-
U.S. Elementary and Secondary Schools: Equalizing Opportunityor
Replicating the Status Quo?
Cecilia Elena Rouse, Lisa Barrow
The Future of Children, Volume 16, Number 2, Fall 2006, pp.
99-123(Article)
Published by Princeton UniversityDOI: 10.1353/foc.2006.0018
For additional information about this article
Access provided by University of California, San Diego (6 Dec
2013 13:19 GMT)
http://muse.jhu.edu/journals/foc/summary/v016/16.2rouse.html
http://muse.jhu.edu/journals/foc/summary/v016/16.2rouse.html
-
U.S. Elementary and Secondary Schools:Equalizing Opportunity or
Replicating the Status Quo?
Cecilia Elena Rouse and Lisa Barrow
SummaryAlthough education pays off handsomely in the United
States, children from low-income familiesattain less education than
children from more advantaged families. In this article, Cecilia
ElenaRouse and Lisa Barrow investigate why family background is so
strongly linked to education.
The authors show that family socioeconomic status affects such
educational outcomes as testscores, grade retention, and high
school graduation, and that educational attainment strongly
af-fects adult earnings. They then go on to ask why children from
more advantaged families getmore or better schooling than those
from less advantaged families. For low-income students,greater
psychological costs, the cost of forgone income (continuing in
school instead of getting ajob), and borrowing costs all help to
explain why these students attain less education than
moreprivileged children. And these income-related differences in
costs may themselves be driven bydifferences in access to quality
schools. As a result, U.S. public schools tend to reinforce
thetransmission of low socioeconomic status from parents to
children.
Policy interventions aimed at improving school quality for
children from disadvantaged familiesthus have the potential to
increase social mobility. Despite the considerable political
attentionpaid to increasing school accountability, as in the No
Child Left Behind Act, along with charterschools and vouchers to
help the children of poor families attend private school, to date
thebest evidence suggests that such programs will improve student
achievement only modestly.
Based on the best research evidence, smaller class sizes seem to
be one promising avenue for im-proving school quality for
disadvantaged students. High teacher quality is also likely to be
impor-tant. However, advantaged families, by spending more money on
education outside school, canand will partly undo policy attempts
to equalize school quality for poor and nonpoor children.
V O L . 1 6 / N O. 2 / FA L L 2 0 0 6 99
www.futureofchildren.org
Cecilia Elena Rouse is director of the Education Research
Section and professor of economics and public affairs at Princeton
University. LisaBarrow is a senior economist at the Federal Reserve
Bank of Chicago. The authors thank David Card, Gordon Dahl, and
Lisa Markman forinsightful conversations; participants at the
authors’ conference, particularly Douglas Massey and the issue
editors, for helpful comments;and Eleanor Choi for outstanding
research assistance. The views expressed in this paper are those of
the authors and do not representthose of the Federal Reserve Bank
of Chicago or the Federal Reserve System. All errors in fact or
interpretation are ours.
06 5563 Rouse-Barrow.qxp 8/6/2006 4:56 PM Page 99
-
In 1967 Martin Luther King Jr. wrote,“The job of the school is
to teach sowell that family background is nolonger an issue.” As
King’s remarksuggests, Americans have long hadhigh expectations for
their educational sys-tem. One reason they demand so much fromtheir
schools is that education is closelylinked both to income and to
occupation.Better educated individuals earn more andwork in more
prestigious occupations. In-deed, because education affects both
incomeand occupation, it is traditionally thought tobe important in
determining an adult’s so-cioeconomic status.
Figure 1 shows the relationship betweenyears of completed
schooling and annualearnings, using data from the March 2003and
2004 Current Population Survey (CPS).On average, high school
graduates withtwelve years of schooling earn nearly $26,000a year,
as against about $19,000 for highschool dropouts with only eleven
years ofschooling. Completing a high school degreeis also a
prerequisite for college admission,and the value of a college
degree, particularlya four-year college degree, has
increasedsharply over the past twenty-five years. In1979, adults
with a bachelor’s degree orhigher earned roughly 75 percent more
eachyear than high school graduates. By 2003,their yearly earnings
were more than double(2.3 times) those of high school
graduates.1
Even if an individual does not intend to go onto college, a high
school diploma is a mini-mum education requirement for many
jobs.Although direct information on occupationalrequirements is not
available, high schoolgraduates in the 2004 CPS Outgoing Rota-tion
Group data are more likely than highschool dropouts to work in the
highest-wageoccupation groups—management, architec-
ture and engineering, computers, and thelaw. For example, 7.1
percent of adults agedtwenty-five to sixty-five who have
completedhigh school, but no college, work in one ofthose
occupation groups, as against only 2.6percent of adults who dropped
out of highschool. Conversely, 26 percent of high schooldropouts
work in the lowest-average-wageoccupational groups—food preparation
andservice; farming, fishing, and forestry; andbuilding and grounds
cleaning and mainte-nance—compared with 11.5 percent of highschool
graduates.2
Education is thus an important driver of up-ward mobility in the
United States. But as wedocument below, America’s schools fail to
ful-fill King’s vision. A U.S. child’s educational at-tainment is
strongly linked to his or her fam-ily background, and children of
parents oflow socioeconomic status are likely as adultsto have the
same socioeconomic status astheir parents. In this article we
investigatewhy family background is so important in de-termining a
child’s educational attainment, aswell as how the nation’s K–12
educationalsystem perpetuates this pattern.
How Family Background AffectsEducational AttainmentTheoretically
if everyone, rich or poor, facesthe same cost and reaps the same
benefitfrom additional schooling, educational attain-ment should
not differ by family background.In the real world, however, years
of schoolingcompleted, and educational achievementmore generally,
vary widely by family back-ground. To illustrate we turn to data
from theNational Education Longitudinal Study(NELS) of 1988, which
followed more than20,000 eighth graders from 1988 through1994 (for
many, their sophomore year of col-lege). This survey has rich
information bothabout the educational experiences of the stu-
C e c i l i a E l e n a R o u s e a n d L i s a B a r r o w
100 T H E F U T U R E O F C H I L D R E N
06 5563 Rouse-Barrow.qxp 8/6/2006 4:56 PM Page 100
-
dents and about their parents and schools.Figure 2 shows how
students’ educationalachievements vary by family background. Wehave
divided the students’ families into foureven groups (quartiles)
based on an index ofsocioeconomic status. Those in the
lowestquartile are the most disadvantaged, whilethose in the
highest quartile are the most ad-vantaged. The average family
income in thelowest quartile is about $27,000 (in 2004 dol-lars),
with an average family size of 4.6. In thesecond quartile the
average family income isabout $48,000 (average family size of 4.4);
inthe third quartile it is about $69,000 (averagefamily size of
4.3); and in the fourth quartile itis nearly $110,000 (average
family size of 4.4).
As the figure shows, children from families inthe highest
quartile have higher average testscores and are more likely never
to have beenheld back a grade than children from familiesin the
lowest quartile. Children from familiesin the top quartile are also
more likely never todrop out of high school, and therefore muchmore
likely to have a high school diploma sixyears after they entered
the eighth grade.
Although these patterns are striking, it is notclear they
reflect the causal effect of family
background on a child’s educational achieve-ment. Inherited
genetic ability confounds at-tempts to study the link between
family back-ground and educational achievementbecause to the extent
that ability or intelli-gence is heritable, genetics helps
determinewhether children are successful in school.For example,
evidence suggests that peoplewith lower observed ability earn lower
wagesthan those with higher ability.3 Thus, lessable people will
have lower socioeconomicstatus than more able people. Further,
moreable people likely find it less costly to getmore schooling, in
the sense that it is easierfor them to master the knowledge
required ateach new step of school than it is for an indi-vidual of
similar background but with lowerability. If it is also true that
ability is geneti-cally determined, then less able parentswhose
socioeconomic status is low will alsohave less able children who
will get lessschooling than the children of more able par-ents
whose socioeconomic status is high. Inthis example, the
heritability of ability com-bined with the link between ability and
edu-cational achievement means that low innateability explains both
the parents’ low socio-economic status and the children’s lesser
ed-ucational achievement.
U . S . E l e m e n t a r y a n d S e c o n d a r y S c h o o l
s : E q u a l i z i n g O p p o r t u n i t y o r R e p l i c a t i
n g t h e S t a t u s Q u o ?
V O L . 1 6 / N O. 2 / FA L L 2 0 0 6 101
Figure 1. Average Annual Earnings, by Years of Completed
Schooling
Source: Authors’ calculations from the Current Population
Survey, March 2003 and 2004.
80
70
60
50
40
30
20
10
Average annual earnings (thousands of 2004 dollars)
Years of schooling
1817161514131211109876543210
06 5563 Rouse-Barrow.qxp 8/6/2006 4:56 PM Page 101
-
To disentangle the effects of genetic makeup(which is not
malleable) and family back-ground (which is likely more malleable)
oneducational attainment, a researcher wouldideally conduct an
experiment. The experi-ment would begin with the random assign-ment
of one group of children to disadvan-taged families and another
group to moreadvantaged families—without regard to thechildren’s
“innate” ability. Because assignmentto families would be random,
there would beno link between the genetic ability of the chil-dren
and that of the parents. On average theonly difference between the
two groups ofchildren would be their family background.Years later
the researcher could compare theeducational attainment of these
children.With a large enough sample, differences be-tween the two
groups would provide a credi-ble estimate of how much family
backgroundcausally affects educational attainment.
