
substantially higher than that elsewhere (including
shown
Even those countries with higher before-government child poverty than the United States managed to reduce their poverty levels to far below the level here. Two objections might be raised to the evidence that America leaves so nine other nations not
many of its
children in poverty.
poor because so
many
in the figure).
One
is
that so
many American
live in single-parent families.
Rainwater and Smeeding also looked separately and
in single-parent families. In
figure 6.
1
:
action.
7
The
other objection
being below 50 percent of the median,
how much
real
is
is in
being poor elsewhere. Unfortunately, that ing calculated
American children near
more
real
However,
children in two-parent
likely to
that being
be
left
poor
is
in
after
America,
not so. Rainwater and
purchasing power children
the top and
in
poor
material terms not as terrible as
at the
Smeed-
at the 10th, 50th,
and 90th percentiles of the income distribution had available try.
children are
is true.
each case, the same pattern appears as
American children were exceedingly
government
at
That
in
each coun-
middle did, indeed, have
income than did children near the top and
at the
middle else-
where. But American children near the bottom had less real income than children in the other nations, 25 percent less than poor Canadian children
and 40 percent less than poor West German children. 133
And
again, the
CHAPTER
6
Government in the United States does the least to reduce the proportion of children who are poor: percent of children poor before and after government action.
United States
Australia
Canada
'//////A
United Kingdom
IT Italy
V/////A
West Germany
France
in the 1980s.
They
participate
similar technologies, and have similar types
of industries and occupations. Yet these other countries experienced neither the
same
wage
large increases in
inequality nor the drops in the real
earnings of the less skilled. In Canada, Japan, and
grew, but
much more modestly
and
inequality changed hardly at
Italy,
than
it
Sweden wage
inequality
did in the United States. In France all.
52
(Only
in
Great Britain did the
gap between professionals and blue-collar workers increase as much as did in the United States, and there only because of gains for people with
low wages rose
in
at the top; the
it
pay
Great Britain, just not as quickly.) Low-
wage workers in the United States now earn only about halt as much relaAmerican high-wage workers as low-wage workers in Europe earn
tive to
relative to
Why
high-wage ones
there.
did other advanced industrial countries experience less
wage
in-
equality in the 1980s than the United States did? Because, in large part,
they
made
different policy choices. In particular, other
advanced countries
have different rules about unionization and have different wage-setting institutions.
53
Union Rules Economists Richard Freeman of Harvard and David Card of Princeton mate
that the sharp decline in the percentage of unionized
1980s explains
male workers
at least one-fifth
in the
of the growth
United States.
54
This
is
in
wage
inequality.
differentials
In the
were simply too few union members
Between 1970 and 1990
esti-
in the
among
because unions reduce the pay
gap between higher- and lower-ranking workers. 55 the 1980s, there
workers
United States
to offset
in
growing
the proportion of the labor force that
was unionized dropped more than 45 percent
to only
1 1
percent of the
private sector, virtually the lowest unionization rate in the industrialized
world.
Unionization, in turn, has declined so precipitously in the United States largely because of the unusually hostile political and legal environment here. Especially instructive
is
the
comparison with Canada, because Can-
ada and the United States share similar cultures, economic institutions, and standards of living. In the 1940s Canadians revised their labor laws to re-
semble the United States' 1935 Wagner Act, which established
legal proce-
dures for labor organization and collective bargaining here. Since then.
however, Canadian labor laws have become more favorable while American labor laws have
become 149
less favorable.
56
to
unions
Under current
CHAPTER Canadian law, a union
is
indicating their support.
6
established once a majority of workers sign a card
Under current American
workers have signed such cards, unions must tion
still
go through a long
elec-
campaign, often facing management consultants hired by employers
convince workers that they do not want a union in
law, after a majority of
Canada
ble to
for
do so
employers
in the
many Americans March
1995.
after
all.
57
Also,
to replace strikers permanently, but
it is
it is
permissi-
United States. This was brought to the consciousness of during the baseball strike that ran from August 1994, to
The Canadian government forbade
foreign replacement
workers, and Ontario provincial law prohibited hiring any replacements all.
to
illegal
The Toronto Blue Jays were forced
to schedule their possible
at
1995
"replacement baseball" season in Florida.
when
In the 1950s and 1960s,
labor laws and practices were most similar
two countries, unionization
in the
rates
were also
similar, but since then
unionization rates have risen in Canada and dropped sharply in the United States.
By
the early 1990s the percentage unionized in the private sector in
Canada was almost
three times larger than in the United States. 58 Partly
because of these differences
much
Canada than
less in
in unionization rates,
in the
wage
inequality
grew
United States in the 1980s. 59
Unions and Plant Relocation In the 1980s
some major American companies busted
their unions in cele-
brated cases (after President Reagan had defeated the air traffic controllers' union).
More
often,
however, employers escaped union pressure by mov-
ing from one state to another or out of the United States altogether. scale of
movement during
the decade of the 1980s alone
is
The
staggering.
University of North Carolina sociologist John Kasarda estimated that the
northern and midwestern states lost 1.5 million manufacturing jobs and
$40
billion in
in southern lost to
pay between 1980 and 1990. 60 One-third of the jobs ended up
and western
states;
some of the
rest
moved overseas; some were
automation; and some were simply lost as firms stopped producing
goods.
The competition among
localities for
struggle to land sports franchises
among
cities
and
is
jobs-on-the-move
a vivid illustration.)
states for firms usually turns
commitments, and promises the victors are significant.
intense.
on tax concessions,
to regulate union activity.
They do not
is
61
(The
The competition capital
But the costs
to
necessarily increase their tax bases,
because the bidding frequently requires giving away tax revenues;
also,
local taxpayers often contribute to firms' relocation costs. National policy
150
HOW UNEQUAL' allows states to differ greatly in laws protecting labor and thus to compete
on the basis of who has the weakest ones, thus encouraging to
weakly unionized
— and lower-paying — Wage
employee with ployer. In
his or her
many jobs,
the
or no negotiation
tem, and one result to
Setting
United States, workers' wages are negotiated either by an individual
In the
little
the shift of jobs
states.
is
is
employer or by a
local union with a specific
employer simply
offers the job at a preset
involved. This
is
an extremely decentralized
that differences in the
be high. The variation
gender, and occupation
in
wages
much
is
for
s> s-
wages of similar workers tend
workers of the same age, education,
greater in the United States than
more centralized wage-setting systems.
countries with
em-
wage;
62
it
is
in
That means
greater earnings inequality here. In countries like
Norway and
made up of employers of
Austria, national employer associations,
in different industries, bargain
the national unions to determine
all
sector of the
wage
economy. Local employers and unions
increase (but not decrease)
wages above
with representatives
levels for workers in each
are then allowed to
the national level
if
they agree
to.
France and Germany, bargaining goes on between unions
In countries like
and employers' associations
in
each industry or region; the government
then routinely extends these collective agreements to nonunion workers
and firms
in the relevant industry or region.
These kinds of centralized
arrangements diminish the amount of wage inequality. They do so by ting a
wage
floor for those at the
for those at the top of the
Centralized
wage
wage
worker makes
American top
is
CEO
so
scale, particularly executives.
setting practices are
tween what a typical European
much
set-
bottom of the pay scale and a wage ceiling
CEO
one reason why the disparity be-
makes and what an average European
less than the difference
between what
a typical
reaps and what an average American worker earns. While
American managers might claim
that their
enormous compensation
packages are justified by their productivity, researchers have found only a 63 between executive compensation and productivity. And, as we mentioned in the last chapter, Western Europe's economic growth out-
weak
relation
paced ours between 1970 and 1990. 64
One way in the face
to think about the policy choices different countries
have made
of economic pressures during recent years
The Euro-
is this:
peans have generally chosen to keep workers' real wages up, even
means
a slightly higher level of
if that
unemployment (because employers 151
will
CHAPTER hire
fewer workers
at
6
those wages). Part of that decision
to sustain the basic living standards of the long-term
government transfers and fault
—
to allow real
services.
wages
The United
is a commitment unemployed through
States has decided
to drop, so that slightly
more people
—by de-
are
working
but in lower-paying, less-secure, and often benefit-shorn jobs. Since no
made to assist the workers in these poorer jobs, income has widened more in the United States.
provision has been inequality
Both unionization rules and wage-setting practices are the
result of pol-
And these policy choices have profound effects on the amount of inequality we see in American society today. Recent statistics show that icy choices.
before 1974 American workers' increases in productivity were rewarded
by increased wages. Since 1974, in
this
has no longer been true. Productivity
both manufacturing and services increased by over 50 percent since then,
but wages in both sectors have been essentially
flat.
The
last
twenty years'
gain in productivity instead fueled gains in executive compensation and in stock prices. American workers received no greater slice of the growing pie
because they had no place
at the table.
65
Public Investments: The Case of Higher Education Public investment decisions also shape inequality.
Some
investments, like
clean water or public parks, improve everyone's quality of
down
income
the
others.
Roads
that
ladder.
life
up and
Other investments benefit some of us more than
go from suburbs
to
downtown business
areas of our large
example, tend to advantage middle- and upper-class commuters
cities, for
more than they do
central-city residents.
investments that affects
all
One of the most
of us, but in different ways,
important public is
public higher
education. In a crucial but not too visible manner,
a choice that
moved
Americans a generation ago made
the United States toward greater equality.
From
the
1950s to the 1970s, America invested enormously in higher education. In
1945 there were enough five
slots in
Americans aged eighteen
grown
to
about four for every
sive because
years.
it
postsecondary education for only one of
to
twenty-two.
five.
66
is
the
to serve
number had
especially impres-
happened while baby-boomers were entering
Higher education expanded enough
tion of a
By 1992
The expansion
their college
an ever greater propor-
growing population of young people. 67
Expansion was achieved through a generous commitment of public 152
re-
HOW UNEQL
\l
'
sources. Indeed, private college and university enrollments
grew only
slightly faster than the eligible population, while enrollments in public col-
New
leges and universities soared. States like California and
York
built
elaborate systems of higher education: junior colleges, state colleges, and university
campuses
New
University of size
—
the University of
Michigan from 20,000
15,000 in 1955 to 62,000 state
and local
levels,
tunity, the aspirations
sion,
in
Ohio
to 45,000;
1975. These political choices,
from
State
made
expressed Americans' optimism and belief
largely in
at
oppor-
of states and cities for prestige and economic expan-
and parents' desires to assure
Those who believed sion of opportunity
and campus after campus of the State
in California,
York. Other public universities increased greatly in
in the link
were
right.
their children's futures.
between higher education and the expan-
For those fortunate enough to earn one. a
four-year-college degree levels out family advantages and disadvantages in a
way
that increases equal access to
good
jobs.
no connection between the occupational
there
is
their
own. Children of the working
Among
college graduates,
status of their parents
jobs as are children of the middle class once they have a diploma.
when higher education expanded from 1960 father's place
on the economic ladder determined what
place would be
to the rise in the
was
cut
by
68
So
to 1980, the intergenerational
inheritance of socioeconomic status dropped dramatically.
ter's
and
class are as likely to land prestigious
half, nearly all
of
How much
his son's or
a
daugh-
this decline attributable
69 proportion of the workforce with college degrees. (The
weakening of the connection between parents' and children's rectly contradicts Herrnstein
based intelligence
is
and Murray's argument
becoming more important
in the
statuses di-
that a genetically
modern economy.
If
they were right, the correlation between parents' and children's statuses
should have grown stronger during those years. There are signs, however, that,
with increasing tuitions and stagnating investments
tion, the pattern
of expanding opportunity
is
in
higher educa-
beginning to reverse.
" I
Expansion of higher education increased equality of opportunity b\
weakening the connection between parental and child
status.
But overall
income depends on whether expansion of higher education keeps pace with the economy's demand for educated workers. The great
equality of
development of colleges
in the
1950s and 1960s increased the supply of
educated workers, reducing each graduate's claim on high wages. The
number of managerial and professional jobs
available
from 2.2
fell
71
college diploma-holder in 1952 to 1.6 in the mid-1970s."
workers could
still
ladder, but overall
bump
income
153
each
Better-educated
less-educated workers from jobs farther equality increased.
for
down
the
CHAPTER
Whom
6
Will Colleges Reject?
The expansion of public higher education from 1952 modest expansion of private colleges equality of opportunity in America.
in the
1969 (and a more
to
1970s and 1980s) increased
Young people who lacked wealthy
or
well-educated parents were increasingly able to compete on equal terms
with students from more advantaged backgrounds as seats in colleges creased. However, the picture will be sharply different in the
cade.
Between 1991 and 2001
the
number of 18-to-23-year-olds
increased by 40 percent, as the generation of the "baby
reaches college age. But in the 1990s there fiscal
Thus have
is
in-
coming dewill
have
boom echo"
neither the public will nor the
resources to finance another round of expansion of higher education. the proportion of
young people going on
to higher education will
to fall.*
One of
us,
Michael Hout, has examined
handled similar enrollment squeezes. The
how
Irish
modern
other
societies
between 1970 and 1984
allocated university positions solely on the basis of achievement test scores.
With the
eligible population
growing substantially
faster than uni-
versity places, this policy disadvantaged high-scoring poor
dents equally.
Italy, in
and rich
stu-
a similar squeeze, kept admitting students without
adding faculty or resources, so that students admitted to university increasingly found that they could not get into lectures.
It
took students longer to
graduate, and those from better-off families were likelier to finish. Russia
faced an admissions bottleneck
when
schools without expanding university
it
expanded academic secondary
facilities.
The shortage of places
exacerbated historic inequalities, so that students from poorer back-
grounds became even
The United
likelier than before to
be rejected.
States does not provide the kind of free university education
Ireland offers, so
it is
unlikely that the privileged and the disadvantaged
will bear the brunt of the enrollment
States will probably
mix
crunch equally. Instead, the United
the Italian and Russian patterns. Students
who
qualify for relatively low-cost public institutions will have a harder and
harder time simply finding the classes that allow them to earn degrees. The
more privileged on
to those ity
will
until graduation.
who have
be better able
And access
to
manipulate the system and to hang
to private institutions will increasingly
greater ability to pay.
of opportunity. *
Hout, "The Politics of Mobility."
154
The
result will
go
be reduced equal-
HOW UNEQUAL? After the mid-1970s, however, the supply of educated workers that col-
more slowly than the demand for them. Thus, as we first pointed out in chapter 5, the wages of college graduates rose relative to those of nongraduates. And inequality of income between those who had leges provided rose
who had
and those
who do
not graduated college increased again. 72 Today, those
not graduate from college
high school education or less
—
lege graduates are rising at a time
ratio
1.65.
The earnings of
the earnings of high school gradu-
Between 1979 and 1989
in high-tech
like increased the
the
manufacturing, health services, legal services,
demand
demand
Meanwhile the commerce reduced
for college graduates.
decline of traditional manufacturing, bookkeeping, and for workers with a high school education.
kinds of jobs available in the United States all
a
col-
did not go to college (the "B.A. premium") rose from 1.45 to
Growth
and the
the
who have
so, those
of earnings for college graduates to earnings for high school gradu-
who
ates
when
did not attend college are falling. 73
who
ates
—and even more
face bleak prospects.
These
economy do
shifts in the
not account for
of the increase in earnings inequality in the 1980s, but they do account
for the increased
B.A. premium. 74
It is
we emphasize, market demand
that reversed
a trend,
an earlier one and that reflects not just the
for workers,
but also the supply provided by our decisions about investing in higher
education.
American policy regarding postsecondary education is disMost of our trading partners provide students who do not go to college with more vocational training than we do. Successful systems link In addition,
tinctive.
schools and firms. Firms can explain their labor needs to schools, and schools can draw on firms for technology and job placement. States has given
little
75
The United
systematic attention to vocational education,
al-
though recent research shows that vocational programs tailored to the labor
market notably increase workers' earnings. 76 Overall, then, America's investment decisions about education have had
important
—
if
complicated
education after World
youngsters
who were
War
—
effects
II
on
inequality.
The expansion of higher it gave more
reduced inequality, both because
less affluent the opportunity to attain
high-paying
jobs and because the growth of the supply of educated workers tended to
reduce the B.A. premium. Retreats since those days have increased equality.
At the same time, the
failure of the
generously in vocational training (or
in
in-
United States to invest as
primary and secondary education;
see the next chapter) has increased inequality here relative to other ad-
vanced countries where public investment been greater. 155
in these
kinds of education has
CHAPTER
6
Conclusion: The "Free" Market and Social Policy Influential
commentator George
Will, responding to headlines about
ing inequality in America, voiced what ity is
not bad
A society
if
it
results
from a
many Americans
and
free
fair
market.
that values individualism, enterprise
neither surprised nor scandalized
when
grow-
believe: Inequal-
and a market economy
is
the unequal distribution of market-
able skills produces large disparities in the distribution of wealth. This
does not mean that social justice must be defined as whatever distribution of wealth the market produces. But
does mean that there
it
is
a presump-
tion in favor of respecting the market's version of distributive justice. Certainly there
is
today no prima facie case against the moral acceptability of
increasingly large disparities of wealth. 77
However, "the market's version of a natural market but
Some
from complex
policies determine
distributive justice" results not
political choices,
how unequal
many
the starting points are of those
enter the market's competition; other policies determine selects winners
and
losers.
how
For example, African Americans
were prevented by private discrimination and
from purchasing homes and thereby
from
of them hidden.
lost out
explicit
who
the market
in the
1950s
government policy
on subsidized loans and mort-
gage deductions. They were also unable to leave substantial assets
to their
As another example, think of the businesses in industries subsidies. The market is not a neutral game that distributes
children.
that
receive
just
rewards to the worthy; in biases.
it is
a politically constructed institution with built-
78
As we have shown rising equality of
government
here and in chapter
post-World War
policies,
some
II
5, the
enormous prosperity and
America resulted
legacies of the
New Deal,
in part
from many
policies that provided
old-age security, encouraged homeownership, gave labor increased bargaining power, built massive physical infrastructure, and financed an enor-
mous expansion of public
education. Since the late 1970s, however, public
investment of these sorts has slowed and sometimes actually reversed. At the
same time,
inequality has dramatically increased.
The kinds of inequalities we ral
see reemerging in
America
are neither natu-
nor inevitable, nor do they reflect the distribution of individual
Through our
politics,
Americans have chosen
talents.
to increase equality of op-
portunity (expanding higher education, for example) or equality of result
156
HOW UNEQUAL' (subsidies for to
do so
chosen. the
homeownership, Medicaid, and Medicare,
to a far
We
1
more
for instance), bui
limited extent than citizens in other nations have
extend support to fewer of our citizens, largely the elderly and
middle class; and
we
extend less support. For example,
we
provide
medical insurance for some residents; most nations provide medical care for
all.
We
provide a tax deduction for children of taxpayers; most nations
provide family allowances. Americans have also
made choices
that in-
creased inequality, such as the tax changes of the 1980s and the rules on unionization
we have
the well-off
more than
We have
many programs
to help
the less well-off, such as the subsidies for
home-
accepted.
structured
ownership and medical insurance.
