IS THERE A NATIVITY GAP? - Ricerca sull'intelligenza artificiale specializzata al MIT

IS THERE A NATIVITY GAP?

NEW EVIDENCE ON THE

ACADEMIC PERFORMANCE OF

IMMIGRANT STUDENTS

Amy Ellen Schwartz

(corresponding author)

Professor of Public Policy,

Education and Economics

Steinhardt School of

Education and Wagner

Graduate School

New York University

E-mail:

amy.schwartz@nyu.edu

Leanna Stiefel

Professor of Economics

Wagner Graduate School

New York University

E-mail:

leanna.stiefel@nyu.edu

Astratto
Public schools across the United States are educating an
increasing number and diversity of immigrant students.
Unfortunately, little is known about their performance
relative to native-born students and the extent to which
the “nativity gap” might be explained by school and de-
mographic characteristics. This article takes a step to-
ward ﬁlling that void using data from New York City
Dove 17 percent of elementary and middle school stu-
dents are immigrants. We explore disparities in perfor-
mance between foreign-born and native-born students
on reading and math tests in three ways—using lev-
els (unadjusted scores), “value-added” scores (adjusted
for prior performance), and an education production
function. While unadjusted levels and value-added mea-
sures often indicate superior performance among immi-
grants, disparities are substantially explained by student
and school characteristics. Further, while the nativity
gap differs for students from different world regions,
disparities are considerably diminished in fully speci-
ﬁed models. We conclude with implications for urban
schools in the United States.

C(cid:1) 2006 American Education Finance Association

D
o
w
N
o
UN
D
e
D

F
R
o
M
H

T
T

:
/
/

D
io
R
e
C
T
.

io
T
.

e
D
tu
e
D
P
UN
R
T
io
C
e
–
P
D

F
/

1
1
1
7
1
6
9
2
9
8
1
e
D
P
2
0
0
6
1
1
1
7
P
D

B
sì
G
tu
e
S
T

o
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3

IS THERE A NATIVITY GAP?

1. INTRODUCTION
In 2000 immigrants exceeded 11 percent of the U.S. population, the highest
level since 1930 and more than twice the twentieth-century low of 4.7 per cento
reached in 1970. In large cities, immigrants are particularly important—an
astonishing 59.5 percent of the population of Miami was foreign-born in 2000,
and a more “modest” 35.9 percent of New Yorkers (Singer 2004). As this wave
of immigration adds ethnic and economic diversity to the adult population in
the United States, the population of school children is changing as well. Public
schools across the nation—and particularly urban public schools—educate an
increasing number and diversity of immigrant students speaking a wide array
of languages and hailing from a broad spectrum of countries and cultures
around the world (Qin-Hilliard, Feinauer, and Quiroz 2001). Allo stesso tempo,
little is known about their performance relative to native-born students. Are
there differences in the performance of immigrant and native-born students
beyond what would be predicted by the differences in the sociodemographic
and educational characteristics of these groups? What role do schools play?
Are public schools failing our immigrant students?

There are at least four reasons to be concerned about the academic perfor-
mance of immigrant students. The ﬁrst stems from a concern about equity.
While there may be little consensus on how best to address the needs of
immigrant students, the notion that students’ education should not depend
upon their country of birth per se seems to have fairly broad appeal. To many,
equity would require that two similar students differing only in their birth
country should be treated equally by their schools. A second reason stems
from a concern about the impact of immigrants on native-born students. Fare
immigrants serve to enhance or dilute the quality of the peer group for native-
born students? A third, related concern is that immigrants may increase the
pressure on schools already burdened by the challenges of new accountability
measures, including, Per esempio, those imposed by the federal No Child Left
Behind Act of 2002. Whether (or to what extent) schools have the capacity to
deal with new or different demands created by inﬂows of immigrant students
is key to understanding the impact of immigration on education. Finalmente, one
might be concerned about the impact immigrants may have on the economy
if they do not succeed in American schools. Will an inadequate education
cause them to become a drag on the economy upon entering the labor force?
Given the large ﬂows of immigrants into U.S. urban areas, the productivity
of the U.S. labor force in the twenty-ﬁrst century will be inﬂuenced by how
well urban school districts succeed in teaching immigrants. Così, understand-
ing the success or failure of immigrant children is important, and the dearth
of research—due, perhaps, to the scarcity of data on immigrant children—is
problematic.

EDUCATION FINANCE AND POLICY

D
o
w
N
o
UN
D
e
D

F
R
o
M
H

T
T

:
/
/

D
io
R
e
C
T
.

io
T
.

e
D
tu
e
D
P
UN
R
T
io
C
e
–
P
D

F
/

1
1
1
7
1
6
9
2
9
8
1
e
D
P
2
0
0
6
1
1
1
7
P
D

B
sì
G
tu
e
S
T

o
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3

Amy Ellen Schwar tz and Leanna Stiefel

The New York City school system provides an excellent opportunity to
study the performance of public school immigrant students, deﬁned here as
foreign-born. Of the 600,000-plus students in New York’s public elementary
and middle schools, almost 17 percent are immigrants; they originate in over
200 countries and speak over 160 languages and dialects. Inoltre, we were
able to obtain detailed administrative data on New York City public school
children, including information on their country of birth.

In this study, Poi, we examine and explore the disparity in performance
between foreign-born and native-born students, which we term the “nativity
gap,” for two cohorts of ﬁfth and eighth graders. We explore the nativity gap
in performance on both reading and math tests in three ways—using level
measures of performance (questo è, comparing raw or unadjusted scores), using
“value-added” measures of performance (adjusting for prior academic perfor-
mance), E, ultimately, using an education production function (controlling
for a range of student and school characteristics). We then explore the extent
to which the coefﬁcients of the production function differ between native and
foreign-born students. Finalmente, we analyze differences between immigrants
from different regions of origin.

To preview the results, we ﬁnd that the difference in unadjusted test scores
varies by grade. While ﬁfth-grade immigrants perform better than the native-
born, suggesting a positive nativity gap, in the eighth grade, immigrant and
native-born students do about equally well. Value-added analyses, Tuttavia,
consistently indicate an immigrant advantage; immigrants gain more over a
school year than native-born students do. Controlling for the full range of
individual and school characteristics in a fully speciﬁed education production
function framework, Tuttavia, reduces the estimated nativity gap. In the end,
the results from the fully speciﬁed model indicate that the immigrant advan-
tage is positive but smaller in magnitude than the unadjusted test scores. A
the same time, our regional analyses suggest that while the unadjusted dis-
parities in test scores between regions can be substantively signiﬁcant, IL
magnitudes of these disparities are diminished in the fully speciﬁed model.

