Presidential Essay - IA de Investigación especializada en el MIT

Ensayo presidencial

Amy Ellen Schwartz

Institute for Education

and Social Policy

New York University

665 Broadway, 8th Floor
Nueva York, Nueva York 10012

amy.schwartz@nyu.edu

MAKING RESEARCH IN

EDUCATION FINANCE

AND POLICY MATTER NOW

INTRODUCCIÓN
Research in education ﬁnance and policy has ﬂour-
ished over the past twenty years as No Child Left
Behind (NCLB) and a wide range of school reform
efforts spurred demand for scientiﬁc evidence identify-
ing “what works.” Research funding has been generous,
buoyed by both favorable economic conditions and the
sense that research will provide solutions to persistent
problems in American schooling. It has been a good
time for education research.

The good news is that we have made signiﬁcant
strides over the past decades. The quality and availability
of data have improved substantially, reﬂecting improve-
ments in administrative data from school districts and
estados (p.ej., Texas, Nueva York, and Florida) and survey
data from the U.S. Department of Education contain-
ing longitudinal data on both students (p.ej., National
Education Longitudinal Study, Early Childhood Longi-
tudinal Studies) and schools and school districts (p.ej.,
the Common Core of Data). Perhaps equally important
has been the development and spread of sophisticated
research methodologies that can be used to identify
causal relationships between educational interventions
and outcomes and, more generally, to disentangle the
causes and consequences of student achievement. Alabama-
though not entirely new, natural experiments, instru-
mental variables, regression discontinuity designs, hier-
archical linear models, and randomized controlled trials
are now ﬁxtures in education policy research.

C(cid:2) 2009 American Education Finance Association

D
oh
w
norte
oh
a
d
mi
d

F
r
oh
metro
h

t
t

pag

:
/
/

d
i
r
mi
C
t
.

metro

i
t
.

mi
d
tu
mi
d
pag
a
r
t
i
C
mi
–
pag
d

F
/

5
1
1
1
6
8
9
1
6
4
mi
d
pag
2
0
0
9
5
1
5
1
0
1
pag
d

b
y
gramo
tu
mi
s
t

oh
norte
0
8
S
mi
pag
mi
metro
b
mi
r
2
0
2
3

MAKING EDUCATION RESEARCH MATTER

Desafortunadamente, our successful use of and enthusiasm for new techniques
and data have not been matched by comparable success in identifying solutions
to pressing problems and resolving continuing disputes about efﬁcacy and
efﬁciency. En efecto, the results have been fairly modest—yielding more insight
into “what doesn’t work” than what does. It is not surprising that policy makers
and educational leaders ﬁnd this harvest disappointing.1 As research funding
from foundations and governments tightens with the economic downturn and
a new presidential administration takes the helm, the time is ripe to reevaluate
and consider how to make research in education policy and ﬁnance matter.

In the end, the key to making education policy research matter is asking
questions that matter—about pressing problems that affect large numbers of
students in a broad range of circumstances—and providing useful answers
and solutions that are feasible, practical, and implementable under realistic
circumstances. Why do some students succeed while others do not? What can
and what should the public sector do about it? These are the fundamental
preguntas.

I see three key implications for making education policy research matter
now. Primero, methodology must follow from the question (rather than vice versa).
This means we must look beyond the methodologically neat and fashionable
and focus on bringing to bear the theoretical and methodological tools neces-
sary and appropriate to answering the questions that matter. This may mean
more descriptive work that measures, documentos, and locates problems. Él
may mean fewer randomized controlled experiments that yield internally valid
estimates of programs implemented in a small number of schools or class-
rooms and more emphasis on analyses of the impact of policies and programs
implemented and practiced by states, school districts, y escuelas. Más,
we need to think about the theoretical underpinnings of our empirical work
and bring theory to bear in both design and interpretation. Understanding the
behavioral responses of parents, profesores, estudiantes, and taxpayers is critical to
drawing the link between impact estimates and policy changes.

Segundo, we need to look outside the traditional boundaries of education
research to understand how nutrition, health care, housing, y otros factores
affect student outcomes. We need to look beyond the school day and the current
academic year to understand how and when the past shapes the present and
how out-of-school time complements or complicates school experiences. Mientras
few would argue that academic outcomes depend only on what happens in
the classroom, during the school day, and during the current academic year,
much of our empirical research proceeds as if it were so. Understanding the

Ver, Por ejemplo, Glod 2008 and Viadero 2009.