What in great
all this
implies
measure a
is
that the inequality
result of policy decisions
chosen not to make. Generally, life
do
America
to
We
do
is
—
or
far less to equalize
have chosen to reduce
between the middle class and the upper class somewhat, but
far less to
cans — with
see today in
Americans have made
we have chosen
conditions than have other Western people.
the inequality to
we
reduce the gap between the lower class and other Ameri-
the notable exception of older people.
decades, our choices have these choices; others, like
moved
And
in the last
us farther from equality.
George
Will,
may applaud
Americans constructed the inequality we have.
157
couple of
Some
criticize
them. Either way,
*
CHAPTER
7
*
Enriching Intelligence:
More
Policy Choices
i F we have done our job well, readers of
this
book should by now appre-
ciate that the explanation for inequality lies in the design of society, not in
the
minds or genes of
individuals.
The rewards
that contestants gain in the
race for success are determined by the rules of the race, not the personal traits
of the racers. Even
who wins
or loses the race
is
determined more by
the "nutrition" and "training" they receive than by their "natural speed." In
the last
few chapters, we have directed
social systems
and shown
attention
that inequality
and
its
away from
individuals to
reduction are to be found
of competition and distribution. The leverage for expanding
in structures
opportunity and moderating inequality
lies in policies that
deal with those
structures.
In this chapter, tive skills
—
we
or, for
argue that social policy can also improve the cogni-
shorthand, the intelligence
—of
individuals. This
is
a
worthwhile goal, because a more cognitively skilled population would be a
more productive one,
tunity,
and cognitive
a wider distribution of skills
skills are
valuable in their
own
would expand opporright.
Those who be-
lieve that intelligence is genetically determined or fixed in infancy claim that nothing
That
is
can be done to increase such
skills; intelligence is
immutable.
why, they argue, intervention programs are a waste of taxpayer
money. However, the pessimists are wrong. Evidence shows gence can be changed. Indeed,
it
is
that intelli-
shaped every day by institutions
all
we define intelligence in its broadest sense, as it should be. But even if we define it narrowly, as school skills (see chapter 2), it is still changeable. To improve skills, we need to improve the social around
us.
This
is
true if
environment.
The debate about
the mutability of intelligence has been misfocused;
it
has lost the forest for the trees. Both the pessimists and the optimists have
argued about whether special interventions, such as Head Start and remedial instruction,
episodic events;
can raise it is,
test scores.
But the
real leverage lies not in
instead, in the continual, systematic, everyday
society forms intelligence.
We
such
ways
will therefore forgo debating those inter-
vention programs, only noting for the record that
158
many
scholars have per-
ENRICHING INTELLIGENCE suasively defended their effectiveness (see, e.g., Dickens, et
we
Instead,
will look at four
cognitive skills: schools,
chap.
al.,
3).
examples of how society pervasively shapes
summer
vacations, classroom tracking, and the
structure of jobs.
Do Schools Matter? In
1964 Congress mandated a major study of educational inequality
America, asking whether disparities ferences in students' test scores.
would arouse
findings
in
in
school resources accounted for dif-
Some members
of Congress hoped the
the nation to battle ignorance by equalizing educa-
tional resources across schools. Alas, the best-laid plans often fail. After
painstaking analysis and reanalysis, the late University of Chicago sociologist
James
S.
Coleman was unable
to find that school resources
had much
of an effect on student achievement. Neither the number of books library,
in the
nor the number of credentials on the teachers' resumes, nor the size
of students' classes, nor the length of the school day seemed to affect
achievement scores. Coleman concluded that family environments, rather than schools, determine learning.
The
report set off a firestorm.
parents
move
to
How
could schools be irrelevant, when
expensive suburbs just to enroll their children
in
highly
regarded schools? Did the Coleman Report really lay the blame for students' failures at the feet of the parents?
policy
if
finally,
what good
is
public
resources in schools do not matter?
Sociologists today understand
much
And,
how
schools shape inequality in learning
some commentators ignore the insights gained in the nearly thirty years since the original Coleman report. For example, Herrnstein and Murray cite the 1966 Coleman Report as better than they did in the 1960s. Yet
demonstrating the
futility
of attempting to raise intelligence. But they
ig-
many studies critical of the original 1966 report, including Coleown subsequent work. The primary value of the 1966 Coleman today is as a historical demonstration of how a mis-asked question
nore the
man's report
1
can lead to a mistaken answer. Correcting limitations of the original study and applying better methods, researchers have found that variations
key limitation of the 1966 report
among
that
is
it
schools are important.
assumed
that
every student
One in a
given school has access to the same resources, the same books, and the
same
teachers.
in the library
But we cannot make
this
assumption. Having
many books
can increase the achievement only of those students exposed 159
CHAPTER to the
7
books; quality teachers help only those students
Coleman's approach did not take
classes.
who
enroll in their
into account individual students'
exposure to the resources of the school when looking
at their
achievement
scores, and thus he underestimated the effects of those resources.
Researchers have devised better understandings of
achievement
how
educational
produced inside schools. They have also developed better
is
understandings of
how
children learn, which has led
them
to ask sharper
research questions, such as what affects students' rates of learning. researchers
have expanded
their
that schools
And
beyond simply comparing
how much
students learn during school
are out of school.
These kinds of studies show
schools, looking, for example, at
compared with when they
analysis
and school resources do matter.
Understanding
How People Learn
One important post-Coleman advance was
a deeper understanding of
people learn. Consider learning to ride a bike.
When
one
first
how
begins, one
knows nothing. Learning at the beginning is likely to proceed very slowly. As one begins to feel comfortable with the task, the rate of learning speeds up. Learning does not, however, increase forever; eventually plateau.
There
is
skilled, but after
always more
to learn,
some plateau
is
it
reaches a
and some bikers become extremely
reached additional learning usually pro-
ceeds slowly. Figure 7.1 demonstrates the classic "learning curve."
The slope of
the curve indicates the speed of learning: the steeper the
slope, the faster the learner learns. Different people
curves. is
have different learning
When we say that someone is "bright," we often mean that he or she we can measure people's skills frewe can measure each one's learning curve. We can then curves to see what circumstances accelerate or slow down
a fast learner in just this sense. If
quently enough,
compare those
individual learning.
Speed of knowledge growth is the very issue in which if one is interested in intelligence. For example,
one should be interested
Herrnstein and Murray claim that intelligence makes
some people more
accurate, rapid, and efficient in solving problems and attaining success. If so, then the
speed with which a person's knowledge grows could be de-
fined as their intelligence or, at the very least, as a
good proxy
for their
intelligence. Researchers studying learning curves find powerful evidence
primary determinants of speed of learning. This turns the old Coleman findings upside down. Anthony Bryk and Stephan Raudenbush, of the University of Chicago that schools are the
and Michigan State University, respectively, studied the learning curves of 160
E N R
I
C H
I
NG
INTELLIGhNCI-
Time An S-shaped
7.1.
618 students grade.
86 different schools as they moved from
in
They investigated
reading. This distinction explicitly taught birth
Learning Curve
math
first
is
important because the only place students are
is in
school, while language skills are taught from
and through most social interactions children have.
searchers found that variation the variation in students' skills,
grade to third
students' learning curves in mathematics and in
among
When
earlier re-
schools did not account for
math scores but mattered more
much
of
for language
they concluded that schools could not be important because
if
they
were, schools should matter especially for math. 2
Bryk and Raudenbush found what Coleman found when they asked
same question Coleman asked: reading achievement that
is
due
How much
the
variation in mathematics and
to the schools the children attend?
They found
only 14 percent of the variation in mathematics achievement and 31
percent of the variation in reading achievement
was
attributable to differ-
ences between schools. Unlike Coleman, however, Bryk and Raudenbush
had data that allowed them to investigate learning curves. They came to a startling conclusion.
Over 82 percent of the
ing curves occurred between schools; that
how
variation in mathematics learn-
is,
82 percent of the variation
in
quickly students learned math was attributable to school differences.
Nearly 44 percent of the variation curred between schools. matter: schools are
Thus, average
finding
more important
at its core, the
test
The
in learning
what one would expect
is
for
math than
if
scores
backgrounds, then ship, resources
it
—do
seem
to
A and
in
if
schools
for language skills.
answer of the Coleman report was
scores for students in school
the same, and
curves for reading also oc-
school
B
incorrect. If the
are pretty
much
be accounted for by the students' famih
appears that schools not matter. Yet. 161
—
their funding, structure, leader-
more complete analyses show
that
CHAPTER
7
schools matter very much. Schools matter because differences within schools in
how
(We
students are taught affect learning.
And
instance, tracking, below.)
will look at
one
schools matter because they influence rates
of learning.
may seem
That schools matter
many American
an unexceptional conclusion; after
But the conventional academic wisdom for years was
was not academic. Cognitive
those struggles
only by family background cians, social
—
background for
wisdom
larly for the less
skills
that the payoff for
seemed
genetic background for
to
be affected
many psychometri-
sociologists. (Perhaps the payoff lay in the
social or "networking" advantages of political
all,
parents struggle to put their children into "good" schools.
in recent years
good
schools.)
The conventional
has been that investing in schools, particu-
advantaged, was wasteful. The Bell Curve seems to certify
these views, but here again
it
misleads
Schools appear to be the primary
us.
determinant of children's intellectual advancement.
Summer Vacations Other researchers also interested
in the question of
whether and how
schools affect learning and test scores have taken advantage of a routine
American life, the summer vacation. It is the major reason most American children receive only 180 days of instruction per year. Researchers compared how much test scores change during the school year with feature of
how much they change during in school has
The affect in
on cognitive
logic of
summer in
order to see what effect being
learning research
is this:
We
can see
how
schools
development by comparing how much children learn when they are
school with
that,
summer
the
skills.
how much
they learn
when
they are out of school. To do
researchers compare changes in test scores that occur during the
school year, from
fall to
spring, to changes in test scores that occur during
summer vacation, from spring surprise except to those
to fall.
who doubt
that intelligence is
dren increase their intellectual performance school than during a month of
pens to children's
skills
—and mutable —
Researchers have found
much more
it
is
no
that chil-
during a month in
summer vacation. Furthermore, what hapsummer depends a lot on their summer
during the
environments. Children from disadvantaged circumstances tend to stagnate or even
fall
back
circumstances
intellectually during the
make some
summer, while those
gains during the vacation.
It
in better-off
looks as
if
class parents provide their children with experiences, such as travel or lessons, that
add
to cognitive
162
middle-
camp
or
growth and higher scores, while
ENRICHING INTELLIGENCE summer
poorer parents do not. Just like going to school, having
activities
increases test scores. In studies separated
by some
fifteen years, at different stages
leged destruction of standards in schools, Barbara Heyns, then versity of California, Berkeley,
of the
al-
Uni-
at the
and Doris Entwistle and Karl Alexander
at
The Johns Hopkins University found that all students learn at a faster rate during the school year than they do during the summer. Some students' rates of learning are so
peers. Poverty
is
slow during the summer
the important determinant of
that they fall
behind their
which students tend
to fall
behind during the summer. The researchers also found that the disadvantages of poverty are compensated for during the school year; that
children begin to catch up to middle-class students while school sion. In
summer
reviewing the literature on
entire black-white
gap
is
poor ses-
Heyns found
learning,
reading achievement
in
is,
is in
due
that the
to differences that
emerge during summer vacations. These findings imply
that schools are
very effective because they overcome the disadvantaged backgrounds of students. skills
But schools cannot sustain
their effectiveness in
improving the
of the disadvantaged because schools are in session only nine months
of the year, whereas family disadvantages operate during winter, spring,
summer, and
fall.
Many weaker come
3
students might gain academically
if
schools were to be-
year-round. Herrnstein and Murray concede in passing that year-
round schooling might reduce inequality but then dismiss the point by guing that
it
is
likely die hard.
politically inviable.
However,
it is
4
The
tradition of
summer
a matter of policy; there
ative for school buildings to close during June, July,
is
ar-
vacation will
no natural imper-
and August, nor for
children to take a three-month vacation from formal schooling.
(It is
a left-
when children were needed to help on the farm.) Resummer vacation tradition is also to choose lower levels of
over from the days specting the
intellectual skills ities in
and a higher level of inequality.
cognitive skills
if
we chose
We could
reduce dispar-
to.
Tracking In the to
wake of the Coleman
report,
many
researchers turned their attention
what might be called "schools within schools,"
process of assigning
some
students to
to tracking. This
is
the
more rigorous college preparatory
courses and other students to less rigorous preparation for immediate work careers. Essentially,
it
creates different learning environments within one
school.
163
CHAPTER
7
Higher-scoring students end up in college tracks, but at each level of test score, students have a better chance the higher their class background.
607c
10
20
15
25
30
35
Math Test Score Probability that Students
7.2.
Were
in
College Track by Math Test Score
and Social Class (Source: Adapted from Gamoran and Mare, "Secondary School Tracking")
Researchers have focused on two questions: (1) Are track assignments
Do
fair? (2)
mean
that
track assignments matter?
equally
achievement
By
fairness, researchers typically
well-prepared students
—of whatever
—
as
measured by previous
ethnicity, gender, or class
have an equal chance
of being assigned to the college-preparatory track. The vast majority of analyses
show
that parents' social class strongly influences
5 to the college track. This
ity,
is
who
is
assigned
one way parental background sustains inequal-
by shaping children's academic development.
Figure 7.2 shows the estimated probability of entering the college track for white
boys of varying mathematics achievement. The horizontal axis
shows tenth-grade students' scores on a math screening cal axis
shows
the chances that a student
was
test,
and the
in a college track.
verti-
The heavy
middle line indicates the probability that a student from a family of average 164
ENRICHING INTELLIGENCE social class
would be
above and below off
that
in a college track,
homes and from worse-off homes,
track.
6
At each
given his math score. The lines
one represent the chances
that students
respectively,
would be
from
better-
in a college
from better-off families had
level of test score, students
a
greater chance of being placed in a college track than did students of poorer class backgrounds.
At the highest
level of
measured mathematics achieve-
ment, economically disadvantaged students suffered a fifteen-point penalty
chances of being
in the
lowest lines
achieved
in the college track.
at the far right
at the
(Compare
the highest and
of the graph.) The lower-class student
who
highest possible level had about a 42 percent chance of
being in a college track, while the higher-class student with the identical score had about a 57 percent chance. Put another way, students
one-quarter of the mathematics questions but
who
miss
come from advantaged back-
much chance of being in the college track as the impoverwho answered every question correctly. (These comparisons
grounds had as ished student
are fairly pure because the possibly
confounding effects of the schools'
structure and composition, as well the effects of individual achievement in
other subjects, have been statistically held constant.)
The
class bias in track placement, large as
would not matter
it is,
if
being
tracked did not affect achievement and college entry. But students in college tracks not only learn
more
and learn different
facts
develop a different relationship to knowledge (see box,
Many
studies
show
hoff of land,
p. 166).
7
One example
will
more than do show how tracking
performance. In an especially well-executed study, Alan Kerck-
Duke University used
data on
Northern Ireland, and Wales
compare students placed in
they also
that students in the college track learn
students in the noncollege track. alters test
facts,
in
all
in the
children born in England, Scot-
week of March
3-9, 1958, to
high tracks, students in low tracks, and students
untracked schools. This design allowed Kerckhoff to control for
the factors
we
already
know might be
many
of
related to achievement, such as stu-
dents' social class, aspirations, and prior educational experience. Kerck-
hoff had information on the students at age seven, age eleven, and age sixteen.
He found
that high-track students
gained more academically than
students in the untracked schools, and that low-track students gained less
than students in the untracked schools. This pattern widened preexisting differences. Thus, after students
were assigned
to different tracks, their
achievement levels diverged; tracking exacerbated inequality.
An
obvious rejoinder to the Kerckhoff study
to high tracks
were more
is
that the students assigned
intelligent at the outset.
high-track students to learn so
Thus, one would expect
much more over time
that the
gap be-
tween them and low-track students would grow. But Kerckhoff compared 165
CHAPTER
7
Same Book, Different Lesson Researchers have shown that in
many
school practices undermine instruction
lower tracks. For example, high-track students are often given demand-
ing material, reading Shakespeare, Faulkner, and Walker, while their peers in
low-track classes read "young adult fiction." But even
students are given the
same books,
when low-track
the uses to which the material
result in vastly different learning experiences.
A
put
is
high-track class might
write and produce a play that captures the essence of The Color Purple, or,
using Julius Caesar as a point of departure, argue over whether and
one can justify murder. Students
how
lower tracks are more likely to
in the
fill
out worksheets that have a rote structure of question and response, and
where
all
of the "correct" responses are to be found in the book. The high-
track students are learning to treat
can even produce for one's
knowledge
are taught to see their
*
own
knowledge
as
something one uses and
ends; in contrast, the low-track students
as something outside of themselves,
beyond
power, that stands over and against them.*
Cookson and
Persell,
Preparing for Power, Anyon, "Social Class and School
Knowledge"; Bernstein, Class, Codes, and Control.
who had
students
same
the
signed to different tracks.
initial
achievement level and yet had been
He found that
as-
students assigned to the high track
gained more than otherwise similar students assigned to lower tracks. Thus,
if,
as
many
claim, these tests reflect intelligence, Kerckhoff's study
shows how tracking
selectively nurtures or neglects existing intelligence.
Those whose intelligence nored
is
nurtured gain; those whose intelligence
is ig-
lose.
Kerckhoff studied Great Britain; the same kind of finding appears
in the
United States. 8 The data show that tracking has powerful effects on cognitive
growth and
less than they
offset
that students
excluded from demanding instruction learn
would have otherwise. Because gains
by losses for others, the average
dents in tracked schools
Tracking exists school.
in
Even though
is
level of
for
some
students are
achievement of
not increased. Yet inequality
is
all
the stu-
increased.
elementary and junior high as well as secondary the systems of tracking are far
more varied
lower schools, analysts find the same pattern for younger students
in the
we have
discussed in the case of older students: powerful effects of social class on student placement, and powerful effects of track location on students' suc-
166
ENRICHING INTELLIGENCE 9 cess that exacerbate preexisting cognitive differences. There
broader
common
in other
in the
Western nations
United States
— means
sum, studies of tracking suggest
demic intelligence to
do just
ing;
it
that.
seems
the time.
all
The
that
and even
10
we do change
children's aca-
entire process of tracking
Most Americans may have good reasons
to
— much more than
that entire schools
school districts are, in essence, on different tracks. In
also a
is
which American schools are tracked. The great decentral-
and local control of schools
ization is
sense in
designed
is
for adopting track-
advantage middle- and upper-class students. Less well-off
students, however, are shunted disproportionately to lowei tracks. Inequality in learning,
although not the average level of learning,
through tracking.
By
is
increased
these practices schools demonstrate the pliability of
cognitive skills as well as the powerful effect social factors have on the
success of individuals. Policies alter intelligence.
Adult Development and Jobs So
far
we have looked
at
how
schools affect children's intellectual
That makes sense, because schools are the major institution signed precisely to expand people's cognitive
velopment does not stop with school nor end
abilities.
But
abilities.
we have
de-
intellectual de-
in people's childhoods.