The implication is that much of the difference between the native-born
and the foreign-born and between immigrants from different world regions
derives from differences in their underlying characteristics, such as poverty
and language skills. There is virtually no evidence to suggest that immigrants
are treated inequitably or discriminated against, nor is there evidence to indi-
cate that immigrants form a low-performing peer group for the native-born.
Further, the relative success of immigrants in their early education may allay
concern over their entry into the labor market. To be clear, our analyses
are limited to students entering the American educational system at a rela-
tively young age—while still in elementary or middle school—and results for

D
o
w
N
o
UN
D
e
D

F
R
o
M
H

T
T

:
/
/

D
io
R
e
C
T
.

io
T
.

e
D
tu
e
D
P
UN
R
T
io
C
e
–
P
D

F
/

1
1
1
7
1
6
9
2
9
8
1
e
D
P
2
0
0
6
1
1
1
7
P
D

B
sì
G
tu
e
S
T

o
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3

IS THERE A NATIVITY GAP?

high-school-age entrants may well be different. A study of the experience and
education of later cohorts is a critical next step for researchers. And we examine
disparities over a period of only one year. Over the longer run, the nativity gap
may widen, shrink, or even reverse itself. Questo, pure, is worthy of further study.
In the next section we discuss alternative explanations for a nativity gap,
and in section 3 we review relevant literature on the size of the nativity gap. Noi
present our model and methods in section 4, the data used in the empirical
work in section 5. Results are discussed in section 6, and in section 7 we
present conclusions.

2. WHY SHOULD NATIVITY MATTER?
Prior research on immigrant education has offered a variety of explanations for
differences in performance between immigrant and native-born students in
the United States. Some explanations explore the implications of differences
between immigrants and native-born students, whether driven by selective
migration, settlement patterns in the United States, or underlying differences
between the United States and their home country. Other explanations look to
differences in the experience of immigrants upon arrival in the United States.
To begin, immigrant students may differ, on average, from native-born stu-
dents in their family and home circumstances in ways that inﬂuence academic
performance. Differences in family income, wealth and/or education, for ex-
ample, may well translate into differences in academic achievement, following
the well-established link between these factors and student performance. Or
there may be differences in family composition, such as number and ages
of siblings or parental marital status and history. To the extent that family
composition matters to achievement, differences in composition may drive
differences between immigrant and native-born students. (See Kao [1999] E
Glick and White [2003] for typical examples of quantitative studies that control
for these kinds of factors.)

A second set of explanations focuses on differences in school readiness and
prior academic experiences. While both immigrant and native-born students
may learn a language other than English as a ﬁrst language, limited English
proﬁciency is undoubtedly more common among immigrants than among
native-born students, and this difference may well lead to differences in aca-
demic performance, not only in literacy and language arts but also in other
areas, such as mathematics, in which progress may be impeded by limited
English skills. (See Gandara et al. [2003] and Bleakley and Chin [2004] for evi-
dence of the relationship between English language proﬁciency and school or
market outcomes.) Equally important, immigrants enter U.S. schools having
had a wide range of prior academic experience in their home country; some

EDUCATION FINANCE AND POLICY

D
o
w
N
o
UN
D
e
D

F
R
o
M
H

T
T

:
/
/

D
io
R
e
C
T
.

io
T
.

e
D
tu
e
D
P
UN
R
T
io
C
e
–
P
D

F
/

1
1
1
7
1
6
9
2
9
8
1
e
D
P
2
0
0
6
1
1
1
7
P
D

B
sì
G
tu
e
S
T

o
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3

Amy Ellen Schwar tz and Leanna Stiefel

will be well prepared and others disadvantaged. Per esempio, Russians and
other Eastern Europeans emigrate from countries with highly developed com-
prehensive educational systems, including rigorous math education; and these
students may be better prepared in some areas than their native-born peers. In
other cases, immigrant students such as Mexicans or Dominicans hale from
countries with less rigorous and comprehensive educational systems and are
less well prepared than their native-born peers. Ovviamente, selective migration
may well mean that the preparation of immigrant students differs markedly
from the typical experience in their birth country. That said, the important
insight is that differences in prior academic experience means the nativity gap
may be positive or negative, favoring immigrants or the native-born.

A third explanation offered for differences between immigrants and native-
born students looks to attitudinal differences between these populations. A
put it simply, some researchers have cited the positive attitude of immigrants
toward education as well as the support and encouragement that immigrant
parents provide their children as reasons that immigrant children may succeed
particularly well in the United States (Waters 1999).

In a somewhat different vein, a fourth set of explanations looks to differ-
ences in school and classroom experiences between immigrants and native-
born students. Residential location patterns may lead immigrants to attend
different schools than native-born students; and whether driven by teacher
location preferences, the idiosyncrasies of resource allocation formulas, IL
schooling preferences of parents, differences in political power, or discrim-
ination per se, the implication is that immigrants’ schools may have differ-
ent resources and teachers. (Note, Tuttavia, that Schwartz and Stiefel [2004]
ﬁnd few differences in resources between the schools attended by immigrant
and native-born students in New York City elementary schools beyond those
predicted by differences in student educational traits.) Differences in schools
attended may mean that immigrants are exposed to different peers than native-
born students, perhaps to more immigrants or more students with limited En-
glish proﬁciency. (See Ellen et al. [2002] for some evidence that immigrants
are only slightly segregated in New York City schools.) Ovviamente, there may be
differences within schools and even within classrooms as well—due to ability
grouping, differences in preferences, or language skills, Dire, or even driven by
the teachers’ and other school personnel’s attitudes toward and expectations
of immigrants. As before, whether these favor or harm immigrant students
cannot be predicted theoretically but must be determined empirically.

Finalmente, differences may be driven by the legal status of immigrants and
their parents. While foreign-born students may be citizens and children of
NOI. citizens, Essi (and/or their parents) may be green card holders, tempo-
rary residents, or of uncertain legal status. While the impact of these legal

D
o
w
N
o
UN
D
e
D

F
R
o
M
H

T
T

:
/
/

D
io
R
e
C
T
.

io
T
.

e
D
tu
e
D
P
UN
R
T
io
C
e
–
P
D

F
/

1
1
1
7
1
6
9
2
9
8
1
e
D
P
2
0
0
6
1
1
1
7
P
D

B
sì
G
tu
e
S
T

o
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3

IS THERE A NATIVITY GAP?

status differences may not be direct, they may indirectly inﬂuence perfor-
mance through mobility, incomes, access to supplementary services, or other
pathways.

Notice that all of these pathways suggest that the immigrant experience
is unlikely to be singular. Invece, there will be considerable variation across
immigrants. Infatti, sociological research suggests that the experience of im-
migrants and their assimilation over time can follow very different paths de-
pending on their own characteristics and their reception in this country, among
others factors.1 As an example, some have argued that political refugees from
Cuba (pre-1980) or Vietnam, Per esempio, were welcomed and did well, while
immigrants who came primarily in pursuit of low-skilled jobs, such as Mexi-
cans and Haitians, did worse over time. (See Portes and MacLeod [1996].) Questo
path has also been documented in ethnographic research and interacts with
the race of the immigrant; black students have especially pronounced declin-
ing paths. (See Waters’s 1999 study of Caribbean immigrants in Brooklyn,
New York.) Further, evidence from the 2000 census indicates that there are
substantial differences in the population of foreign-born New Yorkers across
countries of origin. As an example, New York City residents born in Korea and
Japan, the Philippines, the Indian subcontinent, and the former Soviet Union
have college graduation rates and earnings that exceed those of native-born
New Yorkers. In contrasto, New Yorkers born in Mexico, the Dominican Re-
public, China, and Latin America have graduation rates and earnings that fall
below that of native-born New Yorkers. (See Rosen, Wieler, and Periera [2005]
for more on New York City immigrants.)