D
oh
w
norte
oh
a
d
mi
d

F
r
oh
metro
h

t
t

pag

:
/
/

d
i
r
mi
C
t
.

metro

i
t
.

mi
d
tu
mi
d
pag
a
r
t
i
C
mi
–
pag
d

F
/

5
1
1
1
6
8
9
1
6
4
mi
d
pag
2
0
0
9
5
1
5
1
0
1
pag
d

b
y
gramo
tu
mi
s
t

oh
norte
0
8
S
mi
pag
mi
metro
b
mi
r
2
0
2
3

Amy Ellen Schwar tz

contributions of outside-the-classroom factors, out-of-school time, and prior
year experiences is critical to understanding how schools shape and improve
student outcomes.

Finalmente, we need research on the large, diverse population of urban school
niños, the challenges posed by poverty, immigration, movilidad, carrera, y
etnicidad, and the systemic challenges of large urban school districts. Mientras
there are challenges facing students and schools across the country, those
facing cities and their students seem to me to be particularly compelling. I
elaborate on these below.

METHODOLOGY MATTERS . . . THEORY TOO
Good research provides answers that are useful, informative, y, al final,
bien. Desafortunadamente, the incentives for researchers are poorly aligned with
these objectives. Research projects are driven by considerations about what
is fundable and publishable in prestigious journals. These incentives easily
translate into an emphasis on the clever over the insightful, on the method-
ologically fashionable over the useful, and on quickly executed studies with
straightforward results over sustained studies yielding full, nuanced, robusto
answers. We also need to do better at matching the answer to the question. A
some degree, the recent focus on estimating “effects” has eclipsed concerns
about what the estimated effect means. We can bridge this gap by bringing the-
ory from economics, política, sociology, and psychology to bear both in the
modeling and estimation stages and in the interpretation.

One good example can be found in the research on class size. Over the last
décadas, a substantial body of research has focused on estimating the impact
of class size on student performance. Some of this research uses experimen-
tal methods—most notably the randomized controlled trial of the Tennessee
STAR (Student Teacher Achievement Ratio) experimento. Others use quasi-
experimental methods (such as Angrist and Lavy 1999). Still others proceed
by modeling the production function for education and using econometric
methods to address potential biases.2 In large part, the analyses are designed
to estimate the impact of changing class size on the performance of the stu-
dents in those classes. This is not the same as the impact of changing class sizes
in all schools in a district or all schools in a state. Knowing whether reducing class
size in a subset of schools or classrooms has a positive effect on student perfor-
mance is interesting and important in its own right, but it does not answer the
policy question: what is the impact of a statewide policy to reduce class size?

2. Hanushek 1986 is a classic; see also Hanushek 1999 and Nye, Hedges, and Konstantopoulos 1999.
The Handbook of Research in Education Finance and Policy, edited by Ladd and Fiske (2007), es un
useful reference.

D
oh
w
norte
oh
a
d
mi
d

F
r
oh
metro
h

t
t

pag

:
/
/

d
i
r
mi
C
t
.

metro

i
t
.

mi
d
tu
mi
d
pag
a
r
t
i
C
mi
–
pag
d

F
/

5
1
1
1
6
8
9
1
6
4
mi
d
pag
2
0
0
9
5
1
5
1
0
1
pag
d

b
y
gramo
tu
mi
s
t

oh
norte
0
8
S
mi
pag
mi
metro
b
mi
r
2
0
2
3

MAKING EDUCATION RESEARCH MATTER

Large-scale policy changes involve a host of issues that a randomized,
controlled trial will not address, such as induced demand for teachers and
classrooms. If existing school buildings are at capacity, reducing class size
might reduce school size and induce a demand for more school buildings,
which may be of different quality (or character) than the existing schools, con
unknown effects on student performance. Reducing class size might affect
the composition of schools because it offers more opportunity for segregation
and isolation. Por supuesto, implementing a statewide class size reduction policy
may create new demand for teachers and lead to a redistribution of teach-
ers across schools, with considerable consequences for students and schools.
(The California experience with class size reduction is instructive.) En el
end, the ﬁnding that reducing class size on a micro scale improves academic
performance of the students in those classes is neither necessary nor sufﬁ-
cient evidence for a positive impact of class size reduction on a macro scale.
En cambio, we need to assess the general equilibrium effects—the impacts on
input demand and the mobility of students, taxpayers, and teachers—which
requires an entirely different strategy than replicating an experiment to assess
the generalizability of the results to different populations or settings.3