Older
notions of intelligence led psychologists to assume that, after about age twenty, people stopped growing mentally and probably got dumber. Intelli-
gence
is,
by
this view, fixed early in life.
cognitive skills deteriorated after
cussed in chapters 2 and
what people learn of
that.
in
3, their
is
that they
now
abilities,
do tend
to
forties, stay level for ten or
Even
intelli-
twenty more
then, however, not all
Moreover, gerontologists have found that even the
weaken can be
much
including practical problem-solving.
years, and then tend to decline after sixty.
that
dis-
their notions of intelli-
believe that cognitive skills and practical
gence typically increase into the
skills deteriorate.
we
measures of intelligence are measures of
But modern psychologists have expanded
result
as
school and, of course, as people age they forget
gence to include a variety of
The
Psychometricians believed that
young adulthood because,
skills
restored with training. Just five hours of
training can, for example, substantially
improve older people's inductive
reasoning. 11
Not everyone grows does everyone decline
intellectually as they in old age.
What
some
stifle
167
age, nor
some environments stimulate Getting more education, staying
ple face and the experiences that they have;
cognitive development and
move toward middle
matters are the situations that peo-
it.
CHAPTER
even being married for a long time
culturally active,
spouse
—
One
slow their deterioration. 12
crucial experience that affects adults' mental their jobs. Social psychologists
is
working out of the National
Institute
of Mental Health, have conducted
how
it is,
studied
And
The more complex the job, the less rousupervision it entails, the more self-directed the job
and the
is
less
the consequences of self-direction
what they
Kohn and Schooler have
call "intellectual flexibility"
—
in effect, intelligence.
they have found that "job conditions that promote occupational self-
direction increase
.
.
.
whereas jobs
intellectual flexibility,
pational self-direction decrease
A quick and obvious
.
.
.
that is
13
what explains the correlation between
having a demanding job and having good cognitive
much
that limit occu-
intellectual flexibility."
objection might be that intelligent people seek out
and get self-directing jobs;
skills.
That
is true,
and
of the causal connection between job and psychology runs that way.
But, using various techniques and data sets,
show
able to
tioning,
Kohn and Schooler have been
that the causal effects run both
that is intellectually
whatever
Upon reflection so.
different kinds of
self-direction jobs require.
Among
is.
in these
Kohn and Schooler have measured how
jobs affect people psychologically.
tine
development
Melvin Kohn and Carmi Schooler,
decades-long research across three continents on
much
to an intelligent
these are the kinds of activities that spur the growth of cognitive
abilities or at least
ways
7
demanding
ways. Having a complex job
itself increases
people's cognitive func-
their prior intelligence.
—and upon other research—
Spending eight hours a day,
five
it
is
clear
days a week,
fifty
why
that
weeks
would be
a year in a
position that forces one to analyze problems, to calculate, to strategize,
often to persuade others, and to do these tasks independently exercises the
mind, just as a rigorous calisthenics regime exercises the body. Spending those the
same two thousand hours
same routine actions and
a year in a job that only asks one to repeat
to follow orders is to the
hours each day as a "couch potato"
is to
mind what spending
the body.
The psychological consequences of job
structure,
turns out,
it
go beyond
the jobholders themselves to their children. Social scientists have long
known
that middle-class parents tend to teach their children
independence
and self-direction, while working-class parents are more likely size conformity
and being obedient. Melvin Kohn showed
for this difference
is
that middle-class adults usually
that
to
empha-
one reason
have had more sec-
ondary and higher education, a training that stresses independent thinking.
Kohn
also
experience.
showed
"Men
that another reason for the class difference is the
of higher class position,
168
who have
job
the opportunity to be
ENRICHING INTELLIGENCE self-directed in their work,
come
want
to think self-direction possible.
Men
come
to regard
matter of necessity to conform to authority, both on and off the job.
man
—
it
who do
of lower class position,
not have the opportunity for self-direction in work,
does mold the
and
to be self-directed off the job, too,
it
a
The job
can either enlarge his horizons or narrow them." 14
These different orientations then carry over into how adults parent
when and how they discipline their children, workers are more likely to expect their children
for example. Blue-collar to
do what the parent says
because the parent says so; white-collar workers are more likely to expect their children to
be able to explain why the parent says what he or she says.
This training, for obedience or for independence, differently for their
own work
The conformist values and ents, with their
in turn
prepares children
lives.
orientation of lower- and working-class par-
emphases on externals and consequences, often
are inap-
propriate for training children to deal with the problems of middle-class
and professional
life.
ations, solving
new problems
.
.
.
[C]onformity
—
in
is
new
inadequate for meeting
short,
[T]he self-directed orientation of the middle and upper classes ...
new and
adapted to meeting the
situ-
for dealing with change.
the problematic.
At
its
best,
it
is
.
.
.
well
teaches
children to develop their analytic and empathic abilities. These are the essentials for handling responsibility, initiating
Without such
reacting to
it.
The job
one setting
is
skills,
that,
change rather than merely
horizons are severely restricted. 15
well into adulthood, shapes intelligence and
other psychological dispositions that, in turn, affect people's
While most jobs would seem policies,
to
outcomes.
life
be outside the purview of governmental
most are under the control of large private organizations. Evi-
dence shows
mands of
that,
within limits, employers
the jobs they provide also
who
enrich the intellectual de-
improve the cognitive
skills
of their
workers.
The broader point
is
that cognitive skills
keep changing over the
life
course and are changed by experience. Policy can intervene here by, for
example,
increasing
older
people's
opportunities
for
intellectual
stimulation.
Conclusion: The Only Constant
Is
Change
Despite the rhetoric of the intelligence-never-changes school, not be surprised to learn that cognitive skills
169
do change over the
we should life
course.
CHAPTER known are weak
7
change so much
Indeed, researchers have long
that those skills
early intelligence scores
predictors of later intelligence test
scores.
seven
The
correlation between
is .86,
and age
measured intelligence
at
that
age six and age
but the correlation between measured intelligence at age six
only
fifteen is
.69.
Were
the correlation to be close to
1
,
intelligence immutable,
we would
or at least
we would
not to decline as the gap in ages increases. But children's ranks on
change so much
that their scores at
expect
expect the correlation
IQ
tests
age six account for less than half of the
variation in those scores at age fifteen. 16
Why
should such change in IQ occur? Because, as
we have
noted, chil-
dren experience different school environments, and those environments play an important role in
how
fast they
leam. Children, locked out of
school in the summer, cope variously well or poorly with the absence of
formal cognitive training. Children are assigned to different tracks, and
some of those to
tracks lead to heightened self-efficacy and greater exposure
knowledge and
analysis, while other tracks lead to lowered self-efficacy
and reduced exposure. But children are not alone and challenging environment faster
when
which
in
cognitive abilities.)
A common
grow; even adults lose their edge
which they work and
the environments in
capabilities. (Other
to
in requiring a nurturing
theme here
is
that challenges
demanding jobs, grow
to
at birth or at least fixed
narrow, academic intelligence
—
When their skills
when
not, those skills
who
is
rently shape skills,
in
like decisions
like decisions
intelligence.
To
is
de-
—even
about tracking
about vacations and job
the extent that academic intelli-
gence, as measured by grades, SATs, and the
advancement
ability
quite changeable.
American policy choices, whether obvious
for
claim that intelligence
by adolescence, cognitive
and school funding or more subtle
American
demanding schools or
meet those challenges.
tend to atrophy. Contrary to the fatalists
structure, shape
test their
and mental exercise expand
are tested, people usually strengthen those skills;
termined
do not
17
intellectual abilities. People, be they children in
adults in
live
such as television-watching, also shape
activities, too,
like, is
used to sort out people
school and the economy, then these same policies cur-
American
we can reshape
inequality.
Because we already shape cognitive
inequality.
170
CHAPTER
*
8
*
Race, Ethnicity, and Intelligence
N
lo
chapter
in
The Bell Curve received more attention than the one
claiming that African Americans and Latino Americans are inherently less intelligent than white
Americans look ing,
sex,
Americans. This attention
and even sports
Murray raised
every issue
at virtually
—through
this sensitive topic
not surprising, because
is
—crime,
poverty, politics, hous-
the prism of race. Herrnstein and
even though
their essential
argument,
that individual differences in intelligence explain social inequality, nei-
ther stands nor falls
on the question of group differences
Although the issue of race
is
irrelevant to their
main argument
there to justify a critique of affirmative action), their
American obsession with
race.
We, however, needed
ferences in this book, because race has been and
We
our society.
still is
cannot understand inequality
in
in intelligence.
book added
(it
is
to the
to treat racial dif-
the great
chasm
in
America without ad-
dressing the roots of racial inequality and the controversy over
intelli-
gence.
who may have turned first race, we reiterate the basic
For readers it
deals with
The Bell Curve
is
ray use, and
most
tion people
have had, not
much about
social
test
findings of our earlier chapters:
of intelligence Herrnstein and Mur-
measure how much academic instruc-
others, too, really
measure differences very
The
mistaken.
their inherent abilities.
Even
if
in native intelligence, those differences
among
inequality
because
to this chapter precisely
such
tests
did
do not explain
individuals in America; individuals'
environments explain more. More significantly, the distribution of
individual intelligence has
little
to
society; patterns of inequality are
structures of the nation and era.
do with the extent of inequality
in a
produced by the economic and social
Both the conditions
that help or
impede
individuals' race to succeed and the system of inequality within
which
those individuals compete are heavily governed by specific social policies.
Thus, policy choices have shaped the kind of class inequality Policy choices, over the long course of
kind of racial inequality
we
American
have, as well.
171
history,
we
have.
have shaped the
CHAPTER
8
The Argument about Race and Intelligence:
A Preview African Americans and Latino Americans in the United States tend to be
economically worse off and to suffer from more social problems than do whites. Blacks and Latinos
standardized
tests.
1
For those
also tend to score lower than
do whites on
who
"natural," the
second pattern clearly explains the
believe that inequality first:
Ethnic groups are socially unequal
because they are intellectually unequal. But for those societies construct the inequalities they have, as has this
book, the reverse
is
true:
is
who
understand that
been demonstrated
Groups score unequally on
tests
in
because they
are unequal in society.
Consider
this situation:
Members
of a minority,
brought to the country as slave labor, are
They do the siders them
dirty
deadly
bottom of the social
work, when they have work. The
rest
ladder.
of the society con-
violent and stupid and discriminates against them.
years, tension in
at the
many of whom were
Over
the
between minority and majority has occasionally broken out
riots. In the past,
minority children were compelled to go to
segregated schools and did poorly academically. Even now, minority chil-
dren drop out of school relatively early and often get into trouble with the law. Schools with
many
minority children are seen as problem-ridden, so
majority parents sometimes
move
out of the school district or send their
children to private schools. And, as might be expected, the minority chil-
dren do worse on standardized
tests
than majority children do.
What
is this
who
in the
minority?
Koreans
in Japan.
Koreans,
who
are of the
same
"racial" stock as Japanese and
United States do about as well academically as Americans of Japanese origin (that
is,
above average), are distinctively "dumb"
in Japan.
planation cannot be racial, nor even cultural in any simple way. nation
is
that Koreans,
half-century,
whose nation was a colony of Japan
have formed a lower-caste group
in Japan.
The
ex-
The explafor about a
2
Consider another case: Immigrants flood into the United States to take
They are "swarthier" and more "primitive" than most Americans; they seem unwilling or unable to assimilate; they are suspected low-wage
jobs.
of criminal behavior; and they threaten native workers' jobs. Together with other newcomers, these immigrants provoke a backlash against easy immigration into the country. Objective test data
172
show
that the
immigrants are
RACE. ETHNICITY. AND INTELLIGENCE low on intelligence; and
their children
do poorly
in school.
Who
does
this
describe? 3 Polish Jews in the 1900s and 1910s.
These same Polish Jews, whose descendants now do well both economically
and on
were scorned when they came. Many scholars of the day
tests,
believed that the Jews, along with Italians and other "dark" types, were dim-witted.
4
Can
race or genes explain what happened in the intervening
As
seventy or so years? No, but social location can.
a peripheral and sub-
servient group in the early twentieth century, Polish
even
As
in school.
they
came
to
be accepted
in
Jews were
American
failures,
society, their
position and their "intelligence" rose.
Our argument, which we or ethnic group's position rather than vice versa.
will flesh out later in the chapter, is that in society
Some
determines
its
measured
ethnic groups find themselves in inferior posi-
Maori
tions through conquest or capture (e.g., the Irish in Great Britain,
New Zealand,
most
origin
Jews
in Israel,
Turks
closer to equality faster.
mance
in three
and lasting effects of subordina-
grow up
5
(e.g., Italians in the
in
Germany). These groups typically move
In either case, subordination leads to
members of
in suffer
like
a familiar one and the subject of that
reduce
much
test
that this first process
—
is
Low
income,
ill
health,
performance. This process
research.
When
Herrnstein and
socioeconomic differences between individual blacks and
whites do not explain the black-white gap in
ditions
low perfor-
the subordinate groups and the families
socioeconomic deprivation.
poor parental education, and the
Murray say
United States, Eastern-
ways.
First, individual
is
drastic
Other minority groups enter a society through immigration, often
made under economic duress
they
in
Africans, Mexicans, and Indians in the United States). These
are the groups that suffer the tion.
a racial
intelligence
—
test scores,
they are asserting
deprivation based on differences in economic con-
insufficient to account for the race differences. This
process they examine.
The second
is
the only
process, however, cannot be understood
individualistically.
Subordinate groups typically experience segregation. Ethnic ghettoes concentrate disadvantages and accentuate them. (The term "ghetto" originally described Jewish quarters in
grants to
European
cities. Ironically, the
America from those ghettoes were viewed
immi-
in a similar light
gen-
erations ago as today's ghetto residents are viewed.) Segregation also
exposes children
who would
otherwise do well to the problems and the
culture of the disadvantaged.
173
CHAPTER
How inferior ethnic caste position leads to low test scores.
Socioeconomic Deprivation
4
^^
Low Ethnic Status
^\
U Gr ° U
Caste or
r
Position^
> cSegregation ?
I
Low
>
Test
Scofes
v
Stigma of Inferiority
8.1.
The
A Model
of
How Low
Ethnic Position Causes
third process is cultural
Low
Test Scores
and psychological: Young members of sub-
ordinate groups understand that they carry a stigma of inferiority based on
The young people respond in variSome become anxious, fatalistic, and resigned;
the wider society's perception of them.
ous ways to that
identity.
others reject the wider culture's expectations and standards, adopting an
oppositional stance. Either of these reactions to stigmatizing images resignation
or rebellion
—brings
down
average school and
test
per-
formance. Youth from advantaged ethnic groups have mirror-image experiences of
all these.
They
typically benefit
from more
circumstances, from having friends and neighbors
affluent family
who do
well and
who
can help them to do well, and from the confidence that they are destined to succeed. Figure 8.1 outlines the argument.
We
do not claim
groups
that this explanation applies equally well to all ethnic
in all societies;
histories matter.
it is
important to understand that groups' particular
But we do claim
that
it
applies to most cases and that
it
applies particularly to groups that have been severely suppressed, such as
American Indians, Mexican Americans, and African Americans. Understood
this
way, groups'
test
scores are not the beginning of an
explanation for inequality but the end of one. The beginning
That
is
system
why
is history.
the next section reviews the history that created a racial caste
in the
United States. In the subsequent section,
174
we
describe the
RACE, ETHNICITY.
AND INTELLIGENCE
continuing legacy of this caste system in our time. In the explain
why
academic
skills.
We show that these
lower scores
American
fit
a
worldwide pattern of
do with
ethnic group differences that has nothing to Instead, the
we
final section,
blacks and Latinos tend to score below whites on tests of
racial intelligence.
case, like others around the globe,
is
explained by
the three consequences of caste: deprivation, segregation, and stigma. norities score
Mi-
lower because they are lower caste.
Ethnic History and Ethnic Inequality many
Herrnstein and Murray, like
seem
others,
in
America
to believe that the history
of slavery, segregation, and discrimination in America, a history they ac-
knowledge but then
forget,
ended sometime
in the 1960s.
6
The
slate
was
Any group
inequalities in
outcomes since the 1960s, therefore, can only be explained
as the result of
wiped clean; everyone then group inequalities
started out even.
in natural talent.
crimination had ended in the 1960s tion)
—
the weight of history
is
But
not so simple. Even
life is
—and
it
if dis-
certainly did not (see next sec-
oppressively heavy. Three decades of
mod-
erate reform
have narrowed some of the gaps, but Americans were naive
think that
would quickly erase
it
A caste system
ranks groups economically, politically, and socially. The
by law and
privileges of those ranks are usually enforced
belong to the ranked groups by virtue of their sense ethnic or racial.
The system
is
justified
by a
policy.
groups are
birth; the set
People in that
of beliefs, often
shared by the lower-ranked as well as higher-ranked groups, about the tellectual is
to
three centuries of a rigid caste system.
in-
7
and moral superiority of the higher-ranked one. The term "caste"
drawn from
India, but
it
describes perfectly the American race system
until the 1960s.
Although most non-English people who came ination and exploitation, the experiences of a
What
to
America faced discrim-
few were
distinctly different.
separates the historical experience of Africans from that of other
groups
is
that they did not
come
to the
United States to find work but were
instead impressed into labor. They, like
Berkeley anthropologist John
Ogbu
American
Indians,
and are distinct from voluntary immigrant groups such as the and Chinese. 8 The African case
is
form what
has labeled an "involuntary minority" Irish,
Jews,
well-known: Europeans purchased cap-
and sold them in America as slaves. The Mexican American case is more complex; it
tive Africans
175
fits
neither
Ogbu
cate-
CHAPTER gory neatly. (We focus
Mexican experience distinct
from
that of
in particular
on Mexican Americans because the
United States
in the
8
is
closer to that of blacks;
most other Latino groups, such
it
is
Americans
as Central
and Cubans; and Mexicans are the largest Latino population. The case of Puerto Ricans, one-sixth the population of Mexican Americans, however, bears
many
Mexican Americans and African Ameri-
similarities to that of
cans.) After the
Mexican- American War of 1848, when the United States
conquered Texas, California, and the Southwest, many Mexicans were caught under the control of the enemy they had just fought. The Americans treated the
Mexicans both
as beaten foes
and as exploitable
labor.
9
Later
immigrants to the United States, although voluntary, were absorbed into a
conquered group. Moreover, those immigrants were treated by Anglos, as
we
members of a racial
shall see later, as
different in origin but similar in
caste.
outcome
Thus, the Mexican case was
to that of involuntary migrants,
such as Africans.
Although educated Americans ought not history of slavery, a
few
salient points are
slavery was, historically, one of the
more
to
need reminding of our cruel
worth noting.
First,
American
rigid versions of slavery in the
world. Africans were not only taken thousands of miles from their homelands but were also shorn of
much
of their culture. Americans built an
especially high wall between the status of slave and that of freeman and built
it
higher over time. In most states any drop of "black blood" con-
demned one and one's descendants forever to slavery, unless formally emancipated. The race-and-slavery borderline was ferociously guarded with respect to sexuality.
was
A
second distinct feature of American slavery
the elaborate ideology southerners developed to reconcile slavery with
a democratic society.