In this experiment what the researcher wantsto control is the
wealth (or socioeconomicstatus more generally) of the family in
whichthe child was raised. The researcher does not
attempt to control which schools the childrenattended, whether
the children had access togood medical care, their families’
parentingpractices, or other aspects of their lives thatundoubtedly
affect their educational attain-ment. Why not? Because the
researcher isnot interested in the effect of randomly as-signing
students to families of different back-grounds, assuming that the
families do every-thing else the same.
Another way to see this is to consider possi-ble policy
implications. Suppose a new publicpolicy aiming to increase the
educational at-tainment of children were to give $10,000 toeach
family whose income fell below, say, thenational poverty line. The
policy’s intentwould not be for parents to put the moneyinto the
bank and not spend it on their chil-dren. Rather, the intent would
be for them touse the money to buy nutritious food, enrolltheir
children in better schools, purchasesupplementary educational
materials, get ac-cess to better medical care, or purchase
othermaterials that would help their children’s ed-ucational
success. That is, the key policy
C e c i l i a E l e n a R o u s e a n d L i s a B a r r o w
102 T H E F U T U R E O F C H I L D R E N
Figure 2. Educational Outcomes, by Family Socioeconomic
Status
Source: Authors’ calculations from the National Education
Longitudinal Study of 1988.
100
90
80
70
60
50
40
30
20
10
0
Percentile or percent
Highest SES quartile
Lowest SESquartile
Percent receivedhigh school diploma,
6 years after 8th grade
Percent reportingnever dropped outof school, 4 years
after 8th grade
Percent never reporting being held back a grade,
8th grade and 2 and 4 years after 8th grade
Standardized test score percentileranking, 4 yearsafter 8th
grade
06 5563 Rouse-Barrow.qxp 8/6/2006 4:56 PM Page 102
-
question is not whether wealth or social ad-vantage affects
educational attainment perse, but whether the behaviors and
resourcesmade possible by that wealth and social ad-vantage affect
educational attainment.
In a study that comes close to the ideal exper-iment just
described, Bruce Sacerdote exam-ines the educational attainment of
childrenadopted from South Korea who were ran-domly assigned to
U.S. adoptive families.4 Be-cause the children are adopted and
randomlyassigned to their families, there should be norelationship
between the mother’s innate abil-ity and the child’s innate
ability; thus any rela-tionship between the mother’s educational
at-tainment and that of the children is causal.Because many of
these families also have bio-logical children, Sacerdote compares
the linkbetween a mother’s schooling and a child’sschooling for
adopted and biological childrenand estimates how much the mothers’
educa-tional attainment determines that of the bio-logical
children. He calculates that only 23percent of schooling
transmitted from motherto child is the direct effect of the
mother’s ed-ucation, suggesting a very large role for genet-ics. In
contrast, he finds that nurture plays amuch larger role than nature
in transmittinghealth habits such as drinking and smoking:these
habits pass along to biological andadopted children at equal rates.
Sacerdote, aneconomist, notes that under very strong as-sumptions
his finding means that 23 percentof educational attainment is
determined byenvironment, implying that up to 77 percentis
determined by nature. Most psychologistswho examine how genetics
affects academicachievement in young children find
smallerestimates, in the range of 30 to 40 percent.5
Some also argue that adoption studies over-state the importance
of genetics becauseadoptive families are not representative
offamilies in the general population.6
U . S . E l e m e n t a r y a n d S e c o n d a r y S c h o o l
s : E q u a l i z i n g O p p o r t u n i t y o r R e p l i c a t i
n g t h e S t a t u s Q u o ?
V O L . 1 6 / N O. 2 / FA L L 2 0 0 6 103
Researchers have used other strategies to es-timate the extent
to which family income de-termines children’s educational
achievement.Again, because they cannot assume that fam-ily income
is unrelated to other factors (suchas inherited ability) that
determine both chil-dren’s socioeconomic status and their
educa-tional attainment, they must look for changesin family income
that are unrelated to family
characteristics such as whether the parentsare highly educated
or have high genetic“ability.” Pamela Morris, Greg Duncan,
andChristopher Rodrigues take advantage ofvariations in family
income caused by experi-mental welfare programs in the United
Statesand Canada during the 1990s to examine howincome affects
children’s achievement.7 Thewelfare programs were all designed to
in-crease work, and several were also designedto increase income,
either through wage sup-plements or by allowing participants to
keepmore of their welfare payments when theywent to work. Because
no direct family orchild services (such as parenting classes
orchild care subsidies) were provided, anychanges in children’s
achievement must beattributable to changes in their parents’
em-
The key policy question is not whether wealth orsocial advantage
affectseducational attainment perse, but whether the behaviorsand
resources made possibleby that wealth and socialadvantage affect
educationalattainment.
06 5563 Rouse-Barrow.qxp 8/6/2006 4:56 PM Page 103
-
ployment, income, and welfare receipt gen-erated by random
assignment to the differentprograms.
Morris, Duncan, and Rodrigues look at howthese differences in
income (all generated byrandom assignment) affect
children’sachievement. They find that a $1,000 in-crease in annual
income (over three to fiveyears) increases achievement by 6 percent
ofa standard deviation for children who are twoto five years old.
However, it has no effect onachievement for older children (six to
nineyears old and ten to fifteen years old). Thecost and benefit of
the increased income forpreschool-aged children compare favorablyto
the cost and benefit of direct educationalinterventions such as
reducing class size. (Inone experiment, Alan Krueger and
DianeWhitmore Schanzenbach find that class-sizereductions costing
$9,200 per pupil forgrades K–3 increased children’s achievementby
13 percent of a standard deviation.)8
Addressing the question of how changes infamily income affect
children’s academic at-tainment in yet another way, Gordon Dahland
Lance Lochner use the fact that increasesover the past twenty years
in the earned in-come tax credit for working families havecaused
increases in family income to examinehow child achievement is
affected.9 Familieswith two children with earned income of,
say,$10,000 in 1993 would have been eligible for atax credit of
$1,511. That same family wouldhave been eligible for a credit of
$2,528 in1994 and $3,110 in 1995. Thus with no changein nominal
earned income, total family in-come would have increased each year.
Did theadded money improve student test scores?
Dahl and Lochner find that it did. A $1,000increase in income
raised math and readingscores by 2 to 4 percent of a standard
devia-
tion—an improvement large enough to closeroughly 3 to 5 percent
of the achievementgap between children in the bottom incomequartile
(average family income of $14,214 in2000 dollars) and those in the
top quartile(average income $81,137).10 Furthermore,when Dahl and
Lochner estimate how in-come affected test scores for various
sub-groups, they find even larger effects for chil-dren from
disadvantaged families, who aremore likely to receive the maximum
increasein income.
Overall, the evidence suggests that parentalsocioeconomic status
has a causal effect onchildren’s educational outcomes. But
thestudies noted cannot identify precisely howincreases in parental
education or incomeimprove children’s educational outcomes.Economic
theory suggests that people stay inschool until the costs of doing
so (direct costsas well as forgone earnings and the psycho-logical
costs of being in school) outweigh thebenefits. Thus, if the
children of advantagedfamilies stay in school longer, it must be
be-cause they receive greater benefits or facelower costs than do
less advantaged children(for example, forgone earnings are less
im-portant to a wealthy family than to a poorfamily). In the next
sections, we investigatewhy the relationship between family
back-ground and educational attainment may be sostrong.
Does the Economic Value ofEducation Differ by
FamilyBackground?We first examine whether education has adifferent
value for people of different socioe-conomic backgrounds. If
children from moreadvantaged families receive larger gains fromeach
additional year of schooling, they willhave a greater incentive to
stay in school. Be-cause research on the economic value of edu-
C e c i l i a E l e n a R o u s e a n d L i s a B a r r o w
104 T H E F U T U R E O F C H I L D R E N
06 5563 Rouse-Barrow.qxp 8/6/2006 4:56 PM Page 104
-
schooling-income connection may mostly re-flect the fact that
more able people commanda premium for their (innate) skills in
thelabor market. Thus empirical estimates of thereturn to schooling
such as the one just notedare too large. In this view, increasing
fundingfor educational programs for the disadvan-taged will have
little or no effect becauseschooling cannot change innate
ability.
Again, researchers have developed severalmethods to isolate the
economic value of ed-ucation in an effort to disentangle these
twohypotheses. To determine definitivelywhether more schooling
raises income, anideal experiment would involve randomly as-signing
one group of students to completehigh school and another group to
drop out,regardless of the students’ innate ability orfamily
background. Years later researcherswould compare how the two groups
fared inthe labor market. On average the only differ-ence between
the two would be whether theyhad graduated from high school.
Differencesin the earnings of the two groups would pro-vide an
estimate of the economic value of ed-ucation—how much completing
high schoolcauses earnings to increase. To determinewhether this
economic value varies by familybackground, the researcher could
simply es-timate the earnings difference for subgroupsof students
based on their family backgroundat the start of the experiment.
Empirical Estimates of the EconomicValue of SchoolingRecognizing
that no such experiment willever be conducted, researchers have
devel-oped two broad approaches to empirical esti-mation of the
economic value of education.The first approach—so-called natural
experi-ments—locates events or policies that mightbe expected to
alter the schooling decisionsof some people, but would not be
expected to
U . S . E l e m e n t a r y a n d S e c o n d a r y S c h o o l
s : E q u a l i z i n g O p p o r t u n i t y o r R e p l i c a t i
n g t h e S t a t u s Q u o ?
V O L . 1 6 / N O. 2 / FA L L 2 0 0 6 105
cation is extensive, while that on the extent towhich that value
varies by family backgroundis more limited, we begin by discussing
theoverall relationship between education andincome.