To summarize, there are many reasons that the academic performance
of immigrant and native-born students may differ. Some involve differences
in the students and/or their family and neighborhood context; some involve
differences in schools; some vary with time and age while others are time-
invariant. These explanations for differences may imply differences in the
mean level of performance of immigrants, ceteris paribus, or in their respon-
siveness to school inputs, family background, and so on. As described below,
we operationalize this notion in our empirical work by examining the extent
to which disparities persist, even after controlling for differences in student

Sociologists identify three assimilation paths distinguished both by their trajectories toward or
away from native-born performance and their application to ﬁrst, second (native-born child with at
least one foreign-born parent), and third (or native-born with native-born parents) generations. In
straight-line assimilation, the ﬁrst generation performs worse than the native-born, but subsequent
generations do as well or better. In optimistic assimilation, the ﬁrst generation performs better
due to parental determination or optimism. Segmented assimilation describes the differences in
paths that reﬂect the context in which the immigrant group is received. (For more information see
Suarez-Orozco and Suarez-Orozco [1995] or Hirschman [2001].)

EDUCATION FINANCE AND POLICY

D
o
w
N
o
UN
D
e
D

F
R
o
M
H

T
T

:
/
/

D
io
R
e
C
T
.

io
T
.

e
D
tu
e
D
P
UN
R
T
io
C
e
–
P
D

F
/

1
1
1
7
1
6
9
2
9
8
1
e
D
P
2
0
0
6
1
1
1
7
P
D

B
sì
G
tu
e
S
T

o
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3

Amy Ellen Schwar tz and Leanna Stiefel

variables, and exploring whether the coefﬁcients in the regression equations
differ between the native and the foreign-born.

To be more speciﬁc, we include measures of student prior academic per-
formance, poverty, race, genere, age, time in system, language ability, E
learning disabilities, and control for school characteristics using school ﬁxed
effects. We are not, Tuttavia, able to include measures of family context, come
as parental education, number of parents or others in household, or attitudes
toward school, which have been found to be important in much previous re-
search, suggesting that even our fullest speciﬁcations are incomplete, E, A
the extent that the omitted factors vary with nativity, some nativity gap should
emerge. Further, these limitations in the data mean that we are not able to
disentangle the speciﬁc importance of each of these explanations. Invece, Questo
work takes a step toward that goal, laying the foundation for further work on
this subject.

3. LITERATURE REVIEW
In sharp contrast to the large volume of research examining the gap in per-
formance between racial/ethnic groups in the United States, there is scant
quantitative research examining disparities between immigrants and native-
born students to complement the rich qualitative, ethnographic studies of
immigrants from speciﬁc regions or countries, a small number of which are
cited above. Così, in addition to looking at the literature on immigrants per
se, our work has also been informed by the far larger research literature that
has examined the gap in performance between racial/ethnic groups using the
type of data that are available for this study.

As shown in Table 1, there are seven recent studies of immigrant student
academic performance that use individual-level data to examine the perfor-
mance of immigrant students at the K–12 level.2 All use samples of adequate
to large size (ranging from 1,225 A 207,609 individual students). They mostly
focus on the late elementary and early high school grades (fourth to eighth);
only two (4, 6) use students as old as tenth and twelfth graders. All but two stud-
ies (4, 5) analyze the determinants of scores themselves rather than added value
in scores (or how those scores change over time). Additionally, all control in
some way for race/ethnicity, socioeconomic status, and language background,
while some add more variables on schools or, in the case of one (5), include
district or school ﬁxed effects.

The ﬁndings are mixed. Among those examining scores in levels, ﬁrst-
generation (foreign-born) students often outperform native-born students

2. These seven studies are all we have located that use individual data and focus speciﬁcally on the

differences between immigrants and native-born K–12 students.

D
o
w
N
o
UN
D
e
D

F
R
o
M
H

T
T

:
/
/

D
io
R
e
C
T
.

io
T
.

e
D
tu
e
D
P
UN
R
T
io
C
e
–
P
D

F
/

1
1
1
7
1
6
9
2
9
8
1
e
D
P
2
0
0
6
1
1
1
7
P
D

B
sì
G
tu
e
S
T

o
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3

2
4

E
D
U
C
UN
T
O
N
F
IO
N
UN
N
C
E

UN
N
D
P
O
l
IO
C
Y

T
H
E
R
E

N
UN
T
IO

IO
T
Y

G
UN
P
?

Tavolo 1 Summary of Relevant Quantitative Research on Immigrant Performance

Authors/Date

Data

1. Portes and
MacLeod
(1996)

2. Kao

(1999)

Children of

Immigrants; Miami,
Ft. Lauderdale, San
Diego, 1996

5,266 students

NELS-88,a 1988

24,599 original sample

3. Kao and
Tienda
(1995)

NELS-88, 1988

24,599 original sample

Grades

8, 9

Measures

Math, reading

Stanford 9 Ach

Levels

No z-scores

GPA

Math and reading

Levels

No z-scores

Math

Levels

No z-scores

Controls

Findings

Socioeconomic, Race,
Educational, School
composition

Second-generation Mexican and Haitians
outperform other immigrants; second-
generation Vietnamese do better in
math; pre-1980 Cubans do better;
post-1980 Cubans do no differently.

Socioeconomic, Race,

Generally, by race, ﬁrst- E

Psicologico, Language,
Education progress and
experience

second-generation immigrants do better
than same race native-born and as well
as or better than white native-born.

Socioeconomic, Race,

Language

Both ﬁrst- and second-generation

immigrants outperform native-born.

Within race/ethnic groups, only ﬁrst-
generation immigrants uniformly
outperform native-born.

4. Glick and
White
(2003)

HSBb 1980, 1990

10, 12

Dropout rates

Socioeconomic, Race,

In levels, 1980s immigrants perform worse

13,152 students

NELS-88, 1990, 1992

16,376 students

Math and reading

Levels and changes

z-scores

Language

than native-born (0 to–.56 sd); In
1990S, they perform better (0 A 0.31
sd).

In changes, immigrants about on par with
native-born. Dropout rates same for
immigrants and native-born.

D
o
w
N
o
UN
D
e
D

F
R
o
M
H

T
T

:
/
/

D
io
R
e
C
T
.

io
T
.

e
D
tu
e
D
P
UN
R
T
io
C
e
–
P
D

F
/

1
1
1
7
1
6
9
2
9
8
1
e
D
P
2
0
0
6
1
1
1
7
P
D

B
sì
G
tu
e
S
T

o
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3

Tavolo 1 Continued

5. Berg and
Kain
(2003)

Texas School Micro
Panel, 1999

207,609 students

6. Demie
(2001)

Three samples of

London students 1998

Primary and
secondary

2340, 2267, 1225

students

Math

Changes

No z-scores

Average of scores on English,
math, and science national
exams

Levels

No z-scores

Socioeconomic, Race, School
composition or ﬁxed effects

All immigrants perform better than
native-born. Hispanic immigrants
perform better than Hispanic
native-born, but Asians do about the
same.

Poverty, Ethnicity, English skills

Indian, Vietnamese, and Chinese students

perform better than
English/Scottish/Welsh, African;
Caribbean and Portuguese perform
worse.