They began with
religious justifications, that blacks
were descendants of Noah's dark son Ham, whose transgressions con-
demned
his descendants forever to
be servants. They ended with "scientific
racism," scholarly arguments that Africans' biological "nature" fit
made them
only to be servants. Third, American slavery ended later than did slavery
elsewhere in the advanced world.
These distinctive features of American slavery contributed
mous weight of disadvantage passed on
formidable edifice of slavery was, of course, not natural;
Americans decided over
fifty
in
more years
1789
to allow slave importation until
after that
to the enor-
to the descendants of slaves. it
was
The
policy.
1808 and for
decided to permit slavery to expand.
We
continue to live with the consequences of those choices.
Mexicans were not as distinct
slaves, but in the
Southwest they formed a caste
from Anglos as blacks were from whites. Anglo Americans 176
AND INTELLIGENCE
RACE. ETHNICITY.
Intelligence as Rationalization
When
groups find themselves
intelligence
used
this
is
a perfect
of inequality, the dominant one
in situations
generally develops a justification for
its
advantages.
A
theory of innate
example of such a justifying ideology. Americans
theory to explain
why
Africans and Mexicans
such as the Irish and American Indians*
—were
—and
others, too,
subordinate. Although
both Africans and Mexicans had established complex societies
Anglo Americans believed them
native lands,
Thomas
Jefferson provides an example of such an interpretation:
Comparing nation, in
in their
to be intellectually inferior.
[blacks]
by
their faculties of
appears to me, that
it
much
reason
inferior, as
memory
in I
memory, reason, and imagithey are equal to the whites;
think one could scarcely be found ca-
pable of tracing and comprehending the investigations of Euclid;
and
imagination they are
that in
dull, tasteless,
Most of them indeed have been confined homes, and
own
their
society: yet
and anomalous.
to tillage, to their
many have been
to the handicraft arts,
.
.
own
so situated, they
might have availed themselves of the conversation of
many have been brought up
.
their masters;
and from
that cir-
cumstance have always been associated with the whites. Some have been
liberally educated,
and
have lived
all
in countries
where the
arts
and sciences are cultivated to a considerable degree, and have had before their eyes samples of the best works from abroad. ... But
never yet could
I
find that a black
level of plain narration; never see
had uttered a thought above the
even an elementary
trait
of painting
or sculpture.**
Jefferson struggled with the contradiction between his egalitarian principles
and
slavery
his slave practice.
was
Although Jefferson thought the
institution of
tyrannical, he understood certain practical considerations.
Slavery provided the labor needed to sustain the plantation economy.
Abolishing
it
would
create
potential political chaos.
economic hardship
A theory
for the general citizenry and
of innate intellectual inferiority helped
reduce the dissonance. Logic could be sacrificed for
*
in this effort
by implying,
example, that by working near his master a slave might absorb a
On
the Irish, see Roediger,
The Wages of Whiteness.
American Indians, see Rogin. Fathers and Children,
p.
On
bit
of
such arguments regarding
34.
** Jordan. White over Black. Jordan draws his quotation from Jefferson's Notes on Virginia.
.
177
CHAPTER
8
Euclid. Jefferson also dealt with the dissonance by suggesting that the
question of innate inferiority be
left to
future scientific observation, a solu-
tion remarkably similar to that disingenuously
advanced by Herrnstein and
Murray.*** ***
Ibid., pp.
438-39; Herrnstein and Murray, The Bell Curve,
innumerable references
in
p.
317. There are
The Bell Curve concerning the need for future
scientific re-
search to determine the real impact of genetics on the cognitive abilities of different races. Thus, like Jefferson, they
hedge on the
ability to
genetic differences in cognitive ability and leave
it
account scientifically for real
for future scientific inquiry.
considered Mexicans both ethnically different and former enemies but also
saw
them
in
peon system
a source of badly needed labor.
Anglos adopted the patron-
to use that labor. In this system, the
on a landowning patron
to provide
him with
peon ranchworker
a job,
home, security
relied
for his
family and his old age, and even religious instruction of his children (through godparenthood). For these favors, the peon reciprocated with total loyalty throughout his lifetime.
10
This quasi-feudal structure, which oper-
ated not unlike slave plantations in the South, provided a social order useful for
economic development. Anglos did not need
during the early years because the
to segregate
patron-peon system was
Mexicans
sufficient to
maintain caste lines and guarantee Anglo control. Anglos did not impose the
most malignant forms of
adopted the same sort of Jim
racial ideology until the
Crow system
1920s when they
that African
Americans then
11
faced.
Nineteenth-century claims about differences in intelligence provided a rationalization that
was comforting
to whites but a
poor explanation for the
12 disadvantaged positions of African Americans and Mexican Americans.
Physical intimidation, legal codes, and social custom are sufficient to ac-
count for the
latter's
lower caste position. Apologists invoked intelligence
to justify the caste order only after whites
blacks and Mexicans.
The
had physically subordinated
13
slave and feudal systems that had been in place gave
way
at the
end
of the nineteenth and beginning of the twentieth centuries to a rigid and
malevolent caste system based on tenant farming and segregation. Understanding this system
cause
it
is critical
persisted until recently.
for understanding race relations today, be-
Many
white Americans dismiss the oppres-
sion of blacks by saying that slavery ended over a century ago. True, but
178
RACE, ETHNICITY,
AND INTELLIGENCE
serfdom lasted much longer. This serfdom was only yesterday for
virtual
blacks and Mexican Americans.
White southerners established the Jim Crow system over the quartercentury following Reconstruction's end in 1877,
when
federal authority
withdrew from the former Confederacy. Jim Crow laws and vigilante pression such as
Ku Klux Klan
embarrassment nor euphemism
lynchings were designed
—
to sustain white
— with
supremacy
re-
neither
after the
Emancipation Proclamation. Those laws established segregation, eliminated civil liberties, withdrew the right to vote, denied educational and
employment tions.
lords
opportunities, and severely constrained intergroup social rela-
The Jim Crow economic system tied black farmers to white landin debt peonage that was only one step above slavery. For Mexicans
in the 1920s, the
in cattle
new economic system
replaced the patron-peon relations
ranching with those more appropriate to tenant farming and agri-
business.
14
These
rebuilt caste systems
ended brief eras of
relatively
open
opportunity for blacks during Reconstruction and Mexican Americans before the 1920s, ushering in generations of denied rights and repression.
The
history of black education in the South
Immediately
after
the
Civil
schools in the South for freed slave children. the opportunity, so
slaves
much
compared well
is
particularly instructive.
War, northern Reconstructionists opened
The
children quickly grabbed
among the former among white children. With the
so that school attendance
to school attendance
advent of Jim Crow, most of that progress was reversed. Sociologist Stanley Lieberson has
shown
that, starting
about 1880, spending on black
plummeted from rough parity with spending on white schools to about one-third of the amount spent on white children. School terms for black students became shorter; teacher schools in the South, where most blacks lived,
per student ratios in black schools lower; and black teachers' standards
weakened. Outside of a few major
cities,
high school education was, for
all
practical purposes, unavailable to black youth in the early years of the
twentieth century. In 1911, for example, Atlanta had no high school for
black students; in 1930 about one-third of counties in the South had no four-year high schools for blacks.
The dismantling of black education was
especially devastating in those regions with the greatest concentrations of blacks.
went
15
Black teenagers today, for the most
part,
have grandparents
who
to those blatantly inferior southern schools.
Starting at the end of Reconstruction, 1877, for African
Americans and
including Mexican Americans by the 1920s, segregation, sometimes legal
and sometimes informal, became the primary means through which access to schooling, housing, and jobs
was
179
controlled.
16
(One
justification
CHAPTER authorities offered for segregation
was
8
that
mixing of the races would com-
promise the superior culture and apartheid
17
intellect of white Europeans.) American began with schooling. "Separate but [supposedly] Equal" was
many
the law in
states.
No
matter where they lived, African American and
Mexican American children had
to attend separate schools, schools invari-
ably inferior to whites' schools. This profoundly undermined the education as well as the self-esteem of those
surprising that black and in school
who were
forced to attend.
It
not
is
Mexican American children lagged behind whites
achievement. (Note, however, that in the early part of the twenti-
most white immi-
eth century northern blacks did better in school than did
grant groups.) 18 Authorities then used lower academic achievement, a result
of educational segregation, to further justify segregation.
Barred by segregation from good education and from well-paying jobs, as well, blacks and
Mexican Americans had
to live in ghetto housing.
consequences of multiple segregation were profound.
It
The
reduced living
standards and housing quality and reinforced the segregation of the schools.
It
concentrated disadvantage into tight quarters. Not the
segregation told
all
least,
black and Mexican American children that they were
not as good as whites, and this
left
would have enormously negative
a psychological scar of inferiority that
effects for decades to
come. 19
Racial stratification remained governed by legal segregation until after
World War
II
when
the Civil Rights
movement
steadily challenged the
system's constitutional basis. This challenge culminated in the successful case of Brown
v.
The Board of Education ofTopeka, Kansas. The Supreme
Court reasoned in
its
decision that "separate but equal" schooling
herently unequal, that
when
was
in-
a society legally separates people by race,
it
sends a message that they are not equal.
down of Jim Crow, and the passing of subsequent civil rights legislation, many Americans thought that both blacks and Mexican Americans would now have the opportunities they had After the
Brown
ruling, the tearing
been denied for so long, lished. Progress
that a level playing field
was made.
New
had
finally
been estab-
policies such as equal opportunity en-
forcement made a difference. In 1940 employed blacks earned 43 percent as
much
as did whites;
by 1980
that proportion
had risen
economic progress leveled off after the mid-1970s.
20
In
to
73 percent. But
some
areas, includ-
ing education and occupational advancement, moderate advances have reversed. This
new
trend troubles Herrnstein and Murray.
They
point to the
growing number of blacks (and others) who are very poor, have been that way for a considerable time, and have little hope of improvement the
—
so-called underclass.
As we noted
in earlier chapters, Herrnstein
180
and Mur-
RACE, ETHNICITY. ray
AND INTELLIGENCE
combine an economic explanation,
that low-skill jobs are disappearing,
with an explanation based on intelligence, that minorities are insufficiently intelligent to ties in
make
it
in the
new marketplace,
economic fortunes. (Of course,
to explain the
group dispari-
few blacks and Latinos
relatively
were succeeding when "suitable" jobs were available.) This over.
It
is
an explanation that ignores history except to claim that history
assumes
and psychological oppression
amount
economic,
that three centuries of steady physical,
—
to nothing, disappear,
no one contests
a record that virtually
evaporate in thirty years.
and sociologically naive! Parents' advantages
—
is
social,
How
historically
property, learning, per-
sonal contacts, practical crafts, social skills, cultural tools, and so on
—
are
passed on to their children and are also passed from older members of a
community to younger ones. Disadvantages are passed on, too. Jim Crow was banished only one generation ago; its legacy will last much longer.
Ethnic Inequality Today To many,
it
might appear that the
1960s created a level playing
lems with that impression.
civil rights legislation
field for all,
but there are
of the 1950s and at least
two prob-
century or more, African American
First, for a
and Mexican American families faced severe discrimination
in education,
housing, jobs, and other economic opportunities. Such disadvantages cu-
mulate and burden future generations.
It
was naive
Mexican Americans could immediately compete
to think that blacks
and
effectively with whites in
the labor market. Second, the conditions creating deprivation never fully
changed (see box,
p. 182).
Although de jure segregation has been
officially
terminated, de facto segregation and discrimination clearly continue. Critics of
government action on
The Bell Curve, argue to
that discrimination against blacks
occur but does no longer.
groups
still
racial matters, including the authors
If
discrimination
lag behind economically, that
ability or will to succeed. But, despite
discrimination illegal, the refusal of a
it
persists.
News
Denny's restaurant
is
policies
and laws
the
make
such as
to serve black Secret Service officers,
Field "experiments" have demonstrated that there in,
that
stories of discrimination,
repeatedly point this out. But systematic research points
nation
and Latinos used
gone and these minority
must be because they lack
new
of
is
it
out, too.
extensive discrimi-
for instance, the housing market. In the typical study, black
and white researchers, posing as homeseekers with identical credentials,
approach
realtors, agents, lenders, or landlords.
181
At
least half
of the time
CHAPTER
8
Martin Luther King on the Conditions for Equality Martin Luther King,
Jr.,
warn both whites and blacks
tried to
that they
should not think that three hundred years of servitude could be overcome with the changing of laws alone.
were
ties
to
be given a
If
African Americans and other minori-
chance
real
compete, something other than
to
merely allowing them to compete would be needed.
Whenever for the
is
raised;
Negro should be granted
On
nothing more. realistic.
For
it is
wrote:
of compensatory or preferential treatment
this issue
Negro
He
some of our
friends recoil in horror.
The
equality, they agree; but he should ask for
the surface, this appears reasonable, but
obvious that
if
a
man
it
is
not
entering the starting line in
is
a race three hundred years after another man, the
first
perform some impossible
up with
feat in order to catch
would have
to
his fellow
runner.*
*
King,
Why We
Can't Wait,
p. 134.
discrimination occurs. For example, the black applicants are not
shown
properties the whites are; they are told that there are no apartments rent while the white applicants are
shown those same apartments; and
blacks are "steered" to black neighborhoods.
where so much depends on where we
live
21
—
the
left to
the
In a society such as ours,
the quality of our schools,
police protection, access to jobs, tax assessments, etc.
—being turned away
from some neighborhoods and being pushed toward others has profound rippling effects. 22
Other studies show similar discrimination except
when under
are the same.
In
when
qualifications for the job
one study, businessmen admitted as much. Sociologists
Joleen Kirschenman and Kathryn the
job hiring. Employers,
strong affirmative action pressure, prefer white to black
or Latino job applicants three to one, even 23
in
Neckerman interviewed employers
Chicago area and found many who said
in
that they preferred not to hire
black men. 24 Even managers of fast-food places in Harlem, according to another study, prefer to hire nonblacks. 25 In interviews with the
New
York
Times, foremen admitted to hiring whites over blacks for construction jobs.
One foreman explained, 'They Employment discrimination open competition,
sports.
[whites] are the people
I
know
best."
26
also appears in that realm of supposedly
The days of 182
explicit segregation are gone, but
RACE, ETHNICITY, black athletes sionally.
still
There
have
to
than white ones to play profes-
bit better
some evidence
also
is
be just a
AND INTELLIGENCE
because of fan preferences,
that,
black players' salaries are lower than they would otherwise be. 27
And
then there are the everyday harassments and slights that identifiable
minorities suffer in America: being ignored by storekeepers, watching
people cross the street to avoid you, subtle rejection, being questioned by
Survey research shows
police.
that white
Americans are
definitely less
28 prejudiced than they used to be. But blacks do not need to be constantly
victimized, as they were under Jim Crow, for discrimination to
The occasional encounter with discrimination and
lives.
of disparagement, even
show
that whites tend,
if
now
even
if
oblique,
is
would
still
their
a pervasive sense
enough. Experimental studies
unconsciously, to ignore blacks in need or to
convey negative impressions of them. 29 (Were only one bigoted, that
warp
in eight
whites
leave one hostile white for each black in America.
Numbers like these would leave whites with the reasonable impression that bigotry was rare and yet leave blacks with frequent experiences of bigotry one way that racial groups can experience the same reality so differ-
—
ently.)
While each discriminatory incident may be
trivial,
a lifetime of such incidents
mounts up
rare
and perhaps even
to a constant experience of
mistreatment. 30
We
might wish that history's heavy weight were the only burden
minorities carried today, but they
where
is this
more
we just described discourage blacks and
finding housing in white areas.
that minority renters
same housing Americans
—
Crow One
stock.
to face the
same
Latinos from
they have found housing in white flight.
Housing segregation means
situation that they faced during the era of
Jim
of the striking changes in American cities during the twentieth cen-
been the increasing residential segregation of blacks. The number "black residential isolation index."
that blacks tend to live
blacks tend to live only
every black lived solely northern
cities.
among
(There are three line segments
northern
to
cities.
that
among other blacks. A perfect 100 would mean that among other blacks. The data cover eighteen large, in the figure
on somewhat
tions of "neighborhood," but the overall trend
from 1890
Low numbers
High numbers mean
whites.
calculations in different eras had to be based
in
dis-
and home purchasers pay more than whites for the More critically, it forces many blacks and Mexican
in figure 8.2 is called the
that
The
segregation.
tury has
mean
When
has typically precipitated white
it
that
confront active discrimination. No-
blatantly visible than in the housing market.
criminatory practices
areas,
still
is
clear.)
because the
different defini-
The
figure
shows
1970 blacks became increasingly segregated from whites 31
Since 1970, black isolation has leveled off
183
at a
high
CHAPTER
Over the twentieth century, blacks became segregated from whites in northern cities.
1930
much more
1950
1970
1990
Year
8.2.
Index of Black Residential Isolation, Eighteen Northern Cities,
1890-1990 (Source: Calculated from Massey and Denton, American Apartheid, pp. 24, 28, and 64; and from Harrison and Weinberg, "Racial
and Ethnic Segregation")
rate. In the early years,
and lived
in
blacks were a small percentage of the population
pockets scattered around the
southern blacks
moved
in
and whites
cities.
As
fled to other
years passed,
more
neighborhoods while
preventing blacks from following them. The result was the division of northern cities into large, virtually all-black inner neighborhoods and virtually all-white outer
neighborhoods. During the same decades that white
immigrant groups became
less segregated
from native-born whites and
from one another, blacks became more segregated. 32 Through the 1980s, poor blacks
in particular
ety as middle-income in
became increasingly
Americans
isolated
fled the inner cities.
from the wider
America have become increasingly all-minority 33 The schools that black and Mexican American children attend 184
soci-
Poor neighborhoods are segre-
RACE. ETHNICITY. AND INTELLIGENCE gated again. This time
it
is
de facto segregation attributable to a segregated
housing market, but the result
the same. Indeed, de facto segregation
is
may
more detrimental psychologically, because African American and Mexican American children can no longer rationalize their separation well be even
as the result of legal oppression but see
instead as personal rejection.
it
Although middle-income blacks have been leaving the ghettoes of the poor, they usually end up living in or near rarely
low-income black ghettoes,
by choice. Their children are often drawn into the world of low-
income youth. 34 This experience sends a clear message youth that their
life
surety
it
will bring a
neighborhoods
Even
if
is
chances are going
What does
white counterparts'.
nonwhite
be significantly less than their
to
good education mean when
good job? What does
a
there
is little
good job mean when choice of
restricted?
blacks and Latinos went to schools that were equal in quality to
those that whites went
Unfortunately, that
a
to all
to,
it
would be
we cannot even
most of the minority students go
gated inner-city schools
is
difficult for teachers to inspire
to are not equal.
a difficult task.
Many
them.
because the schools
test that possibility,
Learning
in segre-
of the physical structures
are in poor condition, educational supplies are inadequate, teachers are
overwhelmed with disciplinary problems, and, more
generally, the climate
of pessimism weighs on the talented students as well. 35 mobility
filter
back from the world of work
couragement and resentment. ents, teachers,
It
Rumors of blocked
to the schools, fostering dis-
then becomes increasingly difficult for par-
and school administrators
keep minority students com-
to
mitted and focused on their studies. In this climate, high dropout rates
among Latino and African American high school
students persist.