Estimating the Economic Value ofSchooling Is Not
StraightforwardEconomists conventionally measure the eco-nomic
value of additional schooling (or the“return to schooling”) as the
average percent-age difference in mean earnings for each
ad-ditional year of education.11 Estimates basedon the Current
Population Survey, for exam-ple, suggest that on average for each
year ofschooling, a person’s earnings increase byabout 11
percent.12 While the economicvalue of education has been well
docu-mented, the question of why education in-creases income is
more controversial. NobelLaureate Gary Becker theorizes that
educa-tion provides skills, or human capital, thatmake a worker
more productive.13 If so, thenbecause a worker’s income reflects
his or herproductivity, education is a key determinantof upward
social mobility. It follows thatmuch of the gap between the rich
and thepoor arises from a lack of skills among thepoor—with the
policy implication being thateducation and training should form the
cor-nerstone of programs aimed at reducing in-come inequality.
Other researchers, such as Nobel LaureateMichael Spence, argue
that education maynot generate higher incomes—that is, the
re-lationship may not be causal.14 Instead, edu-cation and income
may be linked becausepeople with greater “ability” complete
moreschooling and would likely earn higher wagesand salaries even
without the additionalschooling. In this case, as with the
relation-ship between family socioeconomic statusand a child’s
educational attainment, the
06 5563 Rouse-Barrow.qxp 8/6/2006 4:56 PM Page 105
-
alter their income independently. The idea isstraightforward.
Suppose that researchersknew of an event or policy, such as an
in-crease in the compulsory schooling age, thatwould increase a
group’s years of completedschooling. Suppose, further, that they
werecertain that the policy would have no directeffect on the
group’s earnings. They wouldthen estimate the effect of education
onearnings in two steps. First, they would esti-
mate how much the policy increased thegroup’s educational
attainment. Next, theywould measure how much the same policy
af-fected their earnings. If they find that thegroup’s earnings
have increased, they can besure that education caused the increase
be-cause they are certain the policy had no di-rect effect on
earnings. The ratio of the in-crease in income to the increase in
schoolingis an estimate of the economic value of edu-cation. Many
such studies estimate that thereturn to schooling is at least as
large as esti-mates by conventional procedures that relatethe level
of schooling to income directly.15
Other researchers have used sibling or twinpairs to estimate
empirically the return toschooling. Because siblings and twin
pairsshare genetic material and are raised in simi-lar household
environments, their “ability”
C e c i l i a E l e n a R o u s e a n d L i s a B a r r o w
106 T H E F U T U R E O F C H I L D R E N
The findings of all theseempirical studies . . . aresurprisingly
consistent: thereturn to schooling is notcaused by an
omittedcorrelation between abilityand schooling.
and other unobservable characteristics aremuch more similar than
those of randomlyselected members of the population. As a re-sult,
when researchers relate differences insiblings’ schooling to their
earnings, they im-plicitly account for these unobserved
factors.Although the estimated return to schoolingvaries because of
the widely different timeperiods covered by the studies, the
varioussibling and twin studies find a significant linkbetween
schooling and earnings.16 Further,the more recent and more
sophisticated esti-mates typically do not differ from conven-tional
estimates of the return to schooling.17
The findings of all these empirical studies—those using natural
experiments and thoseusing family relationships—are
surprisinglyconsistent: the return to schooling is notcaused by an
omitted correlation betweenability and schooling. A conventional
esti-mate of the economic value of education isthus likely to be
quite close to that of theideal experiment. In fact Nobel
LaureateJames Heckman, writing with PedroCarneiro, concludes, “By
now there is afirmly established consensus that the meanrate of
return to a year of schooling, as of the1990s, exceeds 10 percent
and may be ashigh as 17 to 20 percent.”18
Do Differences in the Value of EducationExplain Differences in
EducationalAttainment?Although researchers consistently find
thateducation has a causal effect on earnings—that education has
economic value—theyhave not come to a consensus on whetherthat
value varies depending on an individual’sfamily background.
Importantly, they havenot established whether people from
moreadvantaged families complete more schoolingbecause it has
greater value for them. Onestudy, for example, concludes that
individuals
06 5563 Rouse-Barrow.qxp 8/6/2006 4:56 PM Page 106
-
with higher “ability” or from more advan-taged families do not
enjoy greater returns toschooling.19 Other studies find no
variation inthe returns to schooling by the race or eth-nicity of
the individual, or by IQ.20 Still oth-ers, however, find higher
returns to schoolingfor more able individuals.21 Another impor-tant
question is why the return to schoolingmight differ by family
background. Differ-ences in school quality, which we addressbelow,
provide one possible explanation.
Do the Costs of Education Differby Family Background?Education
has various costs, the most obviousof which is the direct cost. For
the 90 percentof U.S. K–12 students who attend publicschool, these
direct costs may be minimal,but parents must still pay for such
school sup-plies as notebooks, pencils, paper, and thelike.22 Based
on our estimates using datafrom the 2002 Consumer Expenditure
Sur-vey, families with children under age eight-een who are headed
by a high school dropoutspend roughly $34 a year on school books
andsupplies, whereas families whose head has agraduate degree spend
roughly $85. Thesedifferences, however, are likely too small
togenerate significant differences in educa-tional attainment.
Education also has psychological costs, infor-mation costs,
opportunity costs, and borrow-ing constraints (the cost of
obtaining funds).At the elementary and secondary levels, it isthese
costs that are likely to be important inexplaining differences in
schooling caused byfamily background.
Differences in Psychological CostsLearning can be frustrating,
and masteringnew material and studying for tests can
betime-consuming. Anything that increases thesepsychological costs
for disadvantaged students
relative to their more privileged peers (that is,makes them
dislike school more) may help ex-plain why they get less
schooling.
As one example, systematic differences in theexpectations of
parents and teachers mayraise the psychological costs for less
advan-taged students. A child from a poorer familymay face
different expectations from parentsand teachers than a child from a
more privi-leged family, even if the two children havethe same
“ability.” If these different expecta-tions, in turn, affect the
children’s academicachievement, then expectations could be
onereason why parental socioeconomic status af-fects schooling.
Data from the NELS indicate that more ad-vantaged parents expect
their children tocomplete more education than less advan-taged
parents do, although virtually all par-ents, regardless of
socioeconomic back-ground, expect their children to completehigh
school. If lower parental expectationscause children to have less
confidence intheir own ability, the children could facehigher
psychological costs. Although we arenot aware of evidence that
parental expecta-tions causally affect children’s
academicachievement, some evidence exists thatteacher expectations
affect both student in-telligence and achievement.
Robert Rosenthal and Lenore Jacobson’sPygmalion in the Classroom
has been widelycited as providing just such evidence.23 Theauthors
administered a baseline intelligencetest to elementary students in
a single schooland then randomly assigned 20 percent ofthe students
to be identified as likely to showa dramatic increase in
intelligence over thenext school year because they were
“latebloomers.” The remaining students served asthe control group.
Rosenthal and Jacobson
U . S . E l e m e n t a r y a n d S e c o n d a r y S c h o o l
s : E q u a l i z i n g O p p o r t u n i t y o r R e p l i c a t i
n g t h e S t a t u s Q u o ?
V O L . 1 6 / N O. 2 / FA L L 2 0 0 6 107
06 5563 Rouse-Barrow.qxp 8/6/2006 4:56 PM Page 107
-
then told the teachers which students hadbeen identified as late
bloomers and later ad-ministered follow-up intelligence tests.
Theyfound that one and two years after being la-beled, the
late-blooming children had gainedmore IQ points than the control
group.Rosenthal and Jacobson’s study has spawnedmany more studies
and has been much criti-cized, but a recent review of the research
by
Lee Jussim and Kent Harber concludes thatteacher expectations do
affect student intelli-gence, though the effects are likely
small.24
A recent study by economist David Figlio alsofinds that teacher
expectations affect aca-demic achievement.25 Starting with the
as-sumption that teachers’ perceptions of achild’s family
background may be based onthe child’s name, Figlio assigns
socioeconomicstatus rankings to student names. Becausesiblings’
names are often assigned differentrankings, Figlio can look for
differences intreatment and outcomes among students withidentical
family background. He finds thatteachers are more likely to
recommend stu-dents with high-status names to gifted and tal-ented
programs than students with similartest scores but low-status
names. In addition,using standardized test scores, he finds
thatchildren with low-status names score lower in
mathematics and reading than their siblingswith higher-status
names.
Findings from both economics and psychol-ogy suggest that
teacher expectations may in-deed help explain why family background
af-fects student achievement. If teachers havelower expectations
for children from disad-vantaged families, regardless of their
ability,and if their perceptions about which childrenare
disadvantaged are on average correct,then the lower expectations
for disadvantagedchildren may raise the psychological costs
ofeducation relative to their more privilegedpeers and thus help
explain why children ofdisadvantaged parents attain less
education.
Differences in social or cultural identity mayalso generate
differences in the psychologicalcosts of schooling. In other words,
those whodrop out may feel more peer or family pres-sure not to
continue in school. Again, how-ever, one might ask why these
cultural orsocial norms about education vary systemati-cally with
socioeconomic status. Culturalnorms may vary because education
helps de-termine socioeconomic status, so that disad-vantaged
children may feel pressure not toraise their own status through
educationabove the average for the social and culturalgroup with
which they most identify.
Information DifferencesAnother potential cost to completing
moreschooling is that of acquiring accurate infor-mation about the
costs and benefits of moreschooling. If students from more
privilegedfamilies can get more or better informationabout the
ramifications of their decision at alower cost than those from
poorer families(for example, a better understanding of thepotential
benefits to continuing in school,perhaps because of better family
social net-works), then they may get more schooling.