The point is that we need to carefully align the design and method with
the question and the theory. If the intention is to estimate a local average
treatment effect of an intervention implemented at a small scale—shedding
light on the science of learning and potential directions for school reforms—
then a randomized controlled trial will do well. If the purpose is to uncover
systemic effects of large-scale reform—which I would argue is the critical policy
question—then we need more than a good clean estimate of a program impact.
Disentangling causal impacts requires theoretical and not just methodolog-
ical sophistication. Como ejemplo, consider the estimation of an education pro-
duction function—the workhorse of education research—estimated to identify
the impact of school resources on student outcomes. Típicamente, a test score (o
gain) is linked to variables capturing school resources and student character-
istics (including a lagged score if needed to create a value-added measure).

Imagine that resources are allocated via a formula, which gives money to
schools based on the number and characteristics of their students (poverty,
special needs, limited English proﬁcient (LEP), etc.) and some institutional
características (grade span, building features, etc.). Some of these resources are
allocated by unit (p.ej., number of teachers) and others by dollars (p.ej., libro
allowance). Notice that the production function regression model includes
muchos, if not all, of the factors that drive the differences in resources. El

The work of Tom Nechyba (2000, 2003a, 2003b) provides good examples.

D
oh
w
norte
oh
a
d
mi
d

F
r
oh
metro
h

t
t

pag

:
/
/

d
i
r
mi
C
t
.

metro

i
t
.

mi
d
tu
mi
d
pag
a
r
t
i
C
mi
–
pag
d

F
/

5
1
1
1
6
8
9
1
6
4
mi
d
pag
2
0
0
9
5
1
5
1
0
1
pag
d

b
y
gramo
tu
mi
s
t

oh
norte
0
8
S
mi
pag
mi
metro
b
mi
r
2
0
2
3

Amy Ellen Schwar tz

reasons might be precisely the same as the reason these factors are included in
the allocation formula: they are likely to capture differences in inputs required
to achieve a given output level. The implication is that the estimated coefﬁcients
on resources capture the impact of resources not allocated via this formula—
eso es, not given because of differences in need.4 Put differently, el modelo
identiﬁes the impact of resources using the variation in resources not generated
by variation in need, which may be capricious or, worse, misallocated—say, pendiente
to reliance on inefﬁcient teacher transfer rules or hold-harmless provisions.
Notice that if resources are strictly allocated according to formula (as weighted
student funding advocates propose), there may be no variation left and the
coefﬁcient will not be estimable.

Al mismo tiempo, one might argue that the remaining variation is essen-
tially random and allows us to estimate the impact of random increases in
resources to schools—that this is just the right thing. While this has some ap-
peal, it also means that the estimated coefﬁcient does not provide a meaningful
answer to the key policy question: does strategically increasing resources to
schools improve student performance? This is the question that matters to
policy makers.

Answering this question will require investigating and understanding the
political economy of education as well as policy adoption and implementation.
What caused the “natural experiment”? What determines eligibility cutoffs
used in regression discontinuity designs? Were they set at the point where
marginal beneﬁts are expected to be zero (ensuring that no student that might
beneﬁt from the program is denied services) or where marginal beneﬁts were
thought to exceed marginal costs (consistent with a different cost-beneﬁt anal-
ysis)? What explains the processes used to allocate resources across schools
within districts (or across districts within states)?

Finalmente, many important questions do not, En realidad, involve causal relation-
ships at all. Some are really about “just the facts.” How big is the black-white
test score gap? Is it bigger for boys or girls (or the same)? Has it increased or de-
creased with NCLB? How many children change schools before the end of the
school year or between school years? Is mobility greater among poor students?
Is the current wave of foreclosures affecting all groups of students, or are they
concentrated in particular schools or districts that may be in need of help?
Does school funding increase with increases in students with special needs
(es decir., LEP, poor, learning disabled)? How much? These kinds of measures are

4. Alternativamente, if additional teachers are routinely allocated to schools based upon the number of
estudiantes, weighted, decir, by the percentage of limited English proﬁcient or special education students,
the regression that controls for these is identifying the impact conditional on the variables used to
allocate the resources.