Recall anthropologist John Ogbu's distinction between voluntary immigrants and involuntary minorities. 36 Voluntary immigrants, such as Italians,
Japanese, and West Indians, chose to
nities.
They
come
feel hopeful. If things
do not work
out, they
America
for
its
opportu-
homeland and
can go home, as millions of
immigrants have before them. Their optimism to
to
contrast their conditions here with those in the
in turn
commitment
sparks
schoolwork. Involuntary minorities, such as blacks and American Indi-
more complex way, were forced to be minorities in the white man's land. They contrast their conditions here with those of their fellow citizens and despair; certainly for American Indians ans,
and Mexican Americans
and blacks there
is
no
in a
realistic
homeland
to
compare with or
to return to.
(Former chairman of the Joint Chiefs of Staff Colin Powell put the distinction this
way:
"My
black ancestors
may have been dragged
chains, but they were not dragged to the United States. That ent emotional and psychological beginning than that of
185
to is
Jamaica
in
a far differ-
American blacks.
CHAPTER Delivering in the Barrio
and Dangerous Minds The popular movie Stand and Deliver
in
is
the Ghetto
instructive about student aspi-
rations in ghetto schools. In this true story, teacher Jaime Escalante takes
a group of Mexican
American students
Angeles and develops
their
in
one of the poorest schools
mathematical
score well on national exams.
An
skills to
in
Los
such a level that they
inspiring story, Stand
and Deliver
still
contains several dispiriting elements. For one, few teachers are as gifted
and dedicated as Jaime Escalante was. He put ings, evenings,
more important, even
a heart attack. Perhaps selves
worked so
hard, their success
regulators charged
portunity to have
how
in extra
hours
in the
morn-
and weekends with his students, overworking himself into after his students
was so unexpected
that
them with cheating. Mr. Escalante fought
them take
the test again.
had them-
examination
They scored high
for the op-
again. Yet
often can students' sincere efforts be sustained in the face of such
cynicism and suspicion? The stigma that young nonwhites face
is
a further
deterrent to striving in the classroom or on achievement tests.
Dangerous Minds
is
another truth-based, popular movie about
how
a
dedicated teacher successfully taught minority students. In the story,
LouAnne Johnson reached and helped youth in a crime-ridden black ghetto. Ironically, Ms. Johnson, in her own book, did not credit herself for the students' success.
She credited a federal grant
that paid for smaller
classes and for time to provide students with individual instruction.*
*
Mosle, "Dissed."
whose ancestors were brought here
in chains.")
untary minorities sparks resistance to
37
The pessimism of
invol-
schoolwork. These minority youth
might even be described as "rationally" pessimistic. Numerous studies
show
that the
economic advantages of staying
great for blacks and Latinos as for whites.
The oppressive weight of history and ination together easily explain
Americans remain behind
in the
why
in school are not nearly as
38
the pressure of continuing discrim-
African Americans and Mexican
American race
for success. Racial stratifi-
cation in the United States cannot be understood by looking at individual traits
such as intelligence; ethnic inequality did not emerge from individual
competition. Blacks were captured in Africa, Mexicans lost a war; neither
186
—
RACE, ETHNICITY. AND INTELLIGENT!-.
Smarts on the Street The debate about race and
intelligence has focused on intelligence as ex-
As we pointed
pressed in classroom tasks.
out in earlier chapters, these
formal vocabulary and mathematics, form but a small
skills, essentially
aspect of people's mental abilities. Paper-and-pencil tests capture poorly,
what
if at all,
it
means
to
be "smart"
in life outside the
Here
is
in
sense of their wider intelligence.
where anthropologists and sociologists have a great advantage
over psychometricians.
everyday
Many
of the former have studied people
in their
settings.
and anthropologists have studied the kind of
In particular, sociologists
people
full
—
Only by observing
business, in personal relations, in politics, and so on.
people in their real lives can one get a
classroom
whom
Herrnstein and Murray dismiss as constitutionally "dull"
black and brown residents of low-income neighborhoods,
many of them
high school dropouts, and some even criminals. Although researchers
have described bleak "dull" residents.
Latinos
who
have described
the ethnographers have found poor blacks and
are foolish and shortsighted, but they have also found poor
blacks and Latinos
More
lives in these settings, they rarely
Of course,
who
are wise and discerning, just as with any group.
striking in the research, however,
is
how
quick-wittedly and
shrewdly residents of these communities navigate through the perilous waters of ghetto
life.
Their "street smarts" often entail the same kinds of
sophisticated calculation required of professionals and executives.
members who operate
as black-market entrepreneurs,
"hustle" a living, single mothers
who
Gang
young men who
balance limited funds and demand-
—
men who juggle multiple low-paying jobs these are the kinds of people who perform poorly on the paper-and-pencil tests of the classroom but who nevertheless perform shrewdly on the survival ing children, working
tests
*
of the "mean streets."*
One
vast literature that speaks to these points consists of ethnographic studies
in
low-income communities, ranging from older ones
to
more recent ones such
as Stack, All
Our
like
ski.
done
specifically of
Islands in the Street; Padilla. The
Cocaine Kids; and Wacquant, "Life
Gang
in the
Street
Up
done
Corner Society.
Kin; Hannerz, Soulside; Anderson.
on the Corner; and Williams and Kornblum, Growing literature includes studies
Whyte.
A Place
Poor. Another applicable
gangs and delinquents, such as Jankow-
as an American Enterprise; Williams. The
Zone."
87
CHAPTER
8
group ended up on the bottom because they
on a
test.
The
caste system had
origins in colonization
its
Americans' need for cheap labor and rationalize their control. tify the existing racial
The
lost a fair race or
their
power
to get
role of intelligence has
order rather than to create
it,
scored poorly
—
to
in
European
keep
it,
been primarily
and
to
to jus-
it.
The conditions faced by blacks and Latinos have not changed as much in we might imagine. 39 Most of the young people
the last several decades as
groups face an educational environment and an occupational future
in these
And
that is simply not equal to that of whites. in the black
and Latino communities appear
conditions for the very poor
to
be getting worse. 40
Given the harsh past and the daunting present we have described, what is
perhaps more remarkable than the persisting gap between the academic is that the gap in test scores is narrowMurray admit grudgingly but must admit
performance of blacks and whites ing.
It is
a point Herrnstein and
nevertheless.
Over
the last twenty years or so, the white advantage over
blacks in various standardized tests has narrowed by the equivalent of several
IQ
points.
41
That alone should cast doubt on the idea that the group
differences are inherent and unchangeable. But
let
us look further.
Ethnicity, Race, and Test Scores The is,
current economic inferiority of ethnic minorities in the United States
we have shown,
sufficiently explained
by
their centuries-long suppres-
sion in a caste system. Purported differences in intelligence are not the
cause of their greater poverty, only a post hoc rationalization for theless,
we
return
now
to the question of
why
average, below whites on standardized school achievement is
what much of the controversy
gence"
tests,
it
is
more accurate
is
it.
Never-
blacks and Latinos score, on tests,
since that
about. (Although often labeled "intelli-
to say that the
exams
assess
how much
instruction students have received; see chapters 2 and 3.) Herrnstein and
Murray arise
pull
up
just short of claiming that the
from group differences
group differences
in genes, but others
in scores
have drawn the conclu-
sions The Bell Curve authors merely implied.
Many
critics
have attacked The Bell Curve both for
its
claim that there
are racial differences in intelligence and for the implication that those dif-
ferences are inherently racial. Critics usually engage the debate within the
psychometric tradition (see chapter box,
p.
2),
attacking the quality of the tests (see
189) or the quality of the statistical controls designed to simulate
an all-else-being-equal comparison of the groups. For the record,
we
will
mention some of these and other criticisms. But our particular claim 188
is
RACE, ETHNICITY, AND INTELLIGENCE
Cultural Bias One common argument made
Tests
in
against racial differences in intelligence
that the intelligence tests are "culturally biased."
knowing words and information more ple than to blacks
World War
is
made
test for
I,
Scoring well depends on
familiar to white, middle-class peo-
and Latino Americans. There
The problem of cultural content goes back in
is
is
some
to the early
which included questions such
as:
truth to this claim.
"Alpha"
tests
used
"The Pierce-Arrow car
Buffalo, Detroit, Toledo, or Flint?"* Similarly, a contemporary
in:
small children requires that they be familiar with a flag pole and
that they
know
the etiquette called for
ing to another person.**
To be
upon having broken an item belong-
in a subordinate minority
is,
in part, to
be
and socially isolated from the majority, thereby reduc-
culturally distinct
ing the chances of answering such questions "correctly."
Herrnstein and Murray respond to these criticisms by echoing Berkeley
psychometrician Arthur Jensen's claim that blacks score below whites on "abstract" test items, too. Thus, the gap cannot be due to cultural content,
What
they argue.
is
"abstract," however,
ing and reciting digits in reverse order
and Murray
stress
—
on such
not at
all clear.
Even memoriz-
"intelligence test" Herrnstein
requires familiarity and comfort with numbers.***
Also, as conservative economist cisely
is
—one
Thomas Sowell
sorts of "abstract" items that
has noted,
it
was
pre-
European immigrants, such as
Jews, demonstrated their supposed dim-wittedness in the early twentieth
century.****
Clarence
S.
Yoakum, Army Mental
Tests (1919), pp. 260-61, quoted by Lears,
Fables of Abundance, p. 220. ** Questions quoted in Science for the People, "IQ."
*** See Stephen
J.
Ceci,
On
Intelligence, esp. chap. 9,
gence?" Herrnstein and Murray also
show black Abilities of
Damned
Mankind,"
Abstract
Is Intelli-
Those studies have, however, been persuasively undermined.
inferiority.
See Kamin, "Lies,
"How
like to cite reaction-time studies that purportedly
p.
Lies,
and
Statistics," pp.
87-89; Irvine and Berry, "The
51; and Nisbett, "Race, IQ, and Scientism," p. 44.
**** Sowell, "Ethnicity and IQ."
broader than ity,
that:
Scores on achievement tests are the products of inequal-
here and elsewhere
We
return
grammed to score
now
in the
world.
explicitly to the explanation
in figure 8.
1
.
we
outlined earlier and dia-
Ethnic groups in lower caste or status positions tend
poorly because their position leads to socioeconomic deprivation,
189
CHAPTER group segregation, and a stigmatized
8
each of which undermines
identity,
performance on psychometric measures of intelligence. As foundation for
we
these claims, is
first
show
that this pattern of ethnic differences in scores
not special to blacks or Latino Americans today;
groups earlier
American
in
it
existed for other
history and exists for other groups around the
world.
Many immigrants faced prejudice and discrimination in coming to America (although not slavery or peonage). As early as the colonial era, Benjamin Franklin objected to allowing Germans "Palatine boors," he
—
called
them
—
into Pennsylvania.
often considered those
Around 1900 immigrants
—
who came
By
earlier to these shores
immoral, subversive, and
dull.
new
later as
time they were
at the
— were
cheaper transportation available least
Those who came
the president of M.I.T., an economist, stated that the
eastern European groups
were the
42
Italians,
Russian Jews, Poles, and other
inferior to earlier immigrants. at the
turn of the century, the
of their kind, were failures, and lacked
fit
Because of
new
talent.
arrivals
43
the 1920s the nativists could point to scientific evidence for such
charges:
New immigrants
on standardized
and
their children did
worse
than did "old-stock" Americans.
tests
in school
and worse
As Thomas
Sowell,
writing in the conservative magazine American Spectator, points out, the
conclusion that European immigrant groups were of below-average
gence "was based on hard data, as hard as any data
in
The Bell Curve. These
groups repeatedly tested below average on the mental
War
I
era,
both in the army and in civilian
poorly on the "abstract"
low
that Carl
test questions.
44
life."
The
test
intelli-
tests
of the World
They scored
especially
scores of Jews were so
Brigham, an early scholar whose work was often drawn upon
in policy debates
during the 1920s, wrote that the results "would rather
disprove the popular belief that the
Jew
Karl Pearson, a founder of modern
statistics, also
is
highly intelligent." (In Britain,
discovered that Jewish
children scored below gentile children on mental ability.) Generally, immigrants from southern and eastern Europe were assessed as significantly inferior to the
scores of the
"Nordic race." 45 Yet a couple of generations
new
later,
the test
groups had risen dramatically, matching or exceeding
those of earlier-arriving white Americans. Sowell explains that low scores reflect groups' positions "outside the cultural
Western society
.
.
.
whatever
their race
European immigrant groups arrived
mainstream of contemporary
might be." Southern and eastern
as cultural outsiders but
became
insid-
ers during the course of the twentieth century.
Sowell's argument
is
not unlike our own. (Sowell's argument and evi-
dence were presumably known
to Herrnstein
190
and Murray, because he
first
RACE. ETHNICITY. AND INTELLIGENCE published them a quarter-century ago.) Today, blacks and Latinos are cultural
than
— and economic and
was
social
—
outsiders in
far
8.
lists
1
is
We
where
are rising.
a sample of studies from around the world that have
ined group differences in test scores.
The information
—and they
still
more profound
true for the Europeans. Today, black test scores are about
those immigrants' scores were in the 1920s
Table
ways
exam-
note the limitations of this table:
not complete; the tests and procedures varied consider-
is no simple way to compare the size of some cases, a few studies have yielded more mixed findings than those shown here, although virtually none found group contrasts opposite to those shown here; and a narrowing of differences appears to be happening in some nations not unlike the narrowing in the United
ably from study to study, so there
group differences; in
between blacks and whites. All
States
this said, the table still captures the
general pattern.
The
John Ogbu has argued,
table shows, as
inferior in status
ligence"
tests.
that ethnic
groups that are
and caste position score worse on achievement and
Ogbu
also noted a tendency, in cases
"intel-
where the subordinate
group differs racially from the dominant one, for members of the superior group
to explain those differences as genetic.
46
Yet genetic or racial expla-
nations cannot explain the pattern of group differences.
A reading cally,
of the table shows that ethnicity or race, understood biologi-
cannot be the cause of the test-score differences. Particularly striking
are the substantial gaps in test scores
between groups of the same ethnicity
or race in countries like Israel, Japan, and South Africa. In Israel, Ashkenazi (Western-origin)
Jews have
(Eastern-origin) Jews.
47
origin
Jews have had
historically scored higher than Mizrachi
Since the founding of Israel in 1948, Eastern-
less wealth,
power, and status than Western-origin
ones; the latter have often considered and treated the former as culturally "primitive." Although the differences have been narrowing and intermarriage has been growing, Eastern-origin children
origin children in
many
still
score below Western-
aptitude tests, with differences as large as two-
thirds of a standard deviation.
(The black-white gap
in
one standard deviation.) Moreover, researchers continue
America
is
about
to find ethnic dif-
ferences in scores after controlling for ethnic differences in social class. In Japan, residents of
Korean ancestry are so
racially indistinguishable
from "pure" Japanese that many Korean youth "pass" their ancestry
to pass is
in
school by hiding
and taking Japanese names; some continue passing
ward by cutting family
ties
and fabricating new
48
identities.
after-
The motivation
strong because the Japanese have historically discriminated
against Koreans and consider the Koreans to be a problem group
191
—
dull,
"
CHAPTER Table
8
8.1
Group Differences Around
the
World
Status or Caste Position
Low
High
Country United States a
_
Test Scores, School Success
Low
High
Whites
Blacks
Whites
Blacks
Whites
Latinos
Whites
Latinos
Whites
American Indians
Whites
American Indians
English
Irish, Scottish
English
Irish, Scottish
Protestants
Catholics
Protestants
Catholics
d
Whites
Aborigines
Whites
Aborigines
Zealand e
Whites
Maoris
Whites
Maoris
South Africa f
English
Afrikaaners
English
Afrikaaners
Belgium g
French
Flemish
French
Flemish
Jews
Arabs
Jews
Arabs
Western Jews
Eastern Jews
Western Jews
Eastern Jews
Nontribals
Tribal people
Nontribals
Tribal people
Brahmin
Harijan
Brahmin
Harijan
High caste
Low
High caste
Low
Slovaks
Gypsies
Slovaks
Gypsies
Non-Burakumin
Burakumin
Non-Burakumin
Burakumin
Japanese Origin
Korean Origin
Japanese Origin
Korean Origin
Great Britain
15
Northern Ireland' Australia
New
Israel
11
India'
Czechoslovakia*
Japan k
a.
The white-black and Anglo-Latino
American Indians:
see, for
differences are reported in The Bell Curve and
Research by Richard Lynn discussed
c.
Lynn
scored higher on tions based
on
"Home Background,"
in
for aspects of parental
other places.
On
Benson, "Ireland's 'Low' IQ."
presents evidence that
among young men
an intelligence test than did Catholics. (There
their table
many
caste
example, Church, "Academic Achievement."
b.
et al.,
caste
4 show
that the religious difference in
socioeconomic
status.
in
Northern Ireland, Protestants
was no difference among young women.)
IQ among males
A few newspaper stories on
Catholics scored one standard deviation below Protestants in IQ, but
persists
even
The Bell Curve reported
we have been
Calcula-
after controlling that, in Belfast,
unable to find the source
for
that claim. d. Klich,
"Aboriginal Cognition and Psychological Science"; Clark and Halford, "Does Cognitive Style Ac-
count for Cultural Differences?" e. f.
Ogbu, Minority Education and Caste;
St.
George, "Cognitive Ability Assessment
in
New
Zealand."
Verster and Prinsloo, "The Diminishing Test Performance Gap."
g.
Raven, "The Raven Progressive Matrices," esp.
h.
On Jews
versus Arabs: Kugelmass et
al.,
fig. 2.
"Patterns of Intellectual Ability";
Jewish high school students passed matriculation exam versus ary 12, 1995;
cf.
Lieblich et
al.,
15%
news item
that, in
1992,
26%
of
of Arab students—Jerusalem Reports, Janu-
"Patterns of Intellectual Ability."
On Western
versus Eastern Jews: Gross,
"Cultural Concomitants of Preschoolers' Preparation for Learning"; Dar and Resh, "Socioeconomic and Ethnic
Gaps." i.
on
For an overview on Indian caste differences, see Das and Khurana, "Caste and Cognitive Processes." Also,
tribal groups, see
Concentration of
'g'
Gupta and Jahan, "Differences
in
Cognitive Capacity"; on Brahmins, Shyam, "Variations
in
Level Abilities"; and on caste: Das, "Level-I Abilities of Socially Disadvantaged Children,"
192
RACE, ETHNICITY, AND INTELLIGENCE ill-mannered, often criminal.
Even
the
word "Korean"
is
considered a
slur.
Children of Korean ancestry have historically done relatively worse in school and on aptitude
tests.