7. Schnepf
(2004)

Four international

4, 8, E

Math and reading test scores

surveys, ten countries

15-year olds

TIMSS, PIRLS, PISAc

1995 A 2001

Samples from 1,314
A 8,115 students

Levels

No z-scores

Socioeconomic, Home
lingua, Immigrant
segregation at school

In six countries, immigrants and

native-born perform no differently; In
three countries (France, Netherlands,
and Germany), immigrants perform
worse.

Notes: aNational Educational Longitudinal Study, 1988, National Center for Edu-
cation Statistics, NOI. Department of Education. bHigh School and Beyond, Na-
tional Center for Education Statistics, NOI. Department of Education. cTrends
International Study Center
in International Mathematics and Science Study,

at Boston College, http://timss.bc.edu/timss1999.html; Progress in Interna-
tional Reading Literacy Study,
International Study Center at Boston College,
http://timss.bc.edu/pirls2001.html; OECD Programme for International Student
Assessment, Paris, France, http://www.pisa.oecd.org/.

UN
M
sì

l
l

e
N

S
C
H
w
UN
R
T
z

UN
N
D

l
e
UN
N
N
UN

S
T
io

e
F
e

2
5

D
o
w
N
o
UN
D
e
D

F
R
o
M
H

T
T

:
/
/

D
io
R
e
C
T
.

io
T
.

e
D
tu
e
D
P
UN
R
T
io
C
e
–
P
D

F
/

1
1
1
7
1
6
9
2
9
8
1
e
D
P
2
0
0
6
1
1
1
7
P
D

B
sì
G
tu
e
S
T

o
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3

IS THERE A NATIVITY GAP?

(1, 2, 3), although distinguishing students by country of origin (1, 6, 7) yields
mixed results—immigrants from only some regions or countries show supe-
rior performance. The authors attribute these mixed results to reception of
immigrants in their destination country—welcomed due to political asylum
versus tolerated in order to ﬁll low-skilled jobs, Per esempio (the latter perform
less well). In some studies, race plays a critical role: black students, in particu-
lar, are more likely to show no difference from native-born black students, at
least in the second generation. The two studies that take a value-added perspec-
tive by controlling for previous test scores (4, 5) show either no difference in
performance between foreign-born and native-born or superior performance
of foreign-born for only one race (Hispanic, 5).

In summary, the evidence across studies on the whether immigrants per-
form better or worse than native-born students mostly ﬁnds superior perfor-
mance by immigrants in the ﬁrst generation. Analyses distinguishing the ori-
gin country or region of the world yield mixed results and thus a more nuanced
version of immigrant performance. Few studies use value-added measures of
performance (changes in scores), but when they do, the positive effect of im-
migrant status is often eliminated. Taken together, these studies complement
the large body of qualitative research, but their data sets are typically limited in
size and in the range of origin countries and regions represented, often only
a small number. Further, the treatment of previous academic performance is
typically limited.

While the quantitative research on the nativity gap is small, the large liter-
ature on the racial gap in test scores provides valuable insights and a useful
perspective. In recent years, a signiﬁcant body of work has examined the ex-
tent to which differences in socioeconomic background and school character-
istics explain the magnitudes and trajectories of gaps in performance between
racial/ethnic groups in the United States; studies in this area typically focus
on the black-white test score gap. As an example, Hedges and Nowell (1998,
1999), using data from six large nationally representative surveys conducted
between 1965 E 1992, ﬁnd that the .7 A 1.0 standard deviation black-white
twelfth-grade test score gap is reduced by less than half when adjustments are
made for socioeconomic factors. Cook and Evans (2000), using National As-
sessment of Educational Progress (NAEP) data on thirteen-year-olds between
1970 E 1988, explore the reduction in the black-white test score gap (cioè., IL
difference in the percentage correct dropped from 17.0 A 9.6 in reading and
17.6 A 11.6 in math) due to family and school factors.3 They ﬁnd that 25 per cento
of these changes can be attributed to shifts in family and school characteristics,

To give a sense of magnitudes, the average percentage correct for whites was 63.7 in reading in
1988.

EDUCATION FINANCE AND POLICY

D
o
w
N
o
UN
D
e
D

F
R
o
M
H

T
T

:
/
/

D
io
R
e
C
T
.

io
T
.

e
D
tu
e
D
P
UN
R
T
io
C
e
–
P
D

F
/

1
1
1
7
1
6
9
2
9
8
1
e
D
P
2
0
0
6
1
1
1
7
P
D

B
sì
G
tu
e
S
T

o
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3

Amy Ellen Schwar tz and Leanna Stiefel

while 75 percent can be attributed to reductions within schools. Finalmente, in un
more recent paper, using the kindergarten through ﬁrst grade cohort from
the Early Childhood Longitudinal Study (ECLS), Fryer and Levitt (2004) ﬁnd
that the black-white test score gap disappears for children entering kinder-
garten when socioeconomic status and several other background factors are
controlled.4 Between the beginning of kindergarten and the end of ﬁrst grade,
Tuttavia, the adjusted gap increases to 0.2 standard deviations. Inoltre, In
an extension using additional years of ECLS, Fryer and Levitt (2005) ﬁnd that
the black-white gap increases by 0.1 standard deviations each year between the
beginning of kindergarten and the end of third grade.

These studies of racial test score gaps are similar in spirit to the ones on
the nativity gap in that they attempt to explain the disparity in performance by
compositional differences in student socioeconomic or educational character-
istics or by differences in schools. They differ in that they generally ﬁnd that
large gaps persist even though controls for student and school differences are
systematically used. Our study builds on these two literatures, applying the
tools developed to explore the race gap to study the nativity gap. We begin with
an analysis of raw scores for orientation. We then turn to value-added anal-
yses, since ascertaining how student and school resources affect the ﬂow of
achievement is most relevant for education policy. Thus our study contributes
to the literature by exploring value-added models of academic performance,
analyzing the extent to which models vary between immigrants and native-
born students by examining the interaction between immigrant status and
other individual characteristics and by exploring the performance of students
from twelve world regions, spanning the globe.

4. MODEL AND METHODS
We estimate three sets of regression models. The ﬁrst set estimates the na-
tivity gap between foreign-born and native-born students; the second explores
differences in the coefﬁcients of the production function between foreign and
native-born students; and the third examines differences in the nativity gap
across immigrants from different world regions. All equations are estimated
using ordinary least squares with robust standard errors corrected for within-
school clustering.5

The centerpiece of our empirical work is a regression model relating stu-
dent performance on standardized tests to student background traits, come
as race, genere, and age; education characteristics, such as language abilities,

4. Reardon (2003), using the same ECLS data set, ﬁnds similar results.
5. Alternative speciﬁcations estimated using instrumental variables to control for the potential endo-

geneity of school resources yielded similar results and are available from the authors.