D
oh
w
norte
oh
a
d
mi
d

F
r
oh
metro
h

t
t

pag

:
/
/

d
i
r
mi
C
t
.

metro

i
t
.

mi
d
tu
mi
d
pag
a
r
t
i
C
mi
–
pag
d

F
/

5
1
1
1
6
8
9
1
6
4
mi
d
pag
2
0
0
9
5
1
5
1
0
1
pag
d

b
y
gramo
tu
mi
s
t

oh
norte
0
8
S
mi
pag
mi
metro
b
mi
r
2
0
2
3

MAKING EDUCATION RESEARCH MATTER

critical to good policy making, both because they suggest directions for further
investigación (what explains observed disparities?) and because they indicate how
and where interventions might be targeted and most effective (p.ej., who needs
ayuda?).

BEYOND THE CLASSROOM, THE SCHOOL DAY, AND THE SCHOOL YEAR
Although classroom activities are important, student performance is clearly
shaped by a wide range of nonclassroom or nonschool factors, such as housing,
salud, or nutrition, and out-of-school activities such as after-school programs,
summer camp, libraries, or museums. While few would argue that academic
outcomes depend only on what happens in the classroom, during the school
día, and during the current academic year, much of our empirical research
proceeds as if it were so.

To some extent, this reﬂects limitations in data. Education data sets are un-
derstandably thin on variables capturing the noneducational or out-of-school-
time features of students’ lives. Similarmente, data sets assembled for health
research or housing research are typically thin on data on schools and ed-
ucation. To some extent, sin embargo, the relative scarcity of research in this area
reﬂects the sectoral compartmentalization of government agencies, founda-
ciones, and researchers. Where does research on the impact of housing on
education (or nutrition and schools) “ﬁt”? Who funds it? Who publishes it?

My own research in New York City suggests that pushing these boundaries
is both possible and important. Como ejemplo, in a recent study Schwartz et al.
(2009) examine the academic performance of students living in public hous-
ing in New York City. Assembling the necessary data was time consuming and
difﬁcult and required matching data on student residences to public housing
projects. The results, sin embargo, are intriguing, indicating that students living in
public housing are uniquely disadvantaged. To be speciﬁc, we ﬁnd that while
there are relatively small differences in the resources provided to schools serv-
ing public housing residents, there are signiﬁcant differences in the student
body. More important, children in public housing score lower on standardized
tests than otherwise similar children in their schools. This descriptive study
raises a host of important questions. Are the observed lower scores the result of
conditions in public housing or unobserved characteristics of families living
in public housing? Alternativamente, are they because of differences in the op-
portunities and resources in the neighborhoods surrounding public housing?
What matters? After-school programs? Libraries? Health clinics? Community
centros?

Claramente, health and nutrition matter to student performance. Estudiantes
with poorer health are more likely to miss school, affecting their academic

D
oh
w
norte
oh
a
d
mi
d

F
r
oh
metro
h

t
t

pag

:
/
/

d
i
r
mi
C
t
.

metro

i
t
.

mi
d
tu
mi
d
pag
a
r
t
i
C
mi
–
pag
d

F
/

5
1
1
1
6
8
9
1
6
4
mi
d
pag
2
0
0
9
5
1
5
1
0
1
pag
d

b
y
gramo
tu
mi
s
t

oh
norte
0
8
S
mi
pag
mi
metro
b
mi
r
2
0
2
3

Amy Ellen Schwar tz

actuación. Asthma or obesity may affect social and physical well-being,
with spillover effects on educational achievement. How does the school food
program ﬁt in? While the percent of students eligible for free lunch is routinely
used as a measure of poverty in education research, there is little research that
focuses on the program directly. Does the price matter? Does the menu? El
limited research available suggests that there is much to be learned here.5

One of the implications of excluding out-of-school-time activities and
outside-the-school resources is that doing so hinders our ability to isolate the
impact of school factors on student outcomes. If these activities shape school
resultados, production functions are misspeciﬁed. If the quality or quantity of
these activities is related to school resources or practices, this misspeciﬁcation
will lead to biased estimates of the impacts of these resources or practices.
Returning to the class size example, parents may spend more (or less) tiempo
teaching their children depending on the success of the teaching enterprise
at school. They may hire more (or fewer) tutors, have their children partici-
pate in more (or fewer) after-school programs, or invest in more (or fewer)
computer instructional programs. If we ﬁnd no effect of increasing class size
in high school, we will want to make sure to distinguish between the partial
efecto (which holds other things, such as out-of-school tutoring or test prepara-
ción, constant) and the full effect. Providing good policy guidance may require
knowing both, but it certainly requires knowing which one you have estimated
and the difference between the two.