Another low-status group
49
Japan also racially indistinguishable from
in
the majority consistently also scores lower than the majority tests.
on aptitude
These are the Burakumin, descendants of people who formed a
sort
of "untouchable" caste in feudal Japan. (They, like India's untouchables, dealt with "filth," such as burials
and carcasses.) Burakumin
who
try to
"pass" can be unmasked only by careful genealogical analysis, often con-
ducted preparatory to marriage. to
Still,
them
the majority Japanese consider
be inferior and often discriminate against them. The Burakumin, like the
Koreans their
in Japan, are
prone to economic failure and social problems.
children have systematically done worse on aptitude
Burakumin children
in the
same
schools.
tests
And
than non-
50
In South Africa during the 1950s, children of English origin scored
higher on aptitude and intelligence tests than did Afrikaaner (Dutch-origin) children.
The gaps between
wide as a half
these
two northern European groups ran
to a full standard deviation. In the
as
1960s the English- Afri-
kaaner differences narrowed, and by the 1970s they seemed to have dis-
The convergence of Afrikaaner and English scores coincides with the rise of Afrikaaners to power in South Africa after generations of subordination to the English (and before conceding power to the native appeared.
black Africans in the 1990s). 51
gap has
In all three of these cases, evidence suggests that the test-score
narrowed over recent generations as subordinated groups began slowly toward political and social parity.
A similar trend is
Still,
move
noticeable over
the twentieth century for African Americans, despite continuing
and geographical
to
economic
isolation.
the basic differences in the table persist.
inferior test scores of eastern
These
results, just like the
and southern European immigrants
to the
United States seventy-five years ago, cannot be reasonably explained by
and Das and Padhee, "Level
II
Abilities of Socially Disadvantaged Children."
Some
studies of
advanced school students do not show caste differences. Rangari, "Caste Affiliation," showed nonsignificant differences in a small sample. Sandhu,
"A Study
of Caste Differences," also found
no differences, but the sample sizes of "backward" and "scheduled" castes
in that
study were also
quite small. j.
k.
Adamovic,
"Intellectual
On Burakumin:
Development and Level of Knowledge
in
Gypsy
Pupils."
Shimahara, "Social Mobility and Education"; on Koreans: Lee, "Koreans
Japan and the United States"; DeVos and Wetherall, Japans Minorities.
193
in
CHAPTER The most
inherent genetic differences.
8
logical explanation
Where
is this:
ethnic groups exist in castelike or near-caste relationships, youths in the
subordinate groups do poorly on so-called intelligence tests and similar
academic assessments. They do so because tural
of the
fewer material and cul-
resources their families have, because of segregation, and because
they understand the limitations placed on their aspirations in those societies.
Black and Latino American youth
like other
more
youth
in the
the United States are simply
in
world with subordinate caste
status.
52
Let us look
closely at this process in the United States.
Socioeconomic Deprivation
Most
critics
of The Bell Curve concede
youth score lower on achievement
that the difference can be entirely
backgrounds of the two groups
—
vation of minority individuals.
On
lower income, much less wealth,
that,
tests than
on average, black and Latino whites ones do, but they assert
accounted for by the differing family
that
is,
by the material and cultural depri-
average, minority children
grow up with
less nutritious diets, unhealthier environ-
ments, worse medical care, and so on than do white children. These conditions impair learning. Black and Latino children also
from poor home environments value education
5^
in
which learning
Their families
but cannot create the physical conditions for
lack the kind of social skills to support
help with
more often come
is difficult.
homework
or
how
it
and often
how
example, knowing
(for
to find the best teachers).
minority students are generally less interested
it
in
The
result
is
to
that
academics, have difficulty
focusing on their studies, and are more physically active. Such conditions require the ghetto schools to do
more than they would have
to
do
in
middle-
class neighborhoods, and yet the ghetto schools usually have fewer re-
sources with which to do
it.
If
blacks and Latinos were, on average, equal
to whites in social conditions, critics of racial explanations say, the perfor-
mance gap would
disappear, showing us that race and ethnicity are irrele-
vant to intelligence.
The
typical
way
that researchers evaluate this
adjust individuals' test scores for differences
economic circumstances, simulating
argument
is
among them
a situation in
statistically to in social
and
which minorities and
whites laced equal disadvantages. (See appendix 2 for an introduction to multiple regression analysis.) Herrnstein and Murray version of this procedure
when
seminated by the media, that
(p.
2X8) perform a
they show, in a graph that was widely disat
every level of "parental socioeconomic
194
RACE, ETHNICITY. AND INTELLIGENCE status,"
from lowest decile
Armed
the
social
to highest decile, blacks scored
Forces Qualifying Test.
From
background cannot account for the race gap
One problem with
work, as
that
we
below whites on concluded
that exercise, they
that
in test scores.
pointed out
in
chapter 4,
Herrnstein and Murray's statistical treatment of parental status
is
is
faulty
that
and
so provides an inadequate test of the argument that differences in deprivation explain differences in scores.
Another problem
is
that there are
many
more conditions besides parental education, income, and occupation disadvantage the minority children and that also affect learning. condition
is
in wealth.
income accumulate over the
two-earner, black couple brings
kin,
and because even small annual gaps
is
median young,
years. So, for example, the
home
8
1
percent as
median young, two-earner, white couple. But
worth
This failure arises,
because annual incomes fluctuate widely, because income figures
in part,
do not capture financial help from
the
One such
wealth. Annual income differences between whites and blacks
do not capture the true scale of the differences
in
that
much annual income
as
the black couple's net
only 18 percent as great as the white's. 54 Other differences be-
tween the races not captured
in the typical analysis
range from rates of
breastfeeding, to exposure to lead poisoning, to parents' clout in the school
system. 55 In a recent study of the that Herrnstein
young children of
the
same survey respondents
and Murray analyzed, Jonathan Crane found
that the black-
white gap in math and reading scores could be totally accounted for by the
following differences between black and white children: family income, size of household, proportion of students in the school the
tended
who were
poor, the age the child
and, most important,
how much
mother had
was weaned, whether the
home was
child
at-
was
read
to,
tive
and cognitively stimulating. Black and white children similar to one
the
another in these conditions performed similarly on the
Crane concludes
emotionally suppor-
tests.
Consequently,
that genetics is irrelevant to explaining the test-score
gap. 56
Such studies are controversial, however. The basic charge against them by
racial theorists is that simulating equal social
cal controls is misleading.
circumstances by
statisti-
These circumstances are themselves, psycho-
metricians such as Arthur Jensen contend, the product of low intelligence.
home
environ-
ments for children; those parents are disproportionately black.
Statisti-
For example, parents with low intelligence provide poor
cally eliminating the effect of the
home environment masks
the fact that
poor intelligence breeds poor intelligence genetically; the poor 195
home
envi-
CHAPTER ronment
is
just a by-product of
turn, researchers
who
8
low parental
intelligence, they argue. In
claim that the racial gap can be explained by the
individuals' personal environments have rejoinders to this charge. 57
debate will continue. Certainly,
is fair
it
The
conclude that some, perhaps
to
most, of the minority-white difference in scores arises from impairing social conditions,
such as the poverty that black and Latino children more
often face for longer periods than white children do. But these disadvantages of family class position are not the only ethnic caste impairs test performance and
may
ways
that being in a lower
not be sufficient to explain
the test-score gap.
Segregation and Isolation
As we have shown, African Americans Americans; Latino segregation regation
is
is less
are severely segregated
from white
58
Social seg-
severe but
also great. Friendships usually
black-white intermarriage
is rare.
still
sizable.
do not cross
racial lines,
Schools, especially in the larger
and
cities,
are also highly segregated.
Residential, social, and school segregation blacks, that
children
But
cussed
some minority more education or
have. Sociologists have found that the better
this is not so for
income does not major reason
so profound, especially for
often overrides middle-class advantages that
it
may
income whites have, the live.
is
is
earlier.
59
and safer the neighborhoods
in
which they
black Americans; for them, more education or
translate as well into better or safer neighborhoods.
The
we
dis-
the subtle and not-so-subtle housing discrimination
Therefore, middle-class blacks often have to live in or near
low-income communities, something middle-class whites need not do. Similarly, poor black families are concentrated together
poor white families ents strive to help
who
are.
60
much more
them academically must
still
live
and learn with children
lack such social support. Again, equally poor white children are
less likely to
have
than
This means that poor black children whose par-
to deal with
much
poor neighbors and disadvantaged fellow
students.
Segregation impairs school and revealed by analyses of individual
test
performance
traits.
in
ways
For one, schools
that are not
in segregated
black neighborhoods tend to be poor schools, and differences schools do affect learning and test scores (see chapter that minority students,
7).
among
Research shows
even with the same amount of schooling as compa-
rable white students and even in nominally academic tracks, are less likely
196
AND INTELLIGENCE
RACE. ETHNICITY.
61 than whites to have had classes in advanced mathematics or science.
Minority children score low on standardized for this reason,
tests
such as the
AFQT in
part
because the schools they must attend do not expose them
to
important curricula. Furthermore, most blacks and Latinos must attend schools with higher classroom size and fewer and older teachers
who
facilities.
Many
are creative leave inner-city schools because they tire of the
daily battles, leaving the ghetto with an oversupply of teachers
burnt out or
62
who
are inexperienced.
Such differences
who
are
affect students' abili-
score high on standardized tests. In addition, persistent de facto seg-
ties to
regation creates conditions that lower youths' self-esteem and ultimately
lower their
test scores.
Blacks and Latinos have to
live
among
increasingly
impoverished and marginalized members of their groups; they must cope with the disaffection and disruption of poorer students. Comparable white youth, in contrast, do not face such barriers to learning, because they are not barred from communities of their choice.
And
Much
then "concentration effects" accentuate the problems.
search has
shown
that
re-
people are influenced by the social climate around
them, influenced, for example, to vote in ways one would not expect given their individual characteristics. Similarly, research
ple are strongly influenced
about the crowd with
by
shows
whom their children hang out).
In
Simply
put,
would be expected given
their individual
even "good" kids can turn "bad"
if
more youth
cally
who might
the setting they are in is
is
negative.
be otherwise destined to do well academi-
do not do well because they are
trated problems, that lead
get
backgrounds. 63
heavily comprised of troubled youth and the social climate
Minority children
who worry
neighborhoods and
schools with high concentrations of delinquent youngsters, into trouble than
young peo-
that
their peers (no surprise to parents
them
in
segregated settings, with concen-
to learn less.
Together, the deprivation and the segregation of blacks and Latinos go a long
way
to
account for the persistent difference
ing to appeal to genetics. But
That
final bit
reality
—by
tinos, or
The
we
of difference cannot be erased
raising the
statistically
without need-
— nor probably
in
economic conditions of individual blacks and La-
perhaps even by moving them into desegregated communities.
residual difference
emerges from the fundamental
group, African American or Latino American.
being black or Latino in America
is
to
be
in a
handicap youth
in test
taking and in
197
identity of the
The profound
reality
of
lower caste position. That
identity creates specific expectations, anxieties, turn,
in scores,
suspect that a residual difference remains.
and reactions. These,
more important
tasks as well.
in
CHAPTER
8
Stigmatized Identity
Youths
who
see themselves as fated to a lower caste position
them
cept the description of in self-destructive
whatever their
ment
ways. Or they do both. Thus,
ability or learning, score poorly
many come
meate and
to ac-
on intelligence and achieve-
tests.
Nonwhite youth have heard fail;
come
Or they rebel against it many nonwhite youth,
the majority provides.
their
that they are unintelligent
to fear that this is true. After
all,
low-income neighborhoods. The fear
When
their preparation.
and fated
examples of
to
failure per-
affects their confidence
tested, then, they find that the results con-
firm what others, and sometimes they themselves, believe. Experiments led
by psychologist Claude Steele show assistants
this
He and
his
drawn from
the
process in operation.
gave black Stanford University students a
test
Graduate Record Exam. The researchers raised the specter of feriority for a
measured
random
set
racial in-
of the black students by telling them that the
test
reminded
their personal abilities; alternatively, the researchers
them of race by asking them to check off their ethnicity on a questionnaire. They told another set of black students, also randomly chosen, that the study was simply psychological research. The first group of students performed notably worse than the second group did. Steele explains that when black students are explicitly confronted with the stereotype of black lectual inferiority, the resulting anxiety
inferiority?" If this
—
happens
interferes with to black
youth
can imagine that the process
— "Am
I
intel-
going to reveal blacks'
performance and so they do perform worse.
who have made it all the way to Stanford, one is yet stronger among other black youth. (Re-
inforcing Steele's theory of "stereotype vulnerability" are studies that
showed
that the
same process occurred among women and among white
men taking math tests. When told in advance that women usually do worse than men on the test, female undergraduates scored poorly; but when told that women did equally well, the female undergraduates scored as highly as male ones did. Similarly, white men performed worse when they were contrasted to Asians.)
Here
is
64
one way The Bell Curve and similar books help create
they claim to explain, the lower performance of minorities
apprehension.
As
university teachers of teenagers,
the popularization of
we
that
—by
which
instilling
are concerned that
The Bell Curve has demoralized our minority
dents, reinforced nagging self-doubts, and
worsened the problem.
also falsely inflate the self-images of white students.
198
It
stu-
may
RACE, ETHNICITY,
AND INTELLIGENCE
Black and Latino youth respond to the caste system with fatalism, with anxiety,
and sometimes also with
experience suggests
that,
hostility.
Their parents' and neighbors'
whatever abstract value they place on education,
schooling will probably not pay off for them.
If the
"program" seems
promise humiliation and dead-end jobs, many decide not
to
to
go along with
They view cooperation with white institutions, such as capitulating to the enemy. They develop an oppositional cul-
the program.
schools, as ture,
one
in
which, for example, performing well
"white" or "doing the Anglo thing." Prestige life,
in
is
school
found
is
such as sports. Even good students face peer pressure
standards.
seen as acting
in other
to
realms of
defy white
65
low self-esteem and
In this climate of collective
resistance, school
suffers. Similarly, in testing situations, students might, in a
of disdain, haphazardly answer
test questions.
complete show
(When one of
Martin Sanchez Jankowski, taught junior high school
work
the authors.
in Detroit,
he en-
among Gypsy children. They questions randomly; some deliberately
countered a consistent test-taking pattern
would answer
the standardized test
answered them incorrectly. They said they did not care about school, the only reason they attended
was
trouble with the law.) There
evidence of such attitudes toward the
among
the minority
Korean youth
in
is
to
NLSY respondents who
Japan
—but not
in
that
avoid having their families get into
took
America
AFQT
66 it.
— seem
to
show
similar pat-
terns: low academic expectations and a history of disruptiveness in the
schools.
Some Korean
Japanese children see sports as the only arena for
success in school. Their parents are also often skeptical that they can suc-
ceed through academics.
hope [my son]
is
One
father told a visiting anthropologist, "I just
physically tough and strong."
67
British sociologist Paul
among white workingclass youth in England. (In England, class differences take on some of the quality of caste differences.) Believing that they are doomed to lousy jobs Willis has described such an oppositional culture
or no jobs at
all
and that the society demeans them, they denigrate school-
work and honor delinquency. 68 Similar reactions have been observed where, for example,
among
Eastern-origin Jews in Israel.
have described are not unique to lower-caste
In a caste it
to
69
The
else-
patterns
we
American blacks or Latinos, but common
groups around the world.
system where race
is
a master
trait
defining people's identities,
should be no surprise that race matters. America has for centuries treated
people according to their race and
Young people understand
it still
does, albeit not as severely now.
their positions in the racial caste system, irre-
199
CHAPTER
and Math: A Parallel Story
Boys, Girls,
The idea
that
males and females
differ in their "natural" talent for
many Americans. Women
matics seems intuitively plausible to likely to take fields.
Not
math-heavy subjects
gap
in
in
matheare un-
school or to pursue careers in such
American media
surprisingly, the
fundamental differences
8
in brain structure or
are full of speculations that
chemistry explain the gender
math.
Recent research by David Baker and Deborah Perkin Jones reveals
wrong
that intuition
is.
Using math
how
given in 1982 to 77,000 eighth
tests
graders around the world, they found that, on average, boys and girls score
about the same. But there countries, boys
do
is
considerable variation
do
better; in others, girls
among
better;
and
nations. In
in
about equally well. Girls tend to perform better relative to boys countries where
modern
mance
more women go
industries. Also, girls in the relative to
boys since
speed with which a nation's
girls'
1982
in those
women
hold jobs in
had improved
their perfor-
and more
test
were conducted
earlier tests
girls
which women's participation sum,
to college
some
some, they do
in
1964.
The
closed the gap varied with the degree to
in their country's
math performance responded
to
workforce had grown. In
women's
career opportunities
in their nations.
These and other
results support
Baker and Jones's interpretation of the
stereotypical gender gap: If
male students are afforded the possibility of future educational and
occupational opportunities as a function of their performance in
mathematics, then they
may
try harder, teachers
more, and parents and friends a
may
domain of performance they should take
hand, female students,
mathematics as
less
number of ways by nity structures
who
may encourage them
help them see that mathematics seriously.
On
are faced with less opportunity,
is
the other
may
see
important for their future and are told so in a
teachers, parents, and friends. In short, opportu-
can shape numerous socialization processes that shape
performance.* Substitute "white" for "male," "minority" for "female," and "academic" for "mathematics" in this quotation. *
The
logic
Baker and Jones, "Creating Gender Equality,"
"Cross-Cultural Gender Differences
in
is
p. 92.
the same.
For similar
Mathematics Education."
200
results, see
Hanna.
RACE, ETHNICITY, AND INTELLIGENCE spective of their families' education or wealth.
some
stance expected of them,
rebel against
One result is the same mance on tests in school. What about Asian Americans? They
Some
all
it,
adopt the resigned
probably worry about
it.
whatever the reaction: poorer than expected perfor-
on standardized the data
Does
tests.
that point to a racial explanation?
on the Asian advantage
ferences are tiny.
70
more highly than whites do
score
in intelligence is
mixed;
No.
First,
at best, the dif-
Second, the great bulk of Asian American youth today
are the children of, or are themselves, "voluntary immigrants," quite dif-
ferent
from the experience of the "involuntary minorities," blacks and
Most have arrived since the 1965 liberalization of the immigralaws, many coming with middle-class backgrounds. Third, early in
Latinos. tion
the century, Asians, like Jews, scored
States
on
position.
An
tests; their
improvement
below native whites
is a result of the
change
in the
United
in their social
71
additional feature of the Asian case does have wider implications for
understanding race and academic performance. The success of Asian and of Asian American children
how much more
in
school can be satisfactorily accounted for by
time, attention, ambition, and effort Asian children and
their families put into education. Ironically,
white Americans' disadvan-
tage relative to Asians seems to rest, in part, on the ural" talent.
American idea of
White mothers, children, and teachers are much more
attribute success in school to innate intelligence than are Asians;
instead typically attribute success
Ogbu
reports that the Chinese
viewed believe to
prove that
sivity; the
much more
it is
so.
Asians
to hard work. Indeed,
American high school students he
that they are better than white students
The American
"nat-
likely to
John inter-
and so work harder
belief in "natural" talent leads to pas-
Asian belief that talents are learned leads to more hard work and
better performance.
assimilate into
72
(It
will
American
be interesting to see what happens as Asians
culture.)