C e c i l i a E l e n a R o u s e a n d L i s a B a r r o w
108 T H E F U T U R E O F C H I L D R E N
Findings from botheconomics and psychologysuggest that
teacherexpectations may indeed helpexplain why familybackground
affects studentachievement.
06 5563 Rouse-Barrow.qxp 8/6/2006 4:56 PM Page 108
-
Similarly, students who drop out may believethat the returns to
schooling are much riskierthan do students who continue, thus
loweringtheir expectations of the value of a highschool diploma.
High school dropouts mayalso discount the future income benefits
ofmore education at a much higher rate thanthose who graduate from
high school, alsoleading them to have lower expectations ofthe
value of more education.
Although such arguments could explain whysome students decide to
drop out of highschool in spite of the seemingly large
economicbenefits of continuing, one needs to ask whyperceptions of
risk or discount rates vary sys-tematically with family background.
Further,low-income students appear to understand thepotential
economic benefits of college atten-dance about as well as more
advantaged stu-dents.26 Although research is far from conclu-sive,
it suggests that a simple asymmetry instudents’ understanding of
the costs and bene-fits of schooling is unlikely to fully explain
dif-ferences in educational attainment.27
Opportunity Costs and BorrowingConstraintsBecause students
cannot work during thehours when they are attending school,
theyforgo income to attend school. In some fami-lies that income is
a nontrivial share of familyincome. If instead the family could
borrowmoney to allow the child to continue inschool, then the
increase in earnings fromgetting, say, a high school diploma
wouldallow the family to repay the loan (and thensome), assuming
that interest rates are lowerthan the return to schooling. If
credit mar-kets are perfect—that is, if all families canborrow as
much money as they need at theprevailing interest rate—then
educational at-tainment should not vary by family back-ground. If,
however, poor families lack access
to competitive credit markets and would haveto borrow money at
much higher interestrates, then the cost of continuing in school
ishigher for them than for wealthier familieswho do not need to
borrow the money (orwho can borrow it at competitive rates). Inthis
case, students from wealthier familieswill complete more schooling
than thosefrom poorer families.
Whether borrowing constraints more gener-ally explain
differences in educational attain-ment, especially college
attendance, by fam-ily background is an unresolved issue.28
There is, however, growing evidence fromoutside education that
individuals, particu-larly teenagers, are credit constrained.29
Fur-ther, racial and gender discrimination incredit markets has
long been documented.30
For example, researchers at the Federal Re-serve Bank of Boston
investigating racial dis-crimination in mortgage lending in
theBoston area in 1990 found that the loan re-jection rates of
African American and His-panic applicants were 8 percentage
pointshigher than those of otherwise similar whiteapplicants.31
Although race is certainly corre-lated with socioeconomic status,
we know ofno direct evidence of discrimination by so-cioeconomic
status.
Overall, the evidence suggests that differ-ences in the cost of
education may help ex-plain differences in educational attainmentby
family background. As we will show, manyof these cost differences
are potentiallydriven by variation in school quality by
familybackground, which may also lead to differ-ences in the value
of schooling.
Can Differences in School QualityExplain the Patterns?Finally,
we consider whether differences inschool quality help explain why
more privi-
U . S . E l e m e n t a r y a n d S e c o n d a r y S c h o o l
s : E q u a l i z i n g O p p o r t u n i t y o r R e p l i c a t i
n g t h e S t a t u s Q u o ?
V O L . 1 6 / N O. 2 / FA L L 2 0 0 6 109
06 5563 Rouse-Barrow.qxp 8/6/2006 4:56 PM Page 109
-
leged students complete more schooling thantheir less privileged
counterparts. We beginby noting that the conventional measure ofan
individual’s education—years of com-pleted schooling—is rather
limited. In partic-ular, it ignores whether students with thesame
level of completed education may havereceived an education of
different quality. Bythe conventional measure, completing oneyear
of education should increase an individ-ual’s human capital by the
same amount re-gardless of the school attended. But becauseone year
at a poor school may increasehuman capital less than does one year
at anexcellent school, school quality could affectthe value of
education. It could also arguablyaffect the cost of education. A
low-qualityschool, for example, may leave a student un-prepared to
master the skills of the nextgrade level, thus raising the costs in
psycho-logical terms (and also in time) of gettingmore
education.
Does Family Background Affect theQuality of a Child’s School?In
the United States, the school a child at-tends is largely
determined by the neighbor-hood where he or she lives. To the
extent thatparental socioeconomic status affects theneighborhood
where a child lives, it may thusalso affect school quality. For
example, lessprivileged parents certainly have fewer finan-cial
resources than more privileged ones.While many forms of financial
aid are avail-able to low-income students who want to at-tend
college, no such credit is available tolow-income parents who want
to live in ahigh-quality school district. These
borrowingconstraints likely cause school quality to varyby family
background. If poor school qualityleads to lower educational
attainment, thenchildren of less privileged parents will havelower
educational attainment than children ofmore privileged parents.
At the school or school district level, some po-tential
indicators of school quality are clearlyrelated to family
background or income(which, in turn, is correlated with family
so-cioeconomic status). An obvious first questionis whether overall
school spending differsfrom one district to the next by the average
so-cioeconomic status of the residents of the dis-trict.
Higher-income school districts, after all,have more money to spend
on education, andin theory more money should buy higherschool
quality. Using data from the 2003 Com-mon Core of Data, we
calculate average perpupil spending in school districts with at
least70 percent of students eligible for free or re-duced-price
school lunch and districts withless than 20 percent of pupils
eligible.
Not surprisingly, we find that average spend-ing per pupil is
rather similar. Districts withthe larger share of disadvantaged
childrenspend an average of $10,414 per pupil, asagainst $9,647 for
districts with a smallershare of such children. The similarity
inspending in part reflects school finance re-forms since the 1970s
that have tried toequalize school funding across poor and
richdistricts. But similar total spending per pupildoes not
necessarily reflect similar schoolquality, because different school
districts mayface different costs. Older school districtswith aging
buildings, for example, may haveto spend more to maintain their
facilitiesthan newer suburban districts do. Some dis-tricts may
have more special education stu-dents, who need smaller classes,
whichmeans hiring more teachers. And urban dis-tricts may face
higher-wage labor marketsthan rural districts. Indeed, the
recognitionthat some groups of students may need extramoney to
compensate for family disadvan-tage underlies the goal of closing
achieve-ment gaps between high- and low-perform-ing children in
Title I of the Elementary and
C e c i l i a E l e n a R o u s e a n d L i s a B a r r o w
110 T H E F U T U R E O F C H I L D R E N
06 5563 Rouse-Barrow.qxp 8/6/2006 4:56 PM Page 110
-
Secondary Education Act of 1965 (of whichthe No Child Left
Behind Act of 2001 is themost recent reauthorization.)
Given that instructional salaries and benefitsmake up more than
50 percent of schools’total current spending, class size could be
an-other way in which school quality could varyby family
background.32 Because data onclass size are not readily available,
we look atpupil-teacher ratios instead. We have alsocalculated the
average pupil-teacher ratiosfor schools by family socioeconomic
back-ground. As with total school spending, thepupil-teacher ratios
are quite similar: 16.9 forschools attended by children of
disadvan-taged family background, as against 17.4 forschools
attended by more privileged chil-dren.33 Does the lower ratio in
schools serv-ing poor children mean that the quality ofschooling is
better in those schools? Such aninterpretation is not likely to be
correct be-cause those schools may have a larger shareof special
education or English-language-learner students than schools serving
moreprivileged children, which have fewer specialeducation
classrooms.34
One aspect of school quality that is less proneto distortion by
compensatory education poli-cies is teacher quality. Although a
districtmay be able to raise salaries as an incentive
tohigh-quality teachers, it cannot force suchteachers to accept its
job offers. One measureof teacher quality is teaching experience,
andit is telling that schools serving poorer stu-dents are likely
to have fewer experiencedteachers. In this case, schools’
socioeconomicstatus is defined by the percentage of stu-dents who
are eligible for free or reduced-price school lunch. Eighty percent
of teach-ers in low socioeconomic status schools(those in the top
quartile by share eligible forfree or reduced-price lunch) have
more than
U . S . E l e m e n t a r y a n d S e c o n d a r y S c h o o l
s : E q u a l i z i n g O p p o r t u n i t y o r R e p l i c a t i
n g t h e S t a t u s Q u o ?
V O L . 1 6 / N O. 2 / FA L L 2 0 0 6 111
three years of experience, compared with 89percent of teachers
in high socioeconomicstatus schools (those in the bottom quartileby
share eligible).35
Hamilton Lankford, Susanna Loeb, andJames Wyckoff look in more
detail at differ-ences in teacher quality by student
character-istics for the state of New York. They findthat poor
students are more likely than non-
poor students to have a teacher who is notcertified in any
subject that he or she isteaching (21 percent versus 16 percent),
whofailed a certification exam on the first attempt(28 percent
versus 20 percent), or who at-tended a college ranked “least
competitive”by Barron’s College Guide (25 percent versus24
percent).36
Schools also vary in facility and peer quality.As figure 3
shows, low socioeconomic statusschools (those with 70 percent or
more chil-dren eligible for free or reduced-price schoollunch) have
worse facilities than high socioe-conomic status schools (those
with fewerthan 20 percent of students eligible for freeor
reduced-price school lunch). Fifty-sevenpercent of low
socioeconomic status schoolshave no temporary buildings, as against
65percent of schools serving high socioeco-nomic status students.
Similarly, 37 percent
One measure of teacherquality is teaching experience,and it is
telling that schoolsserving poorer students arelikely to have
fewerexperienced teachers.