THE IMPORTANCE OF URBAN SCHOOLS
Although American education needs to be improved across the board, grande
urban areas deserve particular attention. The ﬁrst reason is their size. El
sixty-seven large urban school districts of the Council of the Great City Schools
(CGCS) educate 7.1 million children and employ almost half a million teachers
in over 11,700 escuelas (CGCS 2009). This represents 15 percent of public school
students nationally, 14 percent of teachers nationally, y 12 percent of public
escuelas. According to the National Center for Education Statistics (NCES), en
2005–6 the one hundred largest public school districts educated 11.3 millón
niños (22.7 percent of all public school students) en 16,584 escuelas (16.7
percent of all public schools) employing over 20 percent of all teachers in the
United States (Garofano and Sable 2008).

Segundo, students in urban school districts are more likely to be poor, mi-
nority, foreign-born, and limited English proﬁcient than students elsewhere.
More than one-third of the students attending CGCS districts are African

Ver, Por ejemplo, Figlio and Winicki 2005 or Hinrichs 2009.

D
oh
w
norte
oh
a
d
mi
d

F
r
oh
metro
h

t
t

pag

:
/
/

d
i
r
mi
C
t
.

metro

i
t
.

mi
d
tu
mi
d
pag
a
r
t
i
C
mi
–
pag
d

F
/

5
1
1
1
6
8
9
1
6
4
mi
d
pag
2
0
0
9
5
1
5
1
0
1
pag
d

b
y
gramo
tu
mi
s
t

oh
norte
0
8
S
mi
pag
mi
metro
b
mi
r
2
0
2
3

MAKING EDUCATION RESEARCH MATTER

Americano (36 por ciento) and Hispanic (35 por ciento), and nearly two-thirds are
eligible for free or reduced price lunch (61 por ciento). Más, 17 percent are
English language learners (ELLs), y 13 percent have individualized educa-
tion programs (IEPs). En tono rimbombante, these students represent 32 percent of all
African American students, 26 percent of all Hispanic students, 23 por ciento de
all poor students, 29 percent of all ELLs, y 14 percent of all students with
IEPs.

Finalmente, students in city schools have signiﬁcantly worse educational out-
comes: their National Assessment of Educational Progress scores are lower in
reading and math in both fourth and eighth grades. The NCES reports gradu-
ation rates for these districts well below the national average. Como ejemplo,
in 2005–6, Los Angeles, chicago, Dade County (Miami), and New York City
showed freshman graduation rates below 60 por ciento.

Thus the inadequate performance of American students and the dispari-
ties between blacks and whites or between the poor and the nonpoor (entre
otros) signiﬁcantly reﬂect the inadequacies of urban education. Improving ed-
ucation in the cities is a lynchpin to reducing the national race gap in academic
achievement and improving educational opportunities for the disadvantaged.
Desafortunadamente, the research base on urban education is insufﬁcient to guide
urban education policy. Although many of the issues facing urban schools and
students are universal, some are unique, due in part to the difference in scale.
The policies and practices that might be successful or effective in districts
with a handful of, or even several, schools may be ineffective in large districts,
numbering in the hundreds of schools. Como ejemplo, small high schools
might be appealing in districts with a few thousand students but prohibitive
or unwieldy in large ones; it is impractical to educate 100,000 high school
students in 2,000 high schools of 500 estudiantes.

Large urban districts are often quite diverse. While Tiebout sorting may
work well to deliver relatively homogenous suburbs, cities are diverse. Y-
derstanding the performance of urban students means looking not only at
blacks and whites but also at Hispanics and Asians, at the differences between
native-born and foreign-born students, and at differences within the immi-
grant community—between students from different countries or those who
speak different languages at home.

This challenge also presents opportunities. The size and diversity of urban
school districts offer unique variation that we can exploit to explore the causes
and consequences of differences in academic performance and an opportunity
to understand how and why some schools succeed while others do not. Are
schools that serve white students well also good for black students? Cómo
about Hispanic students? Más, cities offer the possibility of assembling
large longitudinal data sets from administrative data on students, stafﬁng, y

D
oh
w
norte
oh
a
d
mi
d

F
r
oh
metro
h

t
t

pag

:
/
/

d
i
r
mi
C
t
.

metro

i
t
.

mi
d
tu
mi
d
pag
a
r
t
i
C
mi
–
pag
d

F
/

5
1
1
1
6
8
9
1
6
4
mi
d
pag
2
0
0
9
5
1
5
1
0
1
pag
d

b
y
gramo
tu
mi
s
t

oh
norte
0
8
S
mi
pag
mi
metro
b
mi
r
2
0
2
3

Amy Ellen Schwar tz

school characteristics, as well as the important related factors—neighborhoods,
housing, salud, etc..