For African American and Mexican American youth, even equalizing family backgrounds and community settings probably cannot completely close the test-score gap, because the problem tized caste.
rooted in being in a stigma-
But the gap can be closed further. As European immigrant
groups were accepted
some of
is
in
America,
their test scores rose dramatically. In
the societies listed in table 8.1, recent research
seems
show
to
smaller test-score differences than earlier studies did. Those changes Eastern-origin Jews in Israel, Maori in Africa, even
Burakumin
in
Japan
New
Zealand, Afrikaaners
—appear
to
match
the
in
—
for
South
weakening of
ethnic caste and status barriers in those societies. For example, intermar-
201
CHAPTER
8
A Thought Experiment One can
appreciate the importance of race as a social reality in
America
with a thought experiment once suggested by an economist.
We
readers to imagine that a scientist developed a potion that
would
ask white turn a
white person black. Everything else about the person would be exactly the
now
same, except that he or she would
much would
look African, permanently.
the experimenter have to pay
you
in
potion? The answer might be one estimate of what
America even when everything
who
feel that blacks are
How
order for you to take the it
else about the person
costs to be black in the same. (Those
is
advantaged these days should, of course, be will-
ing to pay the experimenter for the privilege of
becoming
black.)
riage, the greatest breach of the caste lines, has increased in these cases. In
between black and
the United States, however, the caste lines, especially
white, remain firm.
Conclusion The Bell Curve treatment of tory and destructive.
It is
racial differences in intelligence is
score below whites on standardized ral" difference,
inflamma-
also wrong. Yes, blacks and Latinos consistently tests.
But notions
that this is a "natu-
one resulting from genetics, are inadequate. Individual
blacks and Latinos confront these tests of school- and school-like knowl-
edge burdened by centuries of disadvantages: family histories rooted servitude, poverty, and cultural isolation.
They
in
also carry heavy disadvan-
tages rooted in conditions today: continuing discrimination, low income,
concentration in problem neighborhoods, and inferior schools, to
name
a
few. Like other lower-caste groups around the world, their poorer perfor-
mance
in school
and school-like situations can be understood as the
result
of socioeconomic deprivation, segregation, and a stigmatized lower-caste identity.
African Americans and Latino Americans score below whites be-
cause to be black or Latino in the United States
But the gap tion, as
it
is
closing,
is
to be
below whites.
by the equivalent of several IQ points a generamoved from the periphery of Ameri-
closed for other groups that
can society toward
its
center. If
we choose 202
to
exaggerate the remaining
RACE, ETHNICITY. AND INTELLIGENCE ethnic differences, to treat
down
the convergence. If
inequality generally,
more
quickly.
Our
is
them
we
as natural and inevitable,
we
will slow
see, instead, that this inequality, like social
under our control,
we can choose
fate as a multiracial nation
is
gap
not in our stars, to para-
phrase Shakespeare, nor in our genes, but in our hands.
203
to close the
CHAPTER
!
*
9
Confronting Inequality
in
America:
The Power of Public Investment
w
have shown
e
that
American inequality cannot be explained
terms of people's "natural" intelligence or other supposedly genetic
why
Understanding inequality requires explaining
individuals end up
where they do on the "ladder" of success and explaining why the
is built
way
ity is the result
desserts"
it is.
For each
who
task, those
of inequality in natural talent
—have
—
in
traits.
that ladder
argue that social inequalthat
people get their "just
vastly underestimated the importance of the social en-
vironment.
—
While genetically assisted advantages looks
—
portant.
affect
how
One reason environment
individual
example,
traits,
in
matters so
more
being male,
is
that
it
is
more im-
determines
how
also matters because
it
would have been a major
trivial.
it is
egalitarian societies,
is it
In societies with severe
a heavy burden on individual
is less so.
Social environment
directly structures the opportunities individuals
have. Family circumstances tural advantages,
much
societies, near-sightedness
on women, being born female
attainment; in
health,
even genetic ones, translate into material advantages. For
most
handicap; in a society with eyeglasses, restrictions
height,
high individuals climb, social environment
— number of
siblings, parental
income, cul-
and so on; the quality and quantity of schooling; neigh-
borhood conditions; job opportunities; and other features of the text significantly boost or hold
back the individual, whatever
social conhis or her
talent (see, especially, chapter 4).
As
for the structure of inequality, individuals' native abilities are largely
irrelevant. Investments in
improving
skills,
such as the expansion of higher
education in the 1950s and 1960s, and major disinvestments, such as
ducing health care for infants, can
But these,
alter the
too, refer to societal policies
re-
shape of inequality in a society.
and structures, not distributions of
"natural" talent. Nations and historical eras differ in the degree of inequality
they have because their economic, cultural, and political circumstances
differ.
For example, the gap between well-off and worse-off Americans'
standards of living widened substantially in the last twenty-five years, even
while American intelligence stayed constant or increased (see, especially,
204
CONFRONTING INEQUALITY chapter
5).
To understand systems of
AMERICA
IN
inequality,
we have
to think
beyond
individuals to social structure.
Moreover, both operations of the social environment tures the ladder of inequality
and the way
it
—
way
the
it
provides advantages for
struc-
some
individuals and disadvantages for others as they clamber up that ladder are themselves shaped, in part,
through
ters 5
7).
by
political choices (see, especially,
Those choices concern
way we provide schooling and job tions, subsidies,
chap-
the rules of the marketplace, the
opportunities, government interven-
and taxes. In these ways, we, as citizens, decide the
equality, both of opportunity
and of
result, that
in-
our nation will have.
The Myth of "Just Desserts" These conclusions are not novel. In ago, in response to an
IQ controversy
Bane, a noted scholar Health and
Human
many have been commonplace
in
and public discourse for decades. Over twenty years
the social sciences
gist
fact,
who
later
stirred
up
became an
in the early 1970s,
Mary Jo
Department of
official in the
Services, and Christopher Jencks, an eminent sociolo-
and policy analyst, summarized the
known
facts
then in an article
about IQ "myths":
IQ
tests
measure only one rather limited variety of intelligence, namely the
kind that schools (and psychologists [more precisely, psychometricians]) value. Scores on such tests
mance
in
The poor
most adult are
show remarkably
little
relationship to perfor-
roles.
seldom poor because they have low IQ scores.
.
.
.
They
are
poor because they either cannot work, cannot find adequately paying jobs, or cannot keep such jobs. This has very
little
to
do with
their test
scores.
Socioeconomic background has about the same influence as IQ on how
much schooling a person gets, on how much money he makes.
the kind of occupation he enters, and
1
Since those words, research has,
ronment
if
anything, indicated that the social envi-
—encompassing family background,
schools, and
community
—
is
even more important than Bane and Jencks thought.
Why, myths
after all this well-established scholarship
persist?
Why
do Americans
—
at least,
debate such issues in magazines and newspapers
205
debunking them, do the
those "opinion leaders"
— seem so receptive
who
to the
CHAPTER
9
idea that inequality simply reflects individuals getting their just rewards for their "natural" abilities?
A simple
There are several possible answers.
one
people to believe in the justice of current inequality
books
like
The Bell Cur\>e who are
suggested
that, "like the divine right
it
society.
Bane and Jencks
economic inequality "help[s] legitimate the
Certainly,
the interest
it
is in
— material and
we
ideological
—of
the "haves" to
myth
are
that.
may be
particu-
when inequality widens, becomes blatant, and cries out The 1920s, for example, was such an era. The economic
high in eras
for explanation.
boom much
of the decade benefited the affluent, especially stock speculators, so that inequality
flourished and
now
2
predetermined and unchangeable. But
is
Receptiveness to messages like that of The Bell Curve larly
in-
status quo."
believe that the bases for belief in the "natural inequality"
deeper and more genuine than
many
of kings," the myth that genetic
equality explains
endorse the theory that inequality
serves
especially readers of
and Murray repeatedly
(as Herrnstein
members of
note) overwhelmingly advantaged
that
is
—
in a similar era.
also be high
in that
decade eugenics
liorate inequality
We
Receptiveness to notions of natural inequality
now because
are
may
of the widespread impression that efforts to me-
—antipoverty —have
tory education, etc.
were even
widened considerably. And
was applied against "darker" European immigrants.
programs, affirmative action, compensa-
"failed."
Whether they
really tried, is a matter of intense
really
have
failed, or
and serious dispute. 3 But
it is
the opinion of the journalists, politicians, policy intellectuals, and talk-
show have
hosts that seemingly matters here. If social intervention appears to
opinion leaders are more likely to also believe that
failed, then
equality
is
in-
fated and immutable.
Yet another part of the answer
that this social theory is consistent with
is
4 longstanding American beliefs about inequality. Those beliefs generally
hold
— although —
fications
the
there are, as in any belief system, contradictions and quali-
that inequality of result
outcome of a
opportunity.
fair contest in
What
is
perfectly acceptable, so long as
which
citizens of the
all
—
after all, the leaders of the
World inequality was
"unfair."
is
contenders have had equality of
new American
nation in the nineteenth
century resented about the Old World was not so
wealth
it
much
its
Revolution were wealthy
Both the privileges
that
inequality of
—
as that
Old
decadent aristocrats
received by luck of birth and the interference of autocratic governments in the
economy robbed hard workers of
free-market the
way and
spirit, let
Americans wanted
the "fruits of their labor."
—
the race be run.
206
still
want
—everyone
5
In true
to get out of
CONFRONTING INEQUALITY If this is
that this is
how
people want their society to be,
how, for most
part, their society
and can say
ically consistent
in
AMERICA
IN
a simple step to believe
is
it
(Most people are not ideolog-
is.
one conversation
that there is equality
of
opportunity in America but then say in another conversation something like,
"Rich kids get
The context of The
the breaks."
all
and the social position of those
who
read about
Bell Curve debate
will favor the first re-
it
sponse, that America offers equal opportunity. Then, inequality must be
To answer
the result of natural talent.)
American
Left critics of the
status
quo, Herrnstein and Murray and their ideological allies offer a robust state-
ment
that
American equality of opportunity
equality of result
Linked
is
each person's "natural"
people can win or lose a
all,
what
fair race
from
the "inner person" standing apart
A
gifts.
American ideology
parents gave them. But
munity.
and thus American
fair race, then,
This
now
see that
treats the "real" individual as
social context,
gave them. This
American
from family or com-
its
makes
virtues,
it
structure (the "ladder of success"
even
we have
6
it
is
because of
who
icans tend not to explain
in
and
talent are
This individual-
referred to) because
that person is or of
outcomes
can only
to discuss matters of social
usually interpret events as the result of individual will. If or poor,
not for
traits.
beliefs about inequality
difficult
traits,
truly unique person
rooted yet more deeply in American individualism. ism, despite
fair
because of the training their
rewards people for their unique
their parents or friends
can
in-
not logically neces-
is
be the "natural" person, the person composed of in-born
We
what a
to these beliefs about inequality is the notion that
contest reveals sary; after
exists
is fair.
Americans
someone
what he or she
did.
is
rich
Amer-
terms of the circumstances people
face.
One
effort
we have made
does matter, individual
in this
traits
book
is
to explain
notwithstanding.
Move
how
social structure
a child
from a chaotic
and impoverished school to an ordered and affluent one, and more often
become "smarter." Move a job-seeker from a region with a 10 percent unemployment rate to one with a 3 percent rate and more often than not that same job-seeker will than not that
same
child will learn more; he or she will
land a better-paying position.
Move
a family headed by
working but poor
parents from a society with minimal family support to one with family
allowances, universal medical care, and other assistance, and more often than not the children in that school, and contribute nizes the differences
more
among
same family
will be healthier,
as adults. In the big picture,
historical eras
do
one
that recog-
and societies, the greatest
ences are from context to context, rather than person to person.
207
better in
differ-
CHAPTER
An American These points can be icans.
was born
in a steel mill.
laid off, so
1915 in Trenton,
in
The next
he returned
a job in a foundry
New
Jersey. His father his
— not
year, with the onset of the Depression, he
to school.
John graduated
any job
1936 John married;
projects. In
and John,
was
Sr.,
in
1934 and found
would seem
the kind of job a high school diploma in the
Depression was a good one.
John's cohort were lucky to land jobs in
in
Amer-
stories are true, repre-
mother had worked as a laundress before
to deserve in those days, but
Some
But the
1929, at age fourteen, John dropped out of school and started
birth. In
was
Sr.,
a truck driver and his
working
are fictional.
and instructive. 7
John Smith,
was
Story: The Smiths
illustrated with the story of three generations of
The names and kinship
sentative,
9
laid off
in
New
Deal public works
1938 his oldest son, John,
Jr.,
was born
from the foundry. Times were no doubt hard and
much help could be expected from John's parents, with their low income and need to hoard for their later years. After twenty-one months of not
unemployment and
as the
a job in a steel mill.
European war unfolded
Work was
in late 1939,
John found
steady through and beyond World
War
II.
(John was too old for military service and was a father.) With growing prosperity,
John was able
to
buy a home. In 1985 John
worked on-and-off
have two more children
after the
war and
to
His and his wife's pensions (she had
retired.
for state government) and their social security checks
provided an annual income of $15,000 in 1990 (about twice the poverty line for a couple).
John,
Jr.,
born in 1938, also grew up
in Trenton.
Unlike his dad, he
continued his education straight through and attended Rutgers University for
two
from 1956
years,
to 1958. His parents paid the tuition of
$500
a
year and he worked for his living expenses. After his sophomore year John,
Jr.,
took a management-trainee job with a firm in Chicago. The next
year he married. In the 1960s the Smiths
where John
III
ployed, John,
moved
to
Richmond,
Jr.,
moved up the ladder through three firms last move to Atlanta in 1990. In that
agement, making his
owned a new in
1995
Virginia,
("Johnny") was born in 1966. Having never been unemto
upper man-
year, John,
Jr.,
house in the suburbs and earned $48,000 a year (over $56,000
dollars).
His wife earned an additional $17,000 as a secretary.
Despite widespread corporate downsizing, John, two, he had a good future in his
new
Jr., felt
that, at
age
fifty-
company. And, given that his elderly
parents were financially secure, he could look forward to helping out his children.
208
—
CONFRONTING INEQUALITY Johnny, born in 1966, grew up
Duke
Although
his father
computer
out to be five
Richmond suburbs and attended
in the
University, earning a bachelor's degree in political science in 1988.
had paid for $32,000 of
graduated owing an additional $32,000 a
AMERICA
IN
firm, trading
still
Carolina, with a
hobby
He landed
computers
in
a
had been married
had been well on
He was
only a sales rep.
his
way
hobby
that turned
for four years and,
to corporate success
living in central Charlotte,
young woman. They had been
were unmarried,
—
a sales job with
his college major. In 1991, at age twenty-
his father already
less education,
Johnny was
years,
his
more valuable than
—an age when
even with
on
Johnny
his college education,
in loans.
North
living together for a
few
and had no expectation of buying a house
childless,
anytime soon. Johnny earned about $36,000 a year and his companion,
working as a substitute teacher, made only $4,000
that year.
Johnny was promotion
not optimistic about his financial future, putting his prospects of at
50:50.
These three men their genetic ferent,
reers
—grandfather,
endowments. Even
greatly.
They
— varied hardly
at all in
their educational credentials, although dif-
were roughly representative of
—humble achievement,
and son
father,
their generations. Yet their three ca-
solid success,
and anxious toehold
differed because the times, the social contexts, in
—
differed
which they
lived differed.
Consider
how even more
dissimilar the stories
would be had we focused
on three generations ol women. As sharply as conditions have changed for men, they have changed much more for women. The typical 1920s stopped working to marry and raise a large family; the 1990s
more often delays marriage
in
woman of the the woman of
order to pursue a career.
The
first
might have hoped for a temporary job as a teacher or nurse; the second
The woman of seventy years ago could
aspires to a well-paid profession. feel financially
secure supported by a "breadwinner" husband; the
woman
of today must consider the real possibility of divorce and single parenting.
Once an average woman might expect
to
woman
on the aid of her children; today's
spend her widowhood depending can look forward to a financially
secure retirement. These have been radical shifts in
life,
not because of
genetic changes, but because of changing contexts.
The differences
in
contexts are not accidental nor totally a result of
forces outside our society.
To be
sure, such forces
world competition, war, for example
much
is
munity. John,
—do
— technological change,
partly determine contexts. But
subject to our control, to our political choices as a national
We
Sr.'s,
can see illustrations adulthood and John,
in the three-generation story.
Jr.'s,
adulthood,
governmental institutions had been created
209
to
all sorts
com-
Between
of financial and
reduce the ferocity of the
CHAPTER
9
business cycle, so that the younger John did need not to pass through a depression. Also, programs were established that relieved John,
weight that had burdened John,
Sr.'s,
Jr.,
of a
generation, dependent parents. Social
security and Medicare directly helped the older generation but also pro-
vided material and psychological freedom for the younger generation.
Other programs
—expansion of higher —aided
homeownership, and so on the
same
education, the
place, others in national priorities,
some
to
why
which they
Implications:
the
Different scientific paradigms
comprehensible (see chapter
is
is
why some
make
Policy Questions
We
Ask?
different questions relevant or even
So, too, with theoretical or ideological
2).
It
assumes
that the critical trait, intel-
—then people can be ranked on one dimension;
—then we can
schools and jobs; that
it
is
—then one can explain family it
is
have pointed
out, for
fails,
it
continuity,
unchangeable
then existing inequality must be accepted and efforts to change
of these elements
that
efficiently sort people into their appropriate
genetic
legitimate inheritance, and predict talent; and that
If any
people get
consider the implications of the "natural inequality" para-
singular
measurable
generation.
gap between the two may be wide or
What
digm, exemplified by The Bell Curve. ligence,
Ill's
live.
Should
First,
generation. At
world market-
be found inside individuals but instead outside of them,
in the society within
paradigms.
Jr.'s,
have frustrated John
In short, the truly powerful forces in determining
ahead and others do not and
Bill, assistance for
in the
the ascent of John,
time, contractions of various kinds,
narrow are not
GI
so do the policy implications.
it
—
are futile.
As many
critics
example, even singular, measurable, and genetic
traits
(such as hair color, myopia, height, and weight) are changeable and are
changed If
all
the time; only unchangeable traits can justify nonintervention.
one believes
that inequality
is
determined by an innate
trait
a "free" market, then the following sorts of policy questions
How
operating in
make
sense:
we best assign children to their appropriate slots? What tests should we use? At what age? How can we reconcile untalented children and young adults to their fates? What can we do to keep their aspirations in line with their limits, so that they are content? Does too much education lead them to frustration? How do we reorganize democracy to take natural inequality into account? Should voting be restricted to those who test well or succeed in life? can
210
CONFRONTING INEQUALITY
How
do we reduce the
birth rates of the
ducement enough, or For the dull
who
will
are born,
it
IN
poor
—
AMERICA that
is,
the "dull"? Is in-
be necessary to compel them?
what do we do
them
to sustain
in a
humane
fashion?
Are moral If so,
same way?
qualities as well as cognitive ones "natural" in the
how do we deal
with people born lacking moral instincts? (Several
early psychometricians claimed or speculated, for example, that
Jews
inherited a genetic tendency for dishonesty.)