D
o
w
N
o
UN
D
e
D

F
R
o
M
H

T
T

:
/
/

D
io
R
e
C
T
.

io
T
.

e
D
tu
e
D
P
UN
R
T
io
C
e
–
P
D

F
/

1
1
1
7
1
6
9
2
9
8
1
e
D
P
2
0
0
6
1
1
1
7
P
D

B
sì
G
tu
e
S
T

o
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3

IS THERE A NATIVITY GAP?

learning disabilities, time of arrival into the school system, and prior test score
as a proxy for all prior education performance; and school ﬁxed effects to
capture school differences in resources or peers or any other school speciﬁc
variables. Finalmente, we include a dummy variable that captures differences in
performance between the foreign-born and the native-born, whether due to
unmeasured characteristics such as parents’ education and attitudes, within
school differences in treatment by teachers or others, or the context of recep-
zione, eccetera., as described above. To be speciﬁc, the resulting education production
function model is:

Testijt = β0 + β1Foreigni

+ β2Testi,j,t−1 + β3Sociodemijt

+β4Educationijt + β5Cohortijt + β6Schoolj + (cid:2)ijt

(1)

where i, j, and t index student, school, and year, rispettivamente; italics represent
vettori; Test is the student’s normalized score on a citywide math or reading
test (and its lagged value Testi, j,t–1); Foreign is an indicator that takes a value
of one if the student is born in a country outside the United States; Sociodem
is a vector of variables capturing the student’s poverty status (measured by
eligibility for free or reduced price lunch), genere, age, and race; Education
is a vector of variables capturing the student’s educational characteristics,
including English language abilities and participation in part-time special
education programs; Cohort represents a set of dummy variables indicating
the year of admission to the New York City public schools (with “Admission
Cohort 1993” indicating ﬁrst admission ﬁve or more years ago); and School is
a set of school ﬁxed effects.6

Our education variables include an unusually rich speciﬁcation of vari-
ables relating to experience with and proﬁciency in the English language. One
dummy variable indicates whether a language other than English is the primary
language spoken at home. A second variable captures whether the student was
given the Language Assessment Battery (LAB), a test for English language
proﬁciency (Took LAB). A third variable is the score the student earned on the
LAB, if taken, and a ﬁnal variable indicates whether the student scored at or be-
low the fortieth percentile, the cutoff score determining eligibility for services
to address limited English proﬁciency (LEP). Taken together, these provide a
more nuanced view of the relationship between performance and language
than the more usual speciﬁcations in the literature, which include only a LEP
indicator. Così, the coefﬁcient on home language, Per esempio, captures the

6. For more on education production functions see Hanushek (1986), or more recently McEwan
(2003) or Todd and Wolpin (2003) for a good presentation of the theory of education production
functions.

EDUCATION FINANCE AND POLICY

D
o
w
N
o
UN
D
e
D

F
R
o
M
H

T
T

:
/
/

D
io
R
e
C
T
.

io
T
.

e
D
tu
e
D
P
UN
R
T
io
C
e
–
P
D

F
/

1
1
1
7
1
6
9
2
9
8
1
e
D
P
2
0
0
6
1
1
1
7
P
D

B
sì
G
tu
e
S
T

o
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3

Amy Ellen Schwar tz and Leanna Stiefel

difference in performance between children from English-speaking homes
and non-English-speaking homes, controlling for the child’s measured English
proﬁciency. Some caution is warranted, Tuttavia, in interpreting these coef-
ﬁcients because of the correlation between language skills and unobserved
background and family characteristics.

Notice that the three testing variables offer a particularly interesting in-
terpretation: the coefﬁcient on Took LAB indicates the difference between
students who do and do not take the LAB; the LAB score coefﬁcient indicates
the way in which performance on the standardized test varies with English
language proﬁciency; and the coefﬁcient on LEP (LAB less than 40) indicates
how much performance is higher (or lower) for students who are LEP-eligible,
compared to otherwise similar students. To put it differently, Took LAB indi-
cates whether there is a discontinuity in the relationship between performance
in reading and the LAB score at the point of LEP eligibility. Così, it provides
a regression discontinuity estimate of the impact of LEP eligibility on perfor-
mance.7

Finalmente, the inclusion of the admission cohort variables is unusual but
important in this context for three reasons: (1) there may be unobserved differ-
ences in cohorts, Per esempio, due to differences in U.S. immigration policies
or international conditions;8 (2) performance may be inﬂuenced by the amount
of time a student has had to acclimate or adjust to new conditions; E (3) per-
formance may be shaped by the grade of entry.9 Since these factors may also
be important for native-born children, we include them as a control variable
for all students.10

Before estimating equation (1), we present results of a parsimonious test
score regression including only the foreign-born dummy as a regressor. Here,
β1 measures the mean difference in performance between foreign-born and
native-born students. Our seond speciﬁcation introduces two controls for prior
performance: a dummy variable indicating whether a lagged test score was
available and the lagged test score (which takes a value of zero if there is no
prior test score). In this model, β1 captures the disparity between immigrant
and native-born students in the value added to their scores between years.

See Jacob (2004) for more on regression discontinuity designs and an application to special educa-
tional programs or Matsudaira (2004) for an analysis of bilingual education and ESL programs.
Per esempio, NOI. immigration policy may be focused on refugees from particular countries (per esempio.,
Vietnamese or Haitians) or on family uniﬁcation.

9. As an example, a student entering in second grade is virtually certain to be a “new kid” among

peers, most of whom entered at a common articulation grade, typically kindergarten.

10. Of course the native-born prior educational experiences were most likely obtained in U.S. schools,
while the foreign-born may have prior education elsewhere in the United States or in their birth
country or other locations. These dummies capture the impacts of any differences in cohorts
unmeasured by other model variables.

D
o
w
N
o
UN
D
e
D

F
R
o
M
H

T
T

:
/
/

D
io
R
e
C
T
.

io
T
.

e
D
tu
e
D
P
UN
R
T
io
C
e
–
P
D

F
/

1
1
1
7
1
6
9
2
9
8
1
e
D
P
2
0
0
6
1
1
1
7
P
D

B
sì
G
tu
e
S
T

o
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3

IS THERE A NATIVITY GAP?

Notice that this speciﬁcation is nonstandard. While many studies use samples
limited to students with prior test score data, this seems inappropriate here
because recent immigrants would be disproportionately represented among
those excluded on this basis.11 Our ﬁnal speciﬁcation follows equation (1),
adding student characteristics and school ﬁxed effects to the independent
variables.

Our second set of analyses considers whether the production function
differs for foreign-born and native-born students. Do the coefﬁcients suggest
differences in the impact of language proﬁciency, Dire, or poverty? To do so, we
estimate equation (1) with a complete set of interactions between the regressors
and the foreign-born dummy:

Testijt = β0 + β1Foreigni

+ β2Testi,j,t−1 + β3Sociodemijt

+ β4Educationijt + β5Cohortijt + β6Schoolj
+ γ2Testi,j,t−1Foreigni
+ γ4EducationijtForeigni
+ (cid:2)ijt
+ γ6SchooljForeigni

+ γ3SociodemijtForeigni
+ γ5CohortijtForeigni

(2)

Here, Poi, the γ ’s capture the differences between the two groups.