Notice that understanding urban education will require expanding our
theoretical tool kit. Models from public ﬁnance explain the behavior of school
districts. Labor economics provides models of individual behavior (for teach-
ers, estudiantes, directores, padres). The theory of school behavior and the in-
tradistrict decisions of districts and schools, which might draw on industrial
organización, es, in contrast, quite thin. How should (hacer) districts organize
schools within a large district? Is there an optimal portfolio of schools or an
optimal distribution of resources and/or students across schools?6 What are
the implications of alternative policies for neighborhoods or the city as a whole?
My own work, with Leanna Stiefel and colleagues, on education in New
York City provides an illustration of both what is possible and what needs to be
done.7 The largest school district in the country, Nueva York, educates over
1.1 million students in more than 1,400 schools and is incredibly diverse. Encima
a third of New York’s students are black, a third are Hispanic, and there is a
large immigrant population, including students from a wide range of countries
and speaking many different languages at home. This diversity is also seen in
the wide array of neighborhoods, ethnic enclaves, and distinct communities
spread across the ﬁve boroughs.

Al mismo tiempo, there is broad variation across schools. While some
are quite small (100–200 students), others are quite large (4,000+ estudiantes).
Some are virtually all poor and some have hardly any poor students. Some are
nearly all black or Hispanic, others are predominantly white or Asian, and still
others are fairly integrated. New York City has both award-winning excellent
schools and failing schools identiﬁed by state and federal accountability pro-
grams as “in need of improvement.” Some schools have a long history, y
there are also new schools opening each year. More generally, New York’s “ex-
perimentation” with policy changes and reforms yields ample opportunities
for researchers to learn about what works and what does not.

Equally important, we have assembled an extraordinary data set on edu-
cation in New York City. This includes more than thirteen years of longitudi-
nal data on more than a million students, including information on testing,
sociodemographics, schools attended, and home language, complemented by

6. Rubenstein et al. (2007) begin to examine this by exploring how resources are allocated across

individual schools in large districts.
I have been fortunate to have terriﬁc colleagues and partners in the New York City research.
Although I cannot list all of our New York City projects and authors here, I want to acknowledge the
important work of Hella Bel Hadj Amor, Vicki Been, Luis Chalico, Colin Chellman, Dylan Conger,
Sean Corcoran, Ingrid Ellen, Patrice Iatarola, Brian McCabe, Charles Parekh, Ross Rubenstein,
Ioan Voicu, Meryle Weinstein, Matt Wiswall, and Jeff Zabel.

D
oh
w
norte
oh
a
d
mi
d

F
r
oh
metro
h

t
t

pag

:
/
/

d
i
r
mi
C
t
.

metro

i
t
.

mi
d
tu
mi
d
pag
a
r
t
i
C
mi
–
pag
d

F
/

5
1
1
1
6
8
9
1
6
4
mi
d
pag
2
0
0
9
5
1
5
1
0
1
pag
d

b
y
gramo
tu
mi
s
t

oh
norte
0
8
S
mi
pag
mi
metro
b
mi
r
2
0
2
3

MAKING EDUCATION RESEARCH MATTER

data on schools, housing, property values, neighborhoods, profesores, y más.
Data on elementary and secondary education in New York City public schools
are linked to data on the City University of New York, which educates many
graduates of the public schools.8

en un 2007 estudiar, Leanna Stiefel, Ingrid Ellen, and I examined the distri-
bution of the black-white test score gap across elementary and middle schools
(Stiefel, Schwartz, and Ellen 2007). The results were interesting. Primero, segre-
gation meant that many schools had too few students to have a meaningful
“gap.” Among those with signiﬁcant populations of both groups, we found
considerable variation in the magnitude of the gap across schools. While a
range of student characteristics explained some of the gap, signiﬁcant dispar-
ities remained. Más, the black-white differential estimated using a school
ﬁxed effects model was little changed by using a classroom ﬁxed effects model,
suggesting that disparities are not simply due to within-school sorting and in-
equities across classrooms. Like the research on public housing, this article
points to more questions about the underlying causes of these disparities. Are
differences due to differences in student health, neighborhoods, or treatment
within shared classrooms?