06 5563 Rouse-Barrow.qxp 8/6/2006 4:56 PM Page 111
-
of schools serving poor children (low socioe-conomic status
schools) have fully adequatebuilding features, compared with 55
percentof schools serving nonpoor children (high so-cioeconomic
status schools).37
Peer quality as measured by college enroll-ment rates and
Advanced Placement coursesis also lower for less privileged
children. Datafrom the NELS show that low socioeconomicstudents
(those with parents in the bottomquartile by socioeconomic status)
attendschools in which only 56 percent of studentsgo on to some
college, as against 75 percentof students in schools serving high
socioeco-nomic status students (those in the top quar-tile by
socioeconomic status). The share ofstudents taking Advanced
Placement coursesis 16.9 percent in schools attended by stu-dents
with low socioeconomic status, com-pared with 26.2 percent for
schools attendedby high socioeconomic status students. Inshort, the
peers of less privileged studentsare not as academically oriented
as the peersof wealthier students.
Finally, we have found some evidence thatschool districts that
are low in socioeconomicstatus may not spend resources as
efficientlyas districts with higher socioeconomic status,suggesting
that they may be more poorlymanaged.38 This finding, in
combinationwith the descriptive data above (in figure 3),leads us
to conclude that school qualityvaries according to parental
socioeconomicstatus.
Does School Quality Affect Children’sEducational Attainment?The
next question is whether these differ-ences in school quality
translate into worseoutcomes for less privileged children. By
theearly 1990s, many people were convincedthat once one took
account of differences infamily background, school
resources—in-cluding money—did not matter for studentachievement.
In a 1996 article economistEric Hanushek wrote, “Three decades of
in-tensive research leave a clear picture thatschool resource
variations are not closely re-lated to variations in student
outcomes and,
C e c i l i a E l e n a R o u s e a n d L i s a B a r r o w
112 T H E F U T U R E O F C H I L D R E N
Figure 3. School Quality, by Family Socioeconomic Status
Source: Daniel P. Mayer, John E. Mullins, and Mary T. Moore,
Monitoring School Quality: An Indicators Report (NCES 2001-030)
(U.S.Department of Education, National Center for Education
Statistics, 2000), figure 2.3
(http://www.nces.ed.gov/pubs2001/2001030.pdf[September 5, 2005]);
T. D. Snyder, A. G. Tan, and C. M. Hoffman, Digest of Education
Statistics, 2003 (NCES-2005-025) (U.S. Depart-ment of Education,
National Center for Education Statistics, 2004), table 101
(http://www.nces.ed.gov/pubs2005/2005025.pdf [February26,
2005]).
90
80
70
60
50
40
30
20
10
0
Low SES
High SES
High school graduates enrolled
in college
12th graders enrolled in Advanced
Placement classes
Schools with no inadequate
building features
Schools with no temporary
buildings
Percent
Teachers with more than 3 years
of experience
06 5563 Rouse-Barrow.qxp 8/6/2006 4:56 PM Page 112
-
by implication, that aggressive spending pro-grams are unlikely
to be good investmentprograms unless coupled with other
funda-mental reforms.”39
Although Hanushek’s analyses of the effectsof school resources
on student achievementhave been very influential, other
researchershave criticized his findings on methodologi-cal
grounds.40 For example, one independentanalysis of one of
Hanushek’s studies con-cludes that the effect of per pupil
spendingon student achievement is large and educa-tionally
significant.41 More recent studiesthat make explicit attempts to
account for thecompensatory nature of much educationalexpenditure
also provide evidence thatmoney matters. One of our own studies
findsthat the market values school spending interms of property
values. And JonathanGuryan finds that a $1,000 increase in perpupil
spending in Massachusetts increasesaverage test scores for fourth-
and eighth-grade students by one-third to one-half of astandard
deviation. Importantly, in sum-marizing the findings of seventeen
federalstudies, Geoffrey Borman and JeromeD’Agostino conclude that
Title I of the Ele-mentary and Secondary Education Act,which aims
to provide additional funding toschools and districts serving
disadvantagedstudents, has indeed improved the educa-tional
outcomes of children it has served.Further, in studying the effect
of state effortsto equalize funding between wealthier andpoorer
school districts, David Card and A.Abigail Payne find that such
reforms havenarrowed gaps in spending as well as in edu-cational
outcomes.42
Whether money matters must depend in parton how the money is
spent. Probably the bestevidence to date on the effect of class
sizecomes from the Tennessee Student-Teacher
Achievement Ratio experiment (known asProject STAR), the
nation’s largest random-ized experiment aimed at understanding
howsmaller class sizes affect student achieve-ment.43 In the
1985–86 school year some6,000 kindergarten students in
Tennesseewere randomly assigned to one of threegroups: small
classes (13–17 students perteacher), regular-sized classes (22–25
stu-dents), and regular-sized classes with a
teacher’s aide. The experiment, ultimately in-volving some
11,600 students, lasted fouryears. After the third grade, all
students re-turned to regular-sized classes.44 The datahave been
analyzed by a variety of re-searchers, with a remarkably consistent
find-ing: smaller classes result in higher studentachievement.45
One study finds that theclass-size effects are larger for students
eligi-ble for free or reduced-price school lunchthan for more
well-to-do students. Anotherreports that the students who were
(ran-domly) placed in smaller classes in gradesK–3 performed better
on standardized testswhen they reached the eighth grade. Theywere
also more likely to take a college en-trance exam (such as the ACT
or SAT)—asignal that they may have been more likely toattend
college as well.46
Yet another study, by David Card and AlanKrueger, relating the
quality of schooling re-ceived by people born between 1920 and
U . S . E l e m e n t a r y a n d S e c o n d a r y S c h o o l
s : E q u a l i z i n g O p p o r t u n i t y o r R e p l i c a t i
n g t h e S t a t u s Q u o ?
V O L . 1 6 / N O. 2 / FA L L 2 0 0 6 113
Peer quality as measured by college enrollment ratesand Advanced
Placementcourses is also lower for less privileged children.
06 5563 Rouse-Barrow.qxp 8/6/2006 4:56 PM Page 113
-
ence. If that finding is accurate, the fact thatschools serving
poorer students have moreteachers with very little experience
suggeststhat these students will have lower achieve-ment as a
result.
Does Improving School AccountabilityImprove Student Performance?
Given already high levels of educationalspending, policymakers are
looking for waysto provide incentives for schools to improvewithout
large increases in revenues. “Schoolaccountability” programs come
in twoforms.50 Institutional school accountabilityprograms, such as
the No Child Left BehindAct of 2001, set up a system of rewards
andsanctions determined by school perform-ance—typically, student
performance onstandardized tests. Significantly, No ChildLeft
Behind makes each school’s perform-ance public. These reforms are
popular be-cause they are relatively inexpensive and be-cause they
aim to make school systems moretransparent, so that parents can
more readilycompare their child’s school with others. Al-though
research on the effects of school ac-countability on student
achievement is grow-ing, it is still fledgling.51 At best,
suchprograms generate small improvements instudent achievement. At
the same time, re-searchers have documented several unin-tended
consequences. For example, onestudy estimates that teachers cheat
in 4–5percent of elementary school classrooms eachyear in Chicago
and suggests that cheating in-creases when teachers have an
incentive to doso, as they have with high-stakes tests.52
Other researchers find that administrators re-classify
low-achieving students as learningdisabled so that the (presumably
low) scoresof these students will not be included in theschool’s
average test score calculation.53
David Figlio reports that schools are morelikely to suspend
students during the testing
C e c i l i a E l e n a R o u s e a n d L i s a B a r r o w
114 T H E F U T U R E O F C H I L D R E N
1949 to their earnings in 1979 found that a re-duction in the
pupil-teacher ratio of 10 stu-dents increased average earnings by
4.2 per-cent.47 Other studies reviewed by these sameauthors in a
later study find that reductions inpupil-teacher ratios are
associated with in-creased average earnings, although several ofthe
estimates are not statistically significant.48
Economic studies also broadly agree thatteacher quality matters,
though they agreemuch less about what makes a
high-qualityteacher.49 Developing credible studies of theeffects of
particular teacher characteristics onstudent achievement is
extremely difficult.Because teachers are not randomly assignedto
schools, studies find ostensibly “better”teachers at schools
attended by more advan-taged students. Thus, as in other areas,
theresearchers can develop links between cer-tain teacher
characteristics and student out-comes but cannot be assured that
the teachercharacteristics caused the change in studentoutcomes. In
addition, such studies typicallyrely on administrative data that do
not con-tain many of the characteristics that likelymake a good
teacher, such as classroom man-agement, motivation,
professionalism, and athorough understanding of how to communi-cate
new concepts to students. That said,some studies have found that
teachers im-prove greatly after one or two years of experi-
Economic studies alsobroadly agree that teacherquality matters,
though they agree much less aboutwhat makes a
high-qualityteacher.
06 5563 Rouse-Barrow.qxp 8/6/2006 4:56 PM Page 114
-
cycle, apparently to alter the composition ofthe testing pool.
Brian Jacob finds some evi-dence that teachers focused more on
thehigh-stakes test material than on the low-stakes test material
following the introductionof Chicago’s school accountability
system.54
Another potential form of accountability isthrough the market.
Because students are as-signed to schools based on their
neighbor-hood, many observers have argued that localpublic schools
are not required to be ac-countable to local citizens. Thus, if
parentscould “vote with their feet,” competitivepressure and the
threat of losing studentswould force such schools to improve.
Twooften talked-about forms of competitive pres-sure are charter
schools—public schools thatare exempt from many of the regulations
thatapply to traditional public schools—andschool vouchers for use
at private schools.Both forms of competition would give par-ents
alternatives to the local public school,thus presumably improving
both the educa-tional achievement of their children and thequality
of the local public schools. Impor-tantly, because the
accountability is enforcedby parental choices rather than the rules
of asystem, there is less scope for the unintendedconsequences
noted above.