Our third set of analyses investigates differences in the performance of
immigrants from different world regions, replacing the single foreign-born
dummy in equation (1) with a series of region dummies (Region) indicating
the student’s birth region.12

Testijt = β0 + β1Regioni

+ β2Testi,j,t−1 + β3Sociodemijt

+β4Educationijt + β5Cohortijt + β6Schoolj + (cid:2)ijt

(3)

We follow the same procedure as outlined earlier: ﬁrst, estimating a parsi-
monious model, resulting in mean differences in performance across regions;
second, controlling for prior performance, yielding estimates of the differences

11. We include only students with a current-year test scores. In the 1997–98 ﬁfth grade, we exclude
roughly 2 percent of the native-born and one quarter of the foreign-born, virtually all of whom
have a home language other than English, roughly 80 percent are LEP and recent immigrants,
pointing to the importance of controlling for language proﬁciency and recentness of immigration,
as we have done. The excluded foreign-born may well be higher performing—they are better-off,
younger, and more likely to be Asian and white than the excluded native-born. These variables are
included in our regressions. Alternatively, the potential selection bias could have been addressed by
a Heckman-style selection correction, if we were able to identify variables that predict participation
but not the test score, which we were unable to do. Finalmente, estimating a speciﬁcation with the
interaction of the missing ﬂag and foreign-born yielded insigniﬁcant coefﬁcients.

12. The appendix provides countries by region. Schwartz, Stiefel, and Conger (2002) have more on

regional classiﬁcations.

EDUCATION FINANCE AND POLICY

D
o
w
N
o
UN
D
e
D

F
R
o
M
H

T
T

:
/
/

D
io
R
e
C
T
.

io
T
.

e
D
tu
e
D
P
UN
R
T
io
C
e
–
P
D

F
/

1
1
1
7
1
6
9
2
9
8
1
e
D
P
2
0
0
6
1
1
1
7
P
D

B
sì
G
tu
e
S
T

o
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3

Amy Ellen Schwar tz and Leanna Stiefel

in value added; E, ﬁnally, including a full set of student factors and school
ﬁxed effects.

5. EMPIRICAL PRELIMINARIES: DATA SOURCES, DEFINITIONS, AND
STYLIZED FACTS
We use individual-level data on ﬁfth and eighth graders in New York City public
schools in 1997–98 and 2000–1 for whom standardized reading or math exam
data were available, excluding students in full-time special education, lacking
exam data, or having a missing or unknown birthplace.13

Student performance is measured citywide in reading (CTB/McGraw-Hill
Test of Basic Skills or New York State English Language Assessment) E
mathematics (California Achievement Test [CAT] or New York State Math
Assessment). To facilitate the comparison of test scores across grades and
years, we convert them to z-scores.14

As shown in Table 2, samples are large, ranging from 57,152 A 72,509, E
immigrants often outperform the native-born—consistently in the ﬁfth grade
and never lower in the eighth grade. The value-added analyses are even more
consistent: in all cases, immigrants earn higher scores, roughly on the order
of one-tenth of a standard deviation and ranging from a low of .037 to a high
Di .148 standard deviations. While these differences are modest, if continued
over many grades, substantial gaps would accumulate, favoring immigrants.
Notice, Tuttavia, that there are many differences between the native- E
the foreign-born students. As shown in Table 3, approximately 14 percent of the
roughly 65,000 ﬁfth graders and one-ﬁfth of the nearly 57,500 eighth graders
in our 1997–98 samples are foreign-born.15 Immigrants are more likely to
be poor. They are disproportionately Asian and are less likely to be black or
Hispanic; Tuttavia, there is little difference in the percentage of whites. Not
surprisingly, immigrants are far more likely to come from homes in which the
language spoken is other than English; more likely to take the LAB; more likely
to score lower on the LAB if they take it; and more likely to score below the
LEP cutoff point. Di conseguenza, while nearly all native-born students have taken
reading and math tests in the previous academic year (roughly 95 per cento),
a smaller percentage of the foreign-born have taken these standardized tests

13. Data were generously provided by the New York City Department of Education.
14. To calculate z-scores, one subtracts the mean for the grade and year and divides by the standard
deviation. Using z-scores facilitates comparisons across tests and years; equations are estimated
separately.

15. Any student reporting a country of birth other than the United States or its territories is considered
foreign-born. Così, some (exceedingly small) number of students termed foreign-born here may
have been born abroad to U.S. citizens. Further, native-born students include those born in Puerto
Rico or other U.S. territories or in the United States to foreign-born parents. Contact authors for
information on the 2000–1 sample.

D
o
w
N
o
UN
D
e
D

F
R
o
M
H

T
T

:
/
/

D
io
R
e
C
T
.

io
T
.

e
D
tu
e
D
P
UN
R
T
io
C
e
–
P
D

F
/

1
1
1
7
1
6
9
2
9
8
1
e
D
P
2
0
0
6
1
1
1
7
P
D

B
sì
G
tu
e
S
T

o
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3

IS THERE A NATIVITY GAP?

Tavolo 2 Regression Coefﬁcients, Mean Difference in Level and Value-Added Reading
and Math Performance, Foreign-Born and Native-Born, 1997–98 and 2000–1

Foreign-born
coefﬁcient:

Reading

Fifth grade

LEVEL

VALUE ADDED

1997–98

2000–1

1997–98

2000–1

0.122∗∗∗
(0.019)

0.083∗∗∗
(0.018)

0.126∗∗∗
(0.010)

0.089∗∗∗
(0.010)

Eighth Grade

64,971

−0.004
(0.024)

71,141

0.014
(0.027)

64,971

71,141

0.037∗∗∗
(0.010)

0.058∗∗∗
(0.013)

Math

Fifth grade

Eighth grade

57,465

57,152

57,465

57,152

0.061∗∗∗
(0.022)

0.115∗∗∗
(0.021)

0.105∗∗∗
(0.010)

0.108∗∗∗
(0.012)

66,629

−0.029
(0.028)

72,509

66,629

72,509

0.099∗∗∗
(0.010)

0.062∗∗∗
(0.012)

0.148∗∗∗
(0.013)

59,749

59,024

59,749

59,024

Notes: aThe dependent variable is test score standardized to mean of 0 and a
standard deviation of 1. bValue-added regressions include a lagged test score
and a ﬂag indicating lagged test score was not missing. cRobust standard errors,
adjusted for within-school clustering, in parentheses.

∗signiﬁcant at 10%, ∗∗signiﬁcant at 5%; ∗∗∗signiﬁcant at 1%.

(somewhat over four-ﬁfths). Notice, Tuttavia, that more than one third of the
native-born students hail from homes in which a language other than English
is spoken, and a good percentage are LEP as well.16

In addition to demographic differences, the foreign-born differ from the
native-born in their tenure in the New York City public schools. Native-born
students on average have attended New York City schools for a longer period.
By the ﬁfth grade, native-born students average nearly ﬁve years in the public
schools, which is consistent with kindergarten entry. The average for foreign-
born students, on the other hand, is less than four years, consistent with
entry in the ﬁrst grade. By the eighth grade, the difference has widened:
native-born students average 7.7 years in the New York City public schools,
again suggesting the dominance of kindergarten entry; while foreign-born
students have an average tenure of only 5.2 years, reﬂecting entry throughout
the elementary and middle school years.

È interessante notare, there are differences in the schools attended by foreign-
born and native-born students (not shown), which will be captured in our

16. Notice that we measure whether students are LEP, not whether they received services or what kind

of services they received.