Curiosamente, our companion investigation of the educational outcomes of
immigrant students found that the “nativity gap” favors immigrants. Schwartz
and Stiefel (2006) found that immigrants outperform the native born on state
tests and have higher attendance rates and higher graduation rates relative to
otherwise similar native-born students (eso es, controlling for English proﬁ-
ciencia, poverty, etcétera). Sí, limited English proﬁciency means lower test
puntuaciones, but this is true for both native- and foreign-born students. In current
work examining the impact of age of entry on the high school performance of
immigrants, we ﬁnd evidence that mobility matters to the native born as well
as to the foreign born (see Conger, Schwartz, and Stiefel 2008). Does mobility
explain performance disparities?

Understanding mobility is challenging, en parte, because there are many
different kinds of mobility: across schools, districts, estados, or countries; dentro
and between academic years; school mandated or discretionary; voluntary or
involuntary; Etcétera. Más, longitudinal data on mobile populations are
hard to ﬁnd. Using data on New York City public school students, Schwartz,
Stiefel, and Chalico (2007) found that eighth-grade performance declines with
the number of schools a student has attended. Al mismo tiempo, movilidad
is highest among at-risk populations, including poor, negro, and Hispanic
estudiantes.

See Schwartz and Stiefel 2006 and Stiefel, Schwartz, and Ellen 2007.

D
oh
w
norte
oh
a
d
mi
d

F
r
oh
metro
h

t
t

pag

:
/
/

d
i
r
mi
C
t
.

metro

i
t
.

mi
d
tu
mi
d
pag
a
r
t
i
C
mi
–
pag
d

F
/

5
1
1
1
6
8
9
1
6
4
mi
d
pag
2
0
0
9
5
1
5
1
0
1
pag
d

b
y
gramo
tu
mi
s
t

oh
norte
0
8
S
mi
pag
mi
metro
b
mi
r
2
0
2
3

Amy Ellen Schwar tz

Even more, Rubenstein et al. (2009) found that K–8 schools deliver higher
test score results than other types of schools, reﬂecting, en parte, the fewer
school moves made by their students. While this limited evidence hardly
warrants large-scale reconﬁguration of elementary and middle schools into K–
8s, further exploration of the potential beneﬁts of K–8s and other interventions
to limit mobility are clearly warranted.

FINAL THOUGHTS
While this essay has focused on research and policies aimed at improving
test scores and academic performance, it is clear that this focus is too narrow.
En efecto, the public demand for public education is not conﬁned to a demand for
high test scores or graduation rates. Parents and students (and taxpayers and
empleadores) also care about athletics, social and emotional development, civic
engagement, arts and cultural education—that is, football games, proms, y
banda. Ignoring these hampers our ability to identify policies that will garner
the political and social support needed for success.

Al mismo tiempo, we must take care in communicating research results to
the policy-making community. Too often heroic leaps are made from narrow
results to policy guidance. This ultimately undermines our credibility. Es
perhaps no one’s fault but our own that we hear policy makers and leaders
advocate interventions that we “know work” such as charters or merit pay—
even when the research base is thin and conﬂicting.

In the end, making research in education ﬁnance and policy matter means
asking the important questions and designing studies that provide thought-
lleno, nuanced answers that inform policy. As the mathematician John Tukey
observado: “Far better an approximate answer to the right question, cual es
often vague, than an exact answer to the wrong question, which can always be
made precise” (1962, pag. 13).

REFERENCIAS
Angrist, Joshua D., and Victor Chaim Lavy. 1999. Using Maimonides’ rule to estimate
the effect of class size on scholastic achievement. Revista trimestral de economía 114 (2):
533–75.

Conger, Dylan, Amy Ellen Schwartz, and Leanna Stiefel. 2008. Do immigrants differ
from migrants? Disentangling the impact of mobility on high school completion and
actuación. New York University, Institute for Education and Social Policy Working
Paper No. 08-01.

Council of the Great City Schools (CGCS). 2009. Urban school statistics. Disponible
www.cgcs.org/about/statistics.aspx. Accedido 14 Agosto 2009.

Figlio, David, and Joshua Winicki. 2005. Food for thought: The effects of school
accountability plans on school nutrition. Journal of Public Economics 89: 381–94.

D
oh
w
norte
oh
a
d
mi
d

F
r
oh
metro
h

t
t

pag

:
/
/

d
i
r
mi
C
t
.

metro

i
t
.

mi
d
tu
mi
d
pag
a
r
t
i
C
mi
–
pag
d

F
/

5
1
1
1
6
8
9
1
6
4
mi
d
pag
2
0
0
9
5
1
5
1
0
1
pag
d

b
y
gramo
tu
mi
s
t

oh
norte
0
8
S
mi
pag
mi
metro
b
mi
r
2
0
2
3

MAKING EDUCATION RESEARCH MATTER

Garofano, Antonio, and Jennifer Sable. 2008. Characteristics of the 100 largest public
elementary and secondary school districts in the United States: 2005–06. NCES 2008-
339. Washington, corriente continua: National Center for Education Statistics, A NOSOTROS. Department of
Educación.