Although these arguments are theoreticallypersuasive, there is
little empirical evidencethat either charter schools or school
vouchersimprove student test scores (which should, inturn, improve
educational attainment). Forexample, three sets of researchers,
usingstatewide data from North Carolina, Florida,and Texas,
respectively, have studied whetherstudents who attend charter
schools havehigher test score gains than students in localpublic
schools.55 Their findings are remark-ably similar: there are no
achievement gainsfor students who attend charter schools, even
after controlling for a rich set of studentcharacteristics. In
fact, the students in char-ter schools appear to perform worse,
perhapsbecause these are often new schools.
Evidence on school vouchers is also decid-edly mixed. The
best-designed study ofschool vouchers was conducted by
WilliamHowell and Paul Peterson in New York City,beginning in
1997.56 It randomly assigned1,300 students to two groups. One group
re-ceived a (privately funded) scholarship to at-tend a private
school; the other, control,group did not. After three years, the
studyfound that overall there were no test scoregains among the
students who were offered avoucher or among the students who
actuallytook advantage of the voucher offer and at-tended private
schools. Howell and Petersonreported educationally large and
statisticallymeaningful gains among African Americanstudents, but
their findings have been dis-puted in a reanalysis of the data by
Kruegerand Pei Zhu.57
Evidence from publicly funded voucher pro-grams in Milwaukee and
Cleveland does nothelp to clarify the issue. One study of
Mil-waukee’s Parental Choice Program, the old-est publicly funded
choice program in theUnited States, suggests that students gainedin
math but not in reading; another suggestsno gains in either math or
reading.58 Themost recent evidence from the ClevelandScholarship
and Tutoring Program suggeststhat vouchers have not significantly
benefitedthe recipient students.59 After five years thetest scores
of voucher students are generallyquite similar to those of a group
of studentswho applied for, but did not receive, avoucher.
Importantly, all these studies examine small-scale programs.
None addresses the question
U . S . E l e m e n t a r y a n d S e c o n d a r y S c h o o l
s : E q u a l i z i n g O p p o r t u n i t y o r R e p l i c a t i
n g t h e S t a t u s Q u o ?
V O L . 1 6 / N O. 2 / FA L L 2 0 0 6 115
06 5563 Rouse-Barrow.qxp 8/6/2006 4:56 PM Page 115
-
of whether a large-scale program would gen-erate enough
competitive pressure on thepublic schools to induce them to
improve.Evidence from Florida’s school accountabil-ity system
(which includes a school voucherfor students attending persistently
“failing”schools) suggests that even the threat of los-ing students
through vouchers may not be aprime motivator for school
improvement.60
Although schools faced with the possibility of
becoming voucher-eligible appear to improveslightly, such
improvement appears to springfrom avoidance of the stigma of being
labeleda failing school rather than the threat ofvouchers per
se.
Although these studies are not likely to bethe last word on the
effectiveness of institu-tional school accountability systems,
charterschools, or school vouchers, together they in-dicate that
the gains from improving schoolaccountability are likely modest, at
best.
ConclusionsWhile efforts such as Title I and state schoolfinance
equalizations have succeeded insmoothing school spending across
school dis-tricts serving more and less advantaged stu-dents, they
have not eliminated the link be-tween socioeconomic status and
educationaloutcomes. Family background continues to
play an important role in determining achild’s educational
attainment. The costs andbenefits of getting further schooling
differaccording to the socioeconomic status of achild’s family, and
these differences may bedriven by differences in access to
qualityschools. Because school attendance bound-aries are largely
determined by neighbor-hood of residence and because families
ofdifferent socioeconomic backgrounds live indifferent
neighborhoods, children from moreand less advantaged backgrounds
attend dif-ferent schools. Descriptive statistics and
moresophisticated analyses find that school qualityis positively
correlated with family back-ground. Children from well-to-do
families at-tend better schools than children from poorfamilies. As
a result, rather than encouragingupward mobility, U.S. public
schools tend toreinforce the transmission of low socio-economic
status from parents to children.
Policy interventions aimed at improvingschool quality for
children from disadvan-taged families thus have the potential to
in-crease social mobility by reducing the trans-mission of low
socioeconomic status fromparents to children through education.
Basedon the best research evidence, smaller classsizes seem to be
one promising avenue forimproving school quality for
disadvantagedstudents. Maintaining teacher quality at thesame time
is also likely to be important.These are but two of the many
avenues thatgrowing evidence shows are effective in rais-ing school
quality. Smaller schools, grade re-tention, and summer school are
examples ofothers.61 Despite the considerable politicalattention
paid to charter schools and vouch-ers that would help the children
of poor fam-ilies attend private school, to date the bestevidence
suggests that increasing competitivepressure in this way will not
significantly im-prove student achievement. In contrast,
C e c i l i a E l e n a R o u s e a n d L i s a B a r r o w
116 T H E F U T U R E O F C H I L D R E N
Rather than encouragingupward mobility, U.S. publicschools tend
to reinforce the transmission of low socioeconomic status
fromparents to children.
06 5563 Rouse-Barrow.qxp 8/6/2006 4:56 PM Page 116
-
growing evidence suggests that institutionalaccountability
systems may generate smallimprovements in student achievement,
al-though they are also vulnerable to unin-tended negative
consequences.
Because a child’s educational achievementdepends on so many
aspects of his or her life,many of which are outside school,
educationpolicy can go only so far. One particular chal-lenge is
that more advantaged families can af-ford to—and will—spend more on
their chil-dren’s education. Thus, these families canpartly undo
policy attempts to equalizeschool quality for poor and nonpoor
children
by spending more money outside school. Asan example, based on
data from the 2002Consumer Expenditure Survey, parents whodrop out
of high school spend an average of$33 a year for recreational
lessons or otherinstruction for children (not including tu-ition),
whereas parents who have graduatedegrees spend nearly $600. Under
these cir-cumstances, it will be extremely difficult forAmerica’s
public schools to live up to MartinLuther King Jr.’s ideal of
teaching students sowell as to make their family background
irrel-evant. That said, such lofty goals are a stan-dard by which
to measure our efforts. We arereminded that we have a long way to
go.
U . S . E l e m e n t a r y a n d S e c o n d a r y S c h o o l
s : E q u a l i z i n g O p p o r t u n i t y o r R e p l i c a t i
n g t h e S t a t u s Q u o ?
V O L . 1 6 / N O. 2 / FA L L 2 0 0 6 117
06 5563 Rouse-Barrow.qxp 8/6/2006 4:56 PM Page 117
-
Notes
1. Based on authors’ calculations using March Current Population
Survey data available from Unicon. We
limit the sample to individuals twenty-five to sixty-five years
of age who worked at least one week in the
past year.
2. Based on authors’ calculations using 2004 Current Population
Survey March Outgoing Rotations Group data
available from Unicon. We limit the sample to individuals aged
twenty-five to sixty-five, omitting those with
wages of less than one-half of the minimum wage or above the
99th percentile of the wage distribution.
3. See, for example, Derek Neal and William R. Johnson, “The
Role of Pre-Market Factors in Black-White
Wage Differences,” Journal of Political Economy 104, no. 5
(1996): 869–95; and Orley Ashenfelter and Ce-
cilia Elena Rouse, “Income, Schooling, and Ability: Evidence
from a New Sample of Twins,” Quarterly
Journal of Economics 113, no. 1 (1998): 253–84.
4. Bruce Sacerdote, “What Happens When We Randomly Assign
Children to Families?” Working Paper
10894 (Cambridge, Mass.: National Bureau of Economic Research,
2004).
5. See William T. Dickens, “Genetic Differences and School
Readiness,” Future of Children 15, no. 1 (2005):
55–69.
6. See Mike Stoolmiller, “Implications of the Restricted Range
of Family Environments for Estimates of Her-
itability and Nonshared Environment in Behavior-Genetic Adoption
Studies,” Psychological Bulletin 125,
no. 4 (1999): 392–409.
7. Pamela Morris, Greg J. Duncan, and Christopher Rodrigues,
“Does Money Really Matter? Estimating Im-
pacts of Family Income on Children’s Achievement with Data from
Random-Assignment Experiments,”
unpublished manuscript, MDRC and Northwestern University,
2004.
8. Alan B. Krueger and Diane M. Whitmore, “The Effect of
Attending a Small Class in the Early Grades on
College Test Taking and Middle School Test Results: Evidence
from Project STAR,” Economic Journal 111
(2001): 1–28. This calculation assumes that per pupil costs
increase 47 percent per year for 2.3 years, on av-
erage, and that per pupil costs equal the average U.S. total
expenditure per pupil in average daily atten-
dance in 1997–98 in 2001–02 dollars ($8,487). Digest of
Education Statistics 2003, table 166.
9. Gordon B. Dahl and Lance Lochner, “The Impact of Family
Income on Child Achievement,” Working
Paper 11279 (Cambridge, Mass.: National Bureau of Economic
Research, 2005).
10. Based on calculations by the authors using data received
through personal correspondence with Dahl.
11. As Jacob Mincer shows, if forgone earnings are the only cost
of school attendance, this is the private mar-
ginal benefit (or “return”) to the investment in a year of
schooling. See Jacob Mincer, Schooling, Experi-
ence, and Earnings (Columbia University Press, 1974).
12. Based on a regression of the natural logarithm of hourly
wages on years of completed education, a quad-
ratic in potential experience controls for sex, race/ethnicity,
marital status, and nine regions using the 2004
March Current Population Survey. The regression was weighted
using the earnings weight.