EDUCATION FINANCE AND POLICY

D
o
w
N
o
UN
D
e
D

F
R
o
M
H

T
T

:
/
/

D
io
R
e
C
T
.

io
T
.

e
D
tu
e
D
P
UN
R
T
io
C
e
–
P
D

F
/

1
1
1
7
1
6
9
2
9
8
1
e
D
P
2
0
0
6
1
1
1
7
P
D

B
sì
G
tu
e
S
T

o
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3

Amy Ellen Schwar tz and Leanna Stiefel

Tavolo 3 Mean Characteristics of Students by Nativity, Fifth and Eighth Grades, Reading Sample, 1997–98

FIFTH GRADERS

EIGHTH GRADERS

Native-Born

Foreign-Born

Native-Born

Foreign-Born

Free lunch eligible

Reduced-price lunch eligible

Female

Asian and other

Black

Hispanic

White

Language other than English

Age

Years in NYC public schools

Took LAB

LAB percentile

Limited English Proﬁcient (LEP)

Part-time special education

Took reading test last year

Took math test last year

0.76

0.07

0.51

0.08

0.39

0.36

0.17

0.36

10.48

4.92

0.06

26.35

0.04

0.10

0.94

0.95

Number of students in sample

55,925

Proportion of students in sample

0.86

0.79

0.08

0.50

0.24

0.26

0.30

0.19

0.64

10.54

3.80

0.14

31.44

0.08

0.05

0.79

0.84

9,046

0.14

0.72

0.09

0.51

0.07

0.40

0.34

0.18

0.35

13.52

7.65

0.04

17.56

0.03

0.09

0.95

0.78

0.09

0.50

0.23

0.30

0.17

0.63

13.59

5.18

0.14

20.32

0.11

0.04

0.83

0.87

45,773

11,692

0.80

0.20

Notes: aEligibility for free lunch is calculated only for students with nonmissing data: approximately
94% of all students. bForeign-born students are students not born on U.S. soil. cLimited English
Proﬁcient students are those students that score less than or equal to the 40th percentile on their
Language Assessment Battery (LAB) exam.

regressions by school ﬁxed effects. As an example, the average school attended
by a native-born student is smaller and has higher spending and a slightly
larger teacher-pupil ratio.17

These disparities between the foreign- and native-born students suggest
that we should expect something of a nativity gap in unadjusted test scores
even if there is no speciﬁc difference due to nativity per se. How much of a gap
exists, and how much is explained by these factors, are empirical questions to
which we now turn.

6. RESULTS
Fifth Grade

As shown in Table 4, the 1997–98 ﬁfth-grade reading and math results
point consistently to an immigrant advantage. In reading, the foreign-born

17.

Information on the differences in schools is available from the authors. See also Ellen et al. (2002).

D
o
w
N
o
UN
D
e
D

F
R
o
M
H

T
T

:
/
/

D
io
R
e
C
T
.

io
T
.

e
D
tu
e
D
P
UN
R
T
io
C
e
–
P
D

F
/

1
1
1
7
1
6
9
2
9
8
1
e
D
P
2
0
0
6
1
1
1
7
P
D

B
sì
G
tu
e
S
T

o
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3

IS THERE A NATIVITY GAP?

score an average 0.122 standard deviations higher than native-born students
(see column 1). Controlling for prior performance has little impact on the esti-
mate of the nativity gap; the value-added estimate of the disparity is 0.126 (Vedere
column 2). And while production function estimates reduce the magnitude
of the advantage by almost half, IL 0.066 advantage is statistically signiﬁ-
cant. Results in mathematics are similar, pointing, Ancora, to an immigrant
advantage.

When we turn to the other variables, the results are largely consistent
with expectations and the ﬁndings of prior research. To begin, prior-year test
scores are strong predictors of current-year performance: 0.770 E 0.802 In
the value-added models in (2) E (5) and somewhat lower in the education
production function models in (3) E (6). Further, students with prior-year test
scores earn higher scores than students without such data, perhaps reﬂecting
the recent entrance of the latter into the New York City public schools, whether
from other U.S. schools or schools in other countries, or perhaps, their exit
from an exempt status—limited English proﬁciency or special education. As is
typical in education production function models, students eligible for free or
reduced-price lunch earn lower scores than ineligible students; performance
is lower among black and Hispanic students than white and Asian students;
and performance declines with age, which may be capturing grade retention.
Here, girls do worse on both reading and math tests.

There are also few surprises in the education variables. Students who take
the LAB do worse on reading tests (as indicated by the −0.921 coefﬁcient
on Took LAB), but reading test scores increase with the score earned on the
test (as indicated by the 0.014 coefﬁcient on LAB Percentile). Finalmente, controllo-
ling for performance on the LAB, LEP-eligible students do somewhat better
than the ineligible (the coefﬁcient is 0.180). Taken together, these provide a
nuanced view of the importance of language proﬁciency on academic perfor-
mance. Students considered at risk for LEP (measured by being tested for
eligibility for services) do worse than those not considered at risk, but within
that population, higher scores on the LAB suggest higher scores on reading
tests, with a positive discontinuity at a score of 40—the score that indicates
eligibility for LEP services. While our analysis does not include information
on whether students received services to improve their English language skills
E, if so, what type of service was received, the results indicate that students
who are LEP-eligible do better than otherwise similar students who are not
eligible.

A particularly interesting ﬁnding is that the coefﬁcient on the Language
Other Than English variable is positive. Questo è, students who live in a home
where a language other than English is spoken earn higher reading scores than
those living in homes where English is spoken, once we control for the direct

EDUCATION FINANCE AND POLICY

D
o
w
N
o
UN
D
e
D

F
R
o
M
H

T
T

:
/
/

D
io
R
e
C
T
.

io
T
.

e
D
tu
e
D
P
UN
R
T
io
C
e
–
P
D

F
/

1
1
1
7
1
6
9
2
9
8
1
e
D
P
2
0
0
6
1
1
1
7
P
D

B
sì
G
tu
e
S
T

o
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3

Amy Ellen Schwar tz and Leanna Stiefel

Tavolo 4 Reading and Math Test Regression Coefﬁcients, Fifth Grade, 1997–98

READING

MATH

(1)

Level

(2)

0.122∗∗∗
(0.019)

0.126∗∗∗
(0.010)

0.770∗∗∗
(0.006)

0.468∗∗∗
(0.023)

(3)

EPF

0.066∗∗∗
(0.009)

0.651∗∗∗
(0.006)

0.369∗∗∗
(0.020)

−0.132∗∗∗
(0.010)

−0.052∗∗∗
(0.012)

−0.037∗∗∗
(0.005)

−0.033∗∗∗
(0.007)

0.061∗∗∗
(0.013)

−0.076∗∗∗
(0.012)

−0.052∗∗∗
(0.012)

0.031∗∗∗
(0.008)

−0.921∗∗∗
(0.055)

0.014∗∗∗
(0.001)

0.180∗∗∗
(0.041)

−0.260∗∗∗
(0.011)

0.183∗∗∗
(0.039)

0.077∗∗∗
(0.018)

0.047∗∗∗
(0.016)

0.024∗
(0.013)

0.005
(0.008)

(4)

Level

(5)

0.061∗∗∗
(0.022)

0.105∗∗∗
(0.010)

0.802∗∗∗
(0.005)

0.335∗∗∗
(0.022)

(6)

EPF

0.049∗∗∗
(0.008)

0.696∗∗∗
(0.005)

0.276∗∗∗
(0.022)