Glod, María. 2008. Searching for science to guide good teaching. Washington Post,
28 Abril.

Hanushek, eric. 1986. The economics of schooling: Production and efﬁciency in public
escuelas. Journal of Economic Literature 49 (3): 1141–77.

Hanushek, eric. 1999. Some ﬁndings from an independent investigation of the Ten-
nessee STAR experiment and from other investigations of class size effects. Educativo
Evaluation and Policy Analysis 21: 143–63.

Hinrichs, Peter. 2009. The effects of the national school lunch program on education
and health. Unpublished manuscript, Instituto de Tecnología de Massachusetts.

muchacho, Helen F., and Edward B. Fiske. 2007. Handbook of research in education ﬁnance
and policy. Nueva York: Routledge.

Nechyba, tomás. 2000. Mobility, targeting and private school vouchers. Americano
Economic Review 90 (1): 130–46.

Nechyba, tomás. 2003a. Public school ﬁnance and urban school policy: General
versus partial equilibrium analysis. Brookings-Wharton Papers on Urban Affairs 139–83.

Nechyba, tomás. 2003b. What can be (and what has been) learned from general
equilibrium simulation models of school ﬁnance. National Tax Journal 56 (2): 387–414.

Nye, Barbara, Larry Hedges, and Spyros Konstantopoulos. 1999. The long-term ef-
fects of small classes: A ﬁve-year follow-up of the Tennessee class size experiment.
Evaluación educativa y análisis de políticas 21: 127–42.

Rubenstein, ross, Amy Ellen Schwartz, Leanna Stiefel, and Hella Bel Hadj Amor.
2007. From districts to schools: The distribution of resources across schools in big city
school districts. Revisión de la economía de la educación 26 (5): 532–45.

Rubenstein, ross, Amy Ellen Schwartz, Leanna Stiefel, and Jeffrey Zabel. 2009. Spend-
En g, size and grade span in K–8 schools. Education Finance and Policy 4: 60–88.

Schwartz, Amy Ellen, Brian McCabe, Ingrid Ellen, and Colin Chellman. 2009. Do
public schools disadvantage students living in public housing? New York University,
Institute for Education and Social Policy Working Paper No. 09-08.

Schwartz, Amy Ellen, and Leanna Stiefel. 2006. Is there a nativity gap? New evidence
on the academic performance of immigrant students. Education Finance and Policy 1:
17–49.

Schwartz, Amy Ellen, Leanna Stiefel, and Luis Chalico. 2007. The multiple dimen-
sions of student mobility and the implications for academic performance: Evidencia
from New York City elementary and middle school students. A condition report for
the New York Education Finance Research Consortium. Available www.albany.edu/
edﬁn/documents/schwartzsteifelmobilitypaper.pdf. Accedido 14 Septiembre 2009.

D
oh
w
norte
oh
a
d
mi
d

F
r
oh
metro
h

t
t

pag

:
/
/

d
i
r
mi
C
t
.

metro

i
t
.

mi
d
tu
mi
d
pag
a
r
t
i
C
mi
–
pag
d

F
/

5
1
1
1
6
8
9
1
6
4
mi
d
pag
2
0
0
9
5
1
5
1
0
1
pag
d

b
y
gramo
tu
mi
s
t

oh
norte
0
8
S
mi
pag
mi
metro
b
mi
r
2
0
2
3

Amy Ellen Schwar tz

Stiefel, Leanna, Amy Ellen Schwartz, and Ingrid Ellen. 2007. Disentangling the racial
test score gap: Probing the evidence in a large urban school district. Journal of Policy
Analysis and Management 26: 7–30.

Tukey, John W. 1962. The future of data analysis. Annals of Mathematical Statistics 33
(1): 1–67.

Viadero, Debra. 2009. “No effects” studies raising eyebrows. Education Week 28 (27):
1, 14-15.

D
oh
w
norte
oh
a
d
mi
d

F
r
oh
metro
h

t
t

pag

:
/
/

d
i
r
mi
C
t
.

metro

i
t
.

mi
d
tu
mi
d
pag
a
r
t
i
C
mi
–
pag
d

F
/