This agenda
is
not fanciful.
tricians, educators,
and
was
It
statisticians
explicitly taken
up by the psychome-
whose work forms
the foundation of
The Bell Curve. For example, Harvard psychologist William McDougall 1925 proposed a be
in class
A and have
educated would be
in
system of citizenship: Intelligent people would
stratified
full rights;
in class
C
people
who
were poorly
tested poorly or
and have no vote; and class
B would
be
largely a temporary status for children before they are selected into class
or C.
8
Colgate University president George Barton wrote
hoped intelligence India," but
testing
would produce
in
A
1922 that he
a "caste system as rigid as that of
one based on the "rational" system of psychology. 9
Herrnstein and Murray's ideas are distant from these draconian extensions of the "natural inequality" perspective, but they are rooted in the
—
same paradigm.
Parts of a similar agenda are hinted at in
for instance, the
concern with differential breeding, with helping everyone
The Bell Curve
find the place in society "appropriate" to his or her intelligence, with recon-
ciling people to their limitations,
10
and the suggestion of a guaranteed an-
nual income for the "dull" losers. This agenda need not be undemocratic, racist,
or inhumane. But
it
follows logically from the "natural inequality"
paradigm.
The
"natural inequality" paradigm, whatever
does not stand up to the evidence, as to
this
its
political implications,
book has demonstrated. Turning
our "social construction" paradigm, which does
fit
the evidence, a differ-
ent set of policy questions arise:
How much equality of opportunity do we want? We, as a nation, have set up and can revise the rules for the "race" to success. To what extent do we want those rules
to provide every child with
an equal
start,
so that where
they end up reflects only their varied talents and not the advantages or
disadvantages of their social backgrounds?
The hasty answer might be "as much as possible," but stop to consider what that would really mean: In a society with full equality of opportunity, each child would have the same material advantages, the same challenging school curricula, the same quality neighborhoods, and so forth. That goal
211
CHAPTER
9
American
directly contradicts another important
value: that parents
work
hard, scrimp, and save precisely in order to give their children an advan-
tage
—
to provide
attend, the
them with
homes
the nicest
to live in, the best schools to
most supportive neighborhoods, extras
so on. This value
is
oppose the idea of big inheritance your children a head
is
to give
is
being unable to do
so.
So
a real one:
the question
is
family value and the value
How much
like
we
taxes.
start;
One of the just rewards
What
we want between
trade-off do
do we want? Once
should the rewards be for the winners and losers? As is
is
counted.
exists,
Is that
because
the race
from market earnings
acceptable?
this level
this
Many
believe
of inequality
is
it is,
to
what
run,
is
we have documented,
extreme among the industrial nations
inequality once everything,
of success
conversely, one of the costs of failure
place on equality of opportunity?
equality of result
the United States
summer camp, and
one reason Americans, even those with low incomes,
in its
degree of
government programs,
as long as a "safety net"
equitable and because perhaps
it
stimulates greater productive energy. Others find American inequality of
and unnecessary. Certainly, evidence shows
result intolerably high
communities and societies with high degrees of inequality tend bled and torn ones. 11 Critics of inequality also argue that
and depresses
initiative,
thereby reducing productivity.
it
to
that
be trou-
stunts ability
Some
in this
camp
argue that more equality would actually increase the whole "pie" for every-
one by bonding people
at the
bottom
into the
wider society, reducing
ill
health and destructive behavior and widening the pool of talent. There
evidence that equality Realistically,
may
well spur economic growth (see chapter
more equality of
result
would require rewriting
competition and restructuring the system of rewards. that?
—and here "we" includes
book,
we who come from
the authors and
is
5).
the rules of
Do we want
to
most of the readers of
do
this
the comfortable classes in America.
These are the most general questions
that derive
sented in this book (they are questions that
from the paradigm pre-
make no sense in the older come onto the agenda. For
paradigm). But other, more specific issues also
example:
Which it?
talents should
Human
skills,
we
nurture
and how much should we spend doing
whatever genetic component they have, are developed.
Even children born with
serious neurological impediments can today, with
timely help, develop into successful adults. So can children born with seri-
ous social impediments. invest in
—
It is
a matter of identifying the skills
analytical logic, creativity,
and deciding
how and how much
to invest.
212
we wish
empathy, leadership, and so on
to
CONFRONTING INEQUALITY
IN
AMI RICA
History shows us that public investment, far from undercutting tive,
frastructure
—
—
roads, technology, agricultural development, and the like
much of
undergirded ple
initia-
has unleashed and stimulated American energies. Investments in in-
the
American economic boom. Investments
in
peo-
quality schools, public health, higher education, enriched jobs, and
the like
—have not only enriched
American community have not
horizons. These joint endeavors of the
sapped individual
How bedded.
sparked and sustained
vitality but
human
us materially but also expanded
it.
The market
is "socially emwe From standardized weights and measures, to the limited corpowork regulations, to tax deductions, governmentally established
ration, to
the market?
regulate
shall " I2
rules shape the market. So,
which ones do we change //we want
to alter the
Some answer "none" and contend minimum wages, health and safety (with little evidence) that certain rules already equalize too much and intercodes, required benefits, and so on inequality that arises in the market?
—
Others answer "a
fere with efficient production.
for
—
lot."
The
example, increasing the bargaining power of labor
raising the
minimum wage,
In
Expanding higher education,
sum, the paradigm
equality
is
paradigm ever
way
is
How
is
to structure the
would
market?
from our
"historical
agenda of policy issues for Americans. Because
may answer
re-
presented, one that recognizes that in-
socially constructed
factually accurate, that
readers
for example,
in turn
home mortgage deduction would
do we want
we have
not "natural" but
acts," leads to a critical
wages, or
shape the market, which
will
duce earnings inequality; capping the reduce wealth inequality.
might urge,
or limiting executive income, or making taxes
more progressive. Whatever we decide will shape inequality.
latter
in setting
agenda
is
the questions
a real and urgent one.
on
this
this
Which-
agenda, they must rec-
ognize that Americans are designing inequality; even to avoid a decision to
is
decide for the status quo. Either way, Americans must accept responsi-
bility for the
design of inequality instead of blaming nature.
Equality and Opportunity
When
posed as alternatives, Americans endorse equality of opportunity
and reject equality of
result.
called full opportunity, that ideal. Jennifer
13
But there
more
is
a third option, which might be
closely captures the
American moral
Hochschild of Princeton University has identified four com-
plementary tenets of the American Dream: Everyone can participate;
everyone has a reasonable chance for success
213
if
he or she plays by the rules;
CHAPTER success and failure
9
under one's control; and success implies
is
we have
equality on the scale
denies
ity truly to participate,
many
a reasonable chance to succeed, and
handicaps people according to accidents of clearly a
badge of
birth, so that
virtue. Greater opportunity
closer to fulfilling the
virtue. In-
today robs some people of the opportun-
success
would bring
American dream. Securing
this
is
not so
country
this
kind of opportunity
depends, in turn, on our social choices. Policies that simply
provide
promote equal opportunity may not be
One can imagine
opportunity.
full
sufficient to
a science fiction society of im-
poverished people where a few are chosen by lottery to
live in luxury.
This
equality of opportunity neither rewards nor encourages development of talent.
The kind of opportunity Americans want for themselves and their would require not only equalizing access to resources such as
children
schools, but also public investments to
make
sure that the schools were
excellent ones.
We
do not have a society of equal opportunity. The evidence
certainly
shows how
background (especially parents' income)
inequalities in family
and of the broader social context (including the quality of schools) despite the difficulty of measuring such contextual factors fully and accurately
— shape outcomes, so opportunities
are not equally distributed. But
we could equalize 1960s when we expanded higher
research also shows that
opportunity more, as America
did in the
education. In a stagnant econ-
omy, however, or
in an
economy with growing inequalities in outcomes, may mean insecurity or decline for the vast
simple equality of opportunity
majority. In such a case, the difference
between an unfair race
children of the privileged are advantaged versus a fair one that
much
of a difference
if
in
may
which the not
make
neither the children of the privileged nor of the
unprivileged can find secure and rewarding positions. Policies that
would promote
full
opportunity require extensive public
investment. First and most clearly, a society that promoted opportunity
would invest
in
most
fully,
primary and higher education, job — —whatever would help people develop
people
ing, school lunches
in health,
so that more, not fewer, people could participate in a
economy. Second, a commitment in
to full opportunity
programs
home
skills. It
loans for
ple of all social
mean
would mean
that built the
modern
would mean investing
job development, so that there would be rewards for people
proved their
train-
their talents
who
im-
extending, not contracting, the kinds of
broad American middle class
— such
as low-interest
young families and low-tuition public universities for peobackgrounds to go to college. And it would almost certainly
a different
way of providing
health care, in a system that encouraged
214
CONFRONTING INEQUALITY employers
to hire
AMERICA
IN
more, rather than fewer, workers, and that increased the
chances that both rich and poor children had a good
most central
start in life.
Third, and
our argument here, more opportunity necessarily requires a
to
the one we have now. If we have a social order in people cannot make an adequate living even when working many which
society
more equal than
hard or cannot find jobs that provide security or cannot find work then they have
reward for their
efforts.
opportunity, too, since off in
at all,
incentive to develop their abilities and no reasonable
little
Such
a social order inevitably decreases equality of
burdens children as they
strains marriages,
it
start
and disrupts the communities upon which children depend. So,
life,
greater inequality of
outcomes necessarily decreases both opportunity and
equality of opportunity.
Americans
resist
what
is
called "equality of result" because the idea of
giving everyone equal benefits no matter what their contributions violates
many
our sense of fairness. But
energy and tion,
initiative. If
egalitarian policies stimulate and reward
American law encouraged higher
more jobs would pay
a decent wage. If
rates of unioniza-
American government once
again regulated corporate wheeling and dealing, more jobs might be pre-
we provided more
served. If
social infrastructure, such as affordable child
care and longer school years,
same
skills at the
time.
Even
care for poor children or
we would
support work and raise childrens'
policies of direct redistribution, like health
income supplements
give children from poor backgrounds a
"Equality of result," in an absolute sense,
man.
No one
is
for poor
working families,
more equal chance is
in life.
thus something of a straw
advocating, nor could one realistically imagine, a society in
which people benefit equally regardless of thriving, wealthy,
modern
their contributions.
But other
societies stimulate initiative and reward ability
with systems of inequality more generous than our own. The real concerns
ought to be
do
how much
inequality
to exacerbate that inequality.
reasonable and what our social policies
is
At what point does inequality become a
powerful drag on our people, giving the poor no real hope of improving their lot,
keeping the middle striving ever more frenziedly just
place, and rewarding the rich ever
more
lavishly,
sometimes
to stay in
just for hold-
ing assets favored by the tax code?
Naysayers
will
warn
that
such moves would undermine our nation's
we pay
Such As we pointed out in chapter 1, they are arguments Americans heard more and more as the antigovernment economic policies of the 1980s failed to help the middle class. They were meant to assuage listeners' anxieties about
economic
health; that inequality
arguments are
in
is
the price
for our wealth.
tune with the claims that inequality
215
is
"natural."
CHAPTER growing
9
But these warnings are
inequality.
We
false.
need not
The contrary seems
gross inequality in order to have growth.
tolerate
true;
such
inequality probably slows our growth (see chapter 5).
We
have emphasized repeatedly
Americans who
that
national politics design the inequality with which
determine inequality
in the
ways we
we
participate in our
live.
For example, we
distribute tax burdens, subsidize
homeownership, regulate business and labor unions, and finance health care. But the inequality we have does not just affect the overall well-being of our citizens.
by whether it
it
also shapes
It
how much
opportunity our society provides,
encourages individuals to develop their
rewards them
abilities
and whether
fairly for their efforts.
Invocation In closing,
we
note that any debate over inequality can rest only in part on
the weight of the social science evidence.
commitments. The
explicit
over two hundred years
—
It
must also
rest
on our moral
moral commitments Americans have made for
them
despite our frequent failures to live up to
include a political dedication to equality. Franklin Delano Roosevelt said in
1937 that the
"test of
our progress
abundance of those who have much; those
who have
too
junctions to charity.
little."
And
God tells
is
not whether
it is
we add more
to the
whether we provide enough for
our commitments include the biblical in-
the Israelites, "If there be
man, one of thy brethren, within any of thy gates
.
.
.
among you
a needy
thou shalt not harden
thy heart, nor shut thy hand from thy needy brother" (Deut. 15:7).
was hungry and you gave me food, I was was a stranger and you welcomed me.
Jesus spoke: "For
gave
me
drink,
I
thirsty
I
.
you, as you did to one of the least of these (Matt. 25:35, 40).
216
my
brethren,
.
.
Truly,
you did
And
and you
it
I
say to to
me"
* APPENDIX
Summary
*
1
of The Bell Curve
H,
.errnstein and Murray begin,
in the preface,
by raising a concern
that
Americans are becoming increasingly and more widely divided between a highly educated and well-paid elite at one extreme and an impoverished
and problem-beset underclass
How
at the other.
In the introduction, Herrnstein
can
we
understand that?
and Murray define "intelligence" concepand applying relationships drawn
tually as "a general capacity for inferring
from experience" and "a person's capacity for complex mental work" but operationally as a person's score on a statistically determined set
(p. 4),
of
test questions, the
testing against
IQ
tests accurately
how
colloquially
famed "IQ"
many
its
critics.
and
test
Most
its
surrogates.
measure variations
Part
I
agree that
"smart" they are; that people's IQ scores are stable over
80 percent "heritable" is
now
in people's "cognitive ability"
their lifetimes; that these tests are not biased culturally;
people in IQ
They defend IQ
scholars, they claim,
due
—
that
that
is,
to variation
40
to
and
that
American
draws high-IQ people into
II,
40
to
in their genes.
outlines the threat of an emerging "cognitive elite."
economic echelons. Since World War
is
80 percent of the variation among
among them
ciety today with increasing efficiency
IQ
its
so-
top
colleges have accepted and gradu-
more young Americans and have done so more on achievement test scores than of family connections. The elite ated vastly
the basis of
colleges, in
have increasingly "creamed" the smartest students. Similarly,
particular,
occupations that recruit high-IQ people have expanded and perhaps (the case here dates.
is
more
On the job,
speculative)
become more
workers' IQ scores predict
selective of high-IQ candi-
how
well they perform and do
so better than do seemingly job-specific tests or workers' other characteristics.
The market, of
course, rewards this superior performance by the
brightest with the highest incomes.
sun"
(p. 92): stratification
The
result
is
"something new under the
of society by cognitive
the formation of a high-IQ elite.
The
ability,
and
in particular,
stratification process is accelerating
because the emerging economic system rewards intelligence above
The economic separation of
the smartest
constantly reassure readers that they are the rest)
is
augmented by
from the
among
rest
of us (the authors
among More and and smart men and
the smartest, not
their increasing physical separation.
more, the smartest work and
live apart
217
from the
rest,
all.
APPENDIX
women
smart
are increasingly marrying
assumption that IQ
is
largely genetic
we
smart babies. In the end,
turally distinct elite "taking
Part
1
one another, which
—makes
the
are seeing the formation of a smart, rich, cul-
on some characteristics of a caste"
of The Bell Curve lays the foundation for most of
II
—given
even more distinctively
for
its
(p. 113).
substantive
arguments. In over 230 pages of text, notes, and appendices, the authors try to
make
they
.
some of lems"
people
become problematic
likely to
"barely
how
the statistical case that individuals' intelligence determines
live. In particular,
.
.
who
score low on
citizens.
By
looking
IQ
at
tests are especially
IQ's effects, until
the mystery that has surrounded the nation's
(p.
now
considered" by social scientists, they will be "clearing away
118).
To make
massive survey and a
their case, Herrnstein
most serious prob-
and Murray introduce a
statistical tool.
The survey Herrnstein and Murray analyzed is the National Longitudinal Survey of Youth (NLSY). Researchers began in 1979 with a sample of over 12,500 youths aged fourteen to twenty-two and have followed them since, asking
them
a
took a battery of
wide variety of questions. In 1980 most of the sample
tests,
including the
(AFQT). Herrnstein and Murray use Curve includes analyses of the
the
Armed
AFQT
Forces Qualifying Test as their
NLSY respondents
IQ
test.
The Bell
through the 1990
inter-
views. (In this part of the book, the authors look only at white, non-Latino respondents.)
The
statistical tool is "logistic multivariate regression analysis."
quick introduction to regression analysis, see appendix
Murray use
it
how much
for estimating
2.)
(For a
Herrnstein and
the variation in white
NLSY
re-
AFQT scores) "explains" variations How much, they ask, does AFQT score an NLSY respondent in the next decade
spondents' IQs (as measured by their
among them in
in certain
outcomes.
1980 explain the chances
became
that
poor, a high school dropout, unemployed, unmarried, an
unwed
mother, on welfare, a neglectful mother, or a criminal? To underline the
importance of the ation in an
AFQT
outcome
it
score, they regularly
explains with
how much
compare how much
variation
is
vari-
explained by the
socioeconomic status of the respondents' parents. Crudely, parents' class represents "nurture" and the
deem
it
AFQT score represents "nature." (Where they AFQT score, they add other
appropriate to testing the effects of the
variables to the analysis.)
From
chapter 5 through chapter 12, Herrnstein and Murray present the
results of statistical
AFQT
analyses claiming that white
scores better identify which of them ended up
tions than does their parents'
socioeconomic
218
NLSY
respondents'
in problematic situa-
status. In particular,
it is
the
SUMMARY OF THE BELL CURVE respondents
label "very dull"
By
extreme low end of the
at the
AFQT bell curve who are at great
For example, 30 percent of the whites
risk.
—
the
whom
bottom 5 percent on the
were poor. Using
logistic multiple regression analy-
appendix 2 for a review of regression analysis) to look
sis (see
and parents' class
at the
same time suggests
position
poor
in
parents of average
and about 10 percent for children of the very lowest class
class position
who had
—and
AFQT
average
AFQT better
sion: the
who had
AFQT
at the
that the risk of being
1989 was about 25 percent for the "very dull"
position
1989.
24 percent of the lowest 5 percent on the scale measuring
contrast,
parents' class position
work
Herrnstein and Murray
AFQT— were poor in
scores. Herrnstein and Murray's conclu-
accounts for adult poverty than does parents' social
problem
similarly with the other
disability,
being divorced, and being
such as having a
statuses,
in jail.
Herrnstein and Murray wrap this section of The Bell Curve by giving
each respondent to the
NLSY
scored as "middle class" in the labor force
a
"Middle Class Values"
first
A man
score.
he had graduated from high school, had been
throughout 1989, was never interviewed
married to his
still
if
in jail,
and was
who met
wife. ("Never-married people
all
the
other conditions except the marital one were excluded from the analysis" 263]. In other words, unmarried people with a problem were kept in the
[p.
analysis; those with
no problems were ignored.)
A woman
needed
to
have
graduated from high school, have never had a baby out of wedlock, have never been interviewed
AFQT, class
in jail,
and be
still
married to her
by
this
support the claim.
cant independent role for
in
who had
IQ remains"
probability of "middle classness"
graduated high school, "a in
AFQT