13. Gary Becker, Human Capital (Columbia University Press,
1964).
14. Michael Spence, “Job Market Signaling,” Quarterly Journal of
Economics 87, no. 3 (1973): 355–74.
C e c i l i a E l e n a R o u s e a n d L i s a B a r r o w
118 T H E F U T U R E O F C H I L D R E N
06 5563 Rouse-Barrow.qxp 8/6/2006 4:56 PM Page 118
-
15. For example, Joshua D. Angrist and Alan B. Krueger, “Does
Compulsory Schooling Affect Schooling and
Earnings?” Quarterly Journal of Economics 106, no. 4 (1991):
979–1014; Thomas J. Kane and Cecilia
Elena Rouse, “Labor Market Returns to Two- and Four-Year
Colleges,” American Economic Review 83,
no. 3 (1993): 600–13; Jeffrey Kling, “Interpreting Instrumental
Variables Estimates of the Returns to
Schooling,” Journal of Business and Economics Statistics 19, no.
3 (2001): 358–64; David Card, “Using Ge-
ographic Variation in College Proximity to Estimate the Return
to Schooling,” Working Paper 4483 (Cam-
bridge, Mass.: National Bureau of Economic Research, 1993); and
Philip Oreopoulos, “Average Treatment
Effects of Education When Compulsory School Laws Really Matter,”
American Economic Review (forth-
coming). Angrist and Krueger use an individual’s quarter of
birth as the natural experiment; Kane and
Rouse, Card, and Kling use proximity to a two- or four-year
college as the natural experiment.
16. For studies using siblings, see, for example, Orley
Ashenfelter and David Zimmerman, “Estimates of the
Returns to Schooling from Sibling Data: Fathers, Sons, and
Brothers,” Review of Economics and Statistics
79, no. 1 (1997): 1–9; and Joseph Altonji and Thomas Dunn, “The
Effects of Family Characteristics on the
Return to Education,” Review of Economics and Statistics 78, no.
4 (1996): 692–704. For studies using
twins, see Jere R. Behrman, Mark R. Rosenzweig, and Paul
Taubman, “Endowments and the Allocation of
Schooling in the Family and in the Marriage Market: The Twins
Experiment,” Journal of Political Economy
102, no. 6 (1994): 1131–74; Orley Ashenfelter and Cecilia Elena
Rouse, “Income, Schooling, and Ability:
Evidence from a New Sample of Twins,” Quarterly Journal of
Economics 113, no. 1 (1998): 253–84; and
Cecilia Elena Rouse, “Further Estimates of the Economic Return
to Schooling from a New Sample of
Twins,” Economics of Education Review 18, no. 2 (1999):
149–57.
17. Unfortunately, the measurement error in reported schooling
poses an econometric challenge for these
models. The reason is that classical measurement error is
exacerbated in within-sibling (or within-twin) es-
timators because sibling education levels are so highly
correlated. Zvi Griliches, “Estimating the Returns to
Schooling: Some Econometric Problems,” Econometrica 45, no. 1
(1977): 1–22. As a result, much of the
more recent literature using this approach has focused on
addressing the measurement error bias as well as
the ability bias.
18. Pedro Carneiro and James J. Heckman, “Human Capital Policy,”
in Inequality in America: What Role for
Human Capital Policies? edited by Benjamin M. Friedman (MIT
Press, 2003), pp. 148–49.
19. Ashenfelter and Rouse, “Income, Schooling, and Ability” (see
note 16).
20. Lisa Barrow and Cecilia Elena Rouse, “Do Returns to
Schooling Differ by Race and Ethnicity?” American
Economic Review 95, no. 2 (2005): 83–87; and Altonji and Dunn,
“The Effects of Family Characteristics”
(see note 16).
21. John Cawley and others, “Understanding the Role of Cognitive
Ability in Accounting for the Recent Rise
in the Economic Return to Education,” in Meritocracy and
Economic Inequality, edited by Kenneth
Arrow, Samuel Bowles, and Steven Durlauf (Princeton University
Press, 2000), pp. 230–65; Carneiro and
Heckman, “Human Capital Policy” (see note 18); and Christopher
Taber, “The Rising College Premium in
the Eighties: Return to College or Return to Unobserved
Ability?” Review of Economic Studies 68, no. 3
(2001): 665–91.
22. U.S. Department of Education, National Center for Education
Statistics, The Condition of Education
2005, NCES-2005-094 (2005), appendix 1, table 2-2, “Trends in
Private School Enrollments.”
U . S . E l e m e n t a r y a n d S e c o n d a r y S c h o o l
s : E q u a l i z i n g O p p o r t u n i t y o r R e p l i c a t i
n g t h e S t a t u s Q u o ?
V O L . 1 6 / N O. 2 / FA L L 2 0 0 6 119
06 5563 Rouse-Barrow.qxp 8/6/2006 4:56 PM Page 119
-
23. Robert Rosenthal, and Lenore Jacobson, Pygmalion in the
Classroom: Teacher Expectation and Pupils’ In-
tellectual Development (New York: Holt, Reinhart, and Winston,
1968).
24. Lee Jessim and Kent D. Harber, “Teacher Expectations and
Self-Fulfilling Prophecies: Knowns and Un-
knowns, Resolved and Unresolved Controversies,” Personality and
Social Psychology Review 9, no. 2
(2005): 131–55.
25. David Figlio, “Names, Expectations and the Black-White Test
Score Gap,” Working Paper 11195 (Cam-
bridge, Mass.: National Bureau of Economic Research, 2005).
26. For example, see Christopher Avery and Thomas J. Kane,
“Student Perceptions of College Opportunities,”
in College Choices: The Economics of Where to Go, When to Go,
and How to Pay for It, edited by Caroline
M. Hoxby (University of Chicago Press, 2004), pp. 355–91; and
Cecilia Elena Rouse, “Low-Income Stu-
dents and College Attendance: An Exploration of Income
Expectations,” Social Science Quarterly 85, no. 5
(2004): 1299–317.
27. Clearly differences in information costs may be much more
important in the transition from high school to
college, when students need information about where and how to
apply to college and how to go about getting
financial aid. Children with college-educated parents have an
advantage over other children in having parents
who have “been there before.” See the article by Robert Haveman
and Timothy Smeeding in this volume.
28. See, for example, James J. Heckman and Lance Lochner,
“Rethinking Education and Training Policy: Un-
derstanding the Sources of Skill Formation in a Modern Economy,”
in Securing the Future: Investing in
Children from Birth to College, edited by Sheldon Danziger and
Jane Waldfogel (New York: Russell Sage
Foundation, 2000), pp. 47–83; and David T. Ellwood and Thomas J.
Kane, “Who Is Getting a College Ed-
ucation? Family Background and the Growing Gaps in Enrollment,”
in Securing the Future, edited by
Danziger and Waldfogel, pp. 283–324.
29. See John T. Warner and Saul Pleeter, “The Personal Discount
Rate: Evidence from Military Downsizing
Programs,” American Economic Review 91, no. 1 (2001): 33–53; and
David B. Gross and Nicholas Souleles,
“Consumer Response to Changes in Credit Supply: Evidence from
Credit Card Data,” mimeo, University
of Pennsylvania (2000).
30. See Helen Ladd, “Evidence on Discrimination in Mortgage
Lending,” Journal of Economic Perspectives
12, no. 2 (1998): 41–62, for a nice review of the evidence on
discrimination in mortgage lending.
31. Alicia H. Munnell and others, “Mortgage Lending in Boston:
Interpreting HMDA Data,” American Eco-
nomic Review 86, no. 1 (1996): 25–53.
32. T. D. Snyder, A. G. Tan, and C. M. Hoffman, Digest of
Education Statistics, 2003, NCES-2005-025 (U.S.
Department of Education, National Center for Education
Statistics, 2004), table 164
(www.nces.ed.gov/pubs2005/2005025.pdf [February 26, 2005]).
33. Authors’ calculations from the 2003 CCD.
34. See Michael A. Boozer and Cecilia Elena Rouse, “Intraschool
Variation in Class Size: Patterns and Implica-
tions,” Journal of Urban Economics 50, no. 1 (2001): 163–89, for
a more complete discussion of this issue.
35. Daniel P. Mayer, John E. Mullins, and Mary T. Moore,
Monitoring School Quality: An Indicators Report,
NCES 2001-030 (U.S. Department of Education, National Center for
Education Statistics, 2000)
(http://nces.ed.gov/pubs2001/2001030.pdf [September 5, 2005]),
figure 2.3.
C e c i l i a E l e n a R o u s e a n d L i s a B a r r o w
120 T H E F U T U R E O F C H I L D R E N
06 5563 Rouse-Barrow.qxp 8/6/2006 4:56 PM Page 120
-
36. Hamilton Lankford, Susanna Loeb, and James Wyckoff, “Teacher
Sorting and the Plight of Urban Schools:
A Descriptive Analysis,” Educational Evaluation and Policy
Analysis 24, no. 1 (Spring 2002): 37–62, table
6, p. 47.
37. Snyder, Tan, and Hoffman, Digest of Education Statistics,
2003, table 101 (see note 32). The building fea-
tures considered are roofs; framing, floors, and foundations;
exterior walls, finishes, windows and doors; in-
terior finishes and trim; plumbing; heating, ventilation, air
conditioning; electric power; electrical lighting;
and life safety features.
38. Lisa Barrow and Cecilia Elena Rouse, “Using Market Valuation
to Assess Public School Spending,” Journal
of Public Economics 88, no. 9–10 (2004): 1749–71.
39. Eric A. Hanushek, “Measuring Investment in Education,”
Journal of Economic Perspectives 10, no. 4
(1996): 9.
40. See Larry V. Hedges, Richard Laine, and Robert Greenwald,
“Does Money Matter? A Meta-Analysis of
Studies of the Effects of Differential School Inputs on Student
Outcomes,” Education Researcher 23, no.
33 (1994): 5–14; and Alan B. Krueger, “Economic Considerations
and Class Size,” Economic Journal 113
(2003): F34–63.
41. Hedges, Laine, and Greenwald, “Does Money Matter?” (see note
40); and Eric A. Hanushek, “The Impact
of Differential Expenditures on School Performance,” Educational
Researcher 18, no. 4 (1989): 45–65.
42. Barrow and Rouse, “Using Market Valuation to Assess Public
School Spending” (see note 38); Jonathan
Guryan, “Does Money Matter? Regression-Discontinuity Estimates
from Education Finance Reform in
Massachusetts,” Working Paper 8269 (Cambridge, Mass.: National
Bureau of Economic Research, 2001);
Geoffrey D. Borman and Jerome V. D’Agostino, “Title I and
Student Achievement: A Meta-Analysis of
Federal Evaluation Results,” Educational Evaluation and Policy
Analysis 18, no. 4 (1996): 309–26; and
David Card and A. Abigail Payne, “School Finance Reform, the
Distribution of School Spending, and the
Distribution of Student Test Scores,” Journal of Public
Economics 83, no. 1 (2002): 49–82.
43. Other recent papers on the effect of class size use
“quasi-experimental” designs. For example, Joshua D.
Angrist and Victor Lavy, “Using Maimonides’ Rule to Estimate the
Effect of Class Size on Scholastic
Achievement,” Quarterly Journal of Economics 114, no. 2 (1999):
533–75, use the nonlinearity in the de-
termination of class size in Israel to identify an effect of
class size, finding effects on the same order of mag-
nitude as those reported by Boozer and Rouse, “Intraschool
Variation” (see note 34); and Caroline Minter
Hoxby, “The Effects of Class Size on Student Achievement: New
Evidence from Population Variation,”
Quarterly Journal of Economics 115, no. 4 (2000): 1239–85,
exploits variation in the size of the school-aged
population in Connecticut to identify an effect of class size,
finding that small class sizes have no effect on
student achievement.
44. Alan B. Krueger, “Experimental Estimates of Education
Production Functions,” Quarterly Journal of Eco-
nomics 114, no. 2 (1999): 497–531.
45. See, for example, Jeremy D. Finn and Charles M. Achilles,
“Answers and Questions about Class Size: A
Statewide Experiment,” American Educational Research Journal 27,
no. 3 (1990): 557–77; and Alan B.
Krueger, “Experimental Estimates of Education Production
Functions,” Quarterly Journal of Economics
114, no. 2 (1999): 497–531.
U . S . E l e m e n t a r y a n d S e c o n d a r y S c h o o l
s : E q u a l i z i n g O p p o r t u n i t y o r R e p l i c a t i
n g t h e S t a t u s Q u o ?
V O L . 1 6 / N O. 2 / FA L L 2 0 0 6 121
06 5563 Rouse-Barrow.qxp 8/6/2006 4:56 PM Page 121
-
46. Krueger and Whitmore, “The Effect of Attending a Small Class
in the Early Grades” (see note 8). Others
believe the evidence on a positive impact of school quality on
subsequent educational attainment and earn-
ings is not very strong. See, for example, the volume edited by
Gary Burtless, Does Money Matter? The Ef-
fect of School Resources on Student Achievement and Adult
Success (Brookings, 1996), for differing view-
points.
47. David Card and Alan Krueger, “Does School Quality Matter?
Returns to Education and the Characteristics
of Public Schools in the United States,” Journal of Political
Economy 100, no. 1 (1992): 1–40.
48. Card and Krueger, “Labor Market Effects of School Quality:
Theory and Evidence,” in Does Money Mat-
ter? edited by Burtless, pp. 97–140 (see note 46).
49. For example, see Daniel Aaronson, Lisa Barrow, and William
Sander, “Teachers and Student Achievement
in the Chicago Public High Schools,” unpublished manuscript,
Federal Reserve Bank of Chicago (2005);
and Eric A. Hanushek, Steve G. Rivkin, and John F. Kain,
“Teachers, Schools, and Academic Achieve-
ment,” Econometrica 73, no. 20 (2005): 417–58.
50. Another form of accountability targets the student. In this
case, students are not permitted to advance to
the next grade until they have demonstrated a predetermined
level of proficiency in academic subjects. Ev-
idence on these so-called no social promotion policies, however,
is mixed. The best evidence comes from
Brian A. Jacob and Lars Lefgren, “Remedial Education and Student
Achievement: A Regression-
Discontinuity Analysis,” Review of Economics and Statistics 86,
no. 1 (2004): 226–44, who study the intro-
duction of such a policy in the Chicago public schools. They
find that retention increases achievement for
third graders but not for sixth graders.
51. See Martin Carnoy and Susanna Loeb, “Does External
Accountability Affect Student Outcomes? A Cross-
State Analysis,” Education Evaluation and Policy Analysis 24,
no. 4 (2002): 305–31; Melissa Clark, “Educa-
tion Reform, Redistribution, and Student Achievement: Evidence
from the Kentucky Education Reform
Act,” mimeo, Princeton University (2002); David Figlio and
Cecilia Elena Rouse, “Do Accountability and
Voucher Threats Improve Low-Performing Schools?” Journal of
Public Economics 90, nos. 1–2 (2006):
239–55; Eric A. Hanushek and Margaret E. Raymond, “Does School
Accountability Lead to Improved Stu-
dent Performance?” Journal of Policy Analysis and Management 24,
no. 2 (2005): 297–327; Walt Haney,
“The Myth of the Texas Miracle in Education,” Education Policy
Analysis Archives 8, no. 41 (2000); and
Brian A. Jacob, “Accountability, Incentives and Behavior:
Evidence from School Reform in Chicago,” Jour-
nal of Public Economics 89, nos. 5–6 (2005): 761–96.
52. Brian A. Jacob and Steven D. Levitt, “Rotten Apples: An
Investigation of the Prevalance and Predictors of
Teacher Cheating,” Quarterly Journal of Economics 118, no. 3
(2003): 843–77.
53. Julie Berry Cullen and Randall Reback, “Tinkering toward
Accolades: School Gaming under a Perfor-
mance Accountability System,” mimeo, University of Michigan
(2003); David Figlio and Lawrence Getzler,
“Accountability, Ability and Disability: Gaming the System?”
Working Paper 9307 (Cambridge, Mass.: Na-
tional Bureau of Economic Research, 2002); and Jacob,
“Accountability, Incentives and Behavior” (see
note 51).
54. David Figlio, “Testing, Crime and Punishment,” Journal of
Public Economics, forthcoming; and Jacob, “Ac-
countability, Incentives and Behavior” (see note 51). It is
worth noting that while these unintended conse-
quences may have short-run benefits, it is unclear whether any
of them would persist in the long run. If a
C e c i l i a E l e n a R o u s e a n d L i s a B a r r o w
122 T H E F U T U R E O F C H I L D R E N
06 5563 Rouse-Barrow.qxp 8/6/2006 4:56 PM Page 122
-
school increases its average test scores by reclassifying
students, for example, it is unclear whether the
school will continue to experience large gains in the future, as
it can only gain by reclassifying new students.
55. Robert Bifulco and Helen F. Ladd, “The Impacts of Charter
Schools on Student Achievement: Evidence
from North Carolina,” Working Paper SAN04-01 (Durham, N.C.:
Terry Sanford Institute of Public Policy,
2004); Tim R. Sass, “Charter Schools and Student Achievement in
Florida,” mimeo, Florida State Univer-
sity (2004); and Eric A. Hanushek and others, “Charter School
Quality and Parental Decision Making with
School Choice,” Working Paper 11252 (Cambridge, Mass.: National
Bureau of Economic Research, 2005).
56. William Howell and Paul Peterson (with Patrick Wolf and
David Campbell), The Education Gap: Vouchers
and Urban Schools (Brookings, 2002).
57. Alan B. Krueger and Pei Zhu, “Another Look at the New York
City Voucher Experiment,” American Be-
havioral Scientist 47, no. 5 (2004): 658–98.
58. Cecilia Elena Rouse, “Private School Vouchers and Student
Achievement: An Evaluation of the Milwaukee
Parental Choice Program,” Quarterly Journal of Economics 113,
no. 2 (1998): 553–602; and John Witte,
“Achievement Effects of the Milwaukee Voucher Program,” mimeo,
University of Wisconsin (1997).
59. Kim K. Metcalf and others, “Evaluation of the Cleveland
Scholarship and Tutoring Program, Summary Re-
port, 1998–2003,” mimeo, Indiana University (2004).
60. Figlio and Rouse, “Do Accountability and Voucher Threats
Improve Low-Performing Schools?” (see
note 51).
61. Janet Quint and others, “The Challenge of Scaling up
Educational Reform: Findings and Lessons from
First Things First, Final Report,” monograph (New York: MDRC,
2005); Jacob and Lefgren, “Remedial
Education and Student Achievement” (see note 50).
U . S . E l e m e n t a r y a n d S e c o n d a r y S c h o o l
s : E q u a l i z i n g O p p o r t u n i t y o r R e p l i c a t i
n g t h e S t a t u s Q u o ?
V O L . 1 6 / N O. 2 / FA L L 2 0 0 6 123
06 5563 Rouse-Barrow.qxp 8/6/2006 4:56 PM Page 123