−0.091∗∗∗
(0.008)

−0.042∗∗∗
(0.011)

−0.048∗∗∗
(0.005)

−0.025∗∗∗
(0.007)

0.086∗∗∗
(0.012)

−0.131∗∗∗
(0.011)

−0.087∗∗∗
(0.010)

0.052∗∗∗
(0.007)

−0.622∗∗∗
(0.056)

0.011∗∗∗
(0.001)

0.208∗∗∗
(0.044)

−0.166∗∗∗
(0.011)

0.027
(0.036)

0.123∗∗∗
(0.018)

0.092∗∗∗
(0.016)

0.060∗∗∗
(0.013)

0.014∗
(0.007)

Foreign-born

Prior-year

test score

Have prior-year
test score

Free lunch

Reduced-price

lunch

Female

Age

Asian and other

Black

Hispanic

Language other
than English

Took LAB

LAB percentile

LEP

Part-time special

formazione scolastica

Admission

Cohort 1997

Admission

Cohort 1996

Admission

Cohort 1995

Admission

Cohort 1994

Admission

Cohort 1993

D
o
w
N
o
UN
D
e
D

F
R
o
M
H

T
T

:
/
/

D
io
R
e
C
T
.

io
T
.

e
D
tu
e
D
P
UN
R
T
io
C
e
–
P
D

F
/

1
1
1
7
1
6
9
2
9
8
1
e
D
P
2
0
0
6
1
1
1
7
P
D

B
sì
G
tu
e
S
T

o
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3

IS THERE A NATIVITY GAP?

Tavolo 4 Continued

READING

MATH

(1)

Level

(2)

(3)

EPF

(4)

Level

(5)

(6)

EPF

Constant

−0.016
(0.019)

−0.466∗∗∗
(0.026)

0.079
(0.077)

−0.008
(0.020)

−0.350∗∗∗
(0.024)

0.082
(0.076)

Observations

64,971

66,629

R-squared

0.00

0.54

0.60

0.00

0.58

0.63

Notes: aRobust standard errors, adjusted for within school clustering, in parentheses. bModels 3
E 6 use school-level ﬁxed effects. Cohort dummies are deﬁned by the number of years in NYC
public schools as of October 31. As an example, Admission Cohort 1997 indicates the student
entered between November 1, 1996, and October 31, 1997. Other cohort variables are deﬁned
conformably. The omitted cohort entered on or before October 31, 1992, and have at least ﬁve
years in the school system.

∗signiﬁcant at 10%; ∗∗signiﬁcant at 5%; ∗∗∗signiﬁcant at 1%.

effects of limited English proﬁciency and the student’s other educational char-
acteristics. Understanding the relationship between academic performance,
home language, and LEP services is interesting and important and we plan to
explore this in future research.18

Turning to the Admission Cohort variables, the results are consistent and
intriguing. Performance is lowest for the earliest cohort—entering in 1993 E
for the most part monotonically increasing for both reading and math. Quello
È, the most recent entrants perform the highest. Ovviamente, these ﬁndings
must be interpreted with care because of the signiﬁcant correlations between
cohort and other regressors. While the cohorts include both native and foreign-
born students, the more recent ones are disproportionately foreign-born. Co-
horts entering in kindergarten or before were less than a tenth foreign-born.
Cohorts entering in ﬁrst grade were closer to one-quarter foreign-born, while
cohorts entering after ﬁrst grade were between one- and two-thirds foreign-
born. Further, more recent entrants may not have prior-year test scores, so
the pure advantage conferred by recent entrance may apply to relatively few
students.

Fifth and Eighth Grade, Reading and Math, 1997–98 and 2000–1

As described earlier, we perform similar analyses in eighth grade and for two
different academic years. Tavolo 5 displays only the coefﬁcients on the foreign-
born dummies, in the interest of simplicity. Columns 1 E 2 show mean
difference in levels and value added (VA) in reading, and column 3 shows

18. The total effect of LEP status can be found by combining the coefﬁcients on LAB score and LAB

less than or equal to 40.

EDUCATION FINANCE AND POLICY

D
o
w
N
o
UN
D
e
D

F
R
o
M
H

T
T

:
/
/

D
io
R
e
C
T
.

io
T
.

e
D
tu
e
D
P
UN
R
T
io
C
e
–
P
D

F
/

1
1
1
7
1
6
9
2
9
8
1
e
D
P
2
0
0
6
1
1
1
7
P
D

B
sì
G
tu
e
S
T

o
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3

Amy Ellen Schwar tz and Leanna Stiefel

Tavolo 5 Regression Results, Reading and Math Tests, Foreign-Born Coefﬁcient Only, by Grade and Year

READING

MATH

(1)

Level

(2)

(3)

EPF

(4)

Level

(5)

(6)

EPF

Fifth Grade, 1997–98

Foreign-born

0.122∗∗∗
(0.019)

0.126∗∗∗
(0.010)

0.066∗∗∗
(0.008)

0.061∗∗∗
(0.022)

0.105∗∗∗
(0.010)

0.049∗∗∗
(0.008)

Observations

64,971

66,629

0.00

0.54

0.60

0.00

0.58

0.63

Fifth Grade, 2000–1

Foreign-born

0.083∗∗∗
(0.018)

0.089∗∗∗
(0.010)

0.046∗∗∗
(0.009)

0.115∗∗∗
(0.021)

0.108∗∗∗
(0.012)

0.055∗∗∗
(0.009)

Observations

71,141

72,509

0.00

0.47

0.52

0.00

0.55

0.60

Eighth grade, 1997–98

Foreign-born

−0.004
(0.024)

0.037∗∗∗
(0.010)

0.035∗∗∗
(0.009)

−0.029
(0.028)

0.062∗∗∗
(0.012)

0.038∗∗∗
(0.009)

Observations

57,465

59,749

0.00

0.58

0.61

0.00

0.56

0.59

Eighth grade, 2000–1

Foreign-born

0.014
(0.027)

0.058∗∗∗
(0.013)

0.030∗∗∗
(0.008)

0.099∗∗∗
(0.027)

0.148∗∗∗
(0.013)

0.068∗∗∗
(0.008)

Observations

57,152

59,024

0.00

0.54

0.62

0.00

0.59

0.65

Prior-year test score

Demographic characteristics No

Educational characteristics

School ﬁxed effects

Cohort variables

Yes

Notes: a Robust standard errors, adjusted for within school clustering, in parentheses. b Demo-
graphic characteristics are: eligible for free lunch, eligible for reduced-price lunch, female, Asian/
other, black, Hispanic, age, and a dummy indicating free lunch data is nonmissing. Educational
characteristics are: language other than English frequently spoken at home, took the Language
Assessment Battery (LAB), percentile on the LAB, scored at or below the 40th percentile on the
LAB, part-time special education participation, prior-year test score, and whether student took test
in prior year. Cohort variables are dummies for the number of years in NYC Public Schools.

∗signiﬁcant at 10%; ∗∗signiﬁcant at 5%; ∗∗∗signiﬁcant at 1%.

the foreign-born coefﬁcients from the estimated reading production function
(EPF). Columns 4, 5, E 6 present comparable results for math performance.
È interessante notare, while the results in columns 1 E 4 provide mixed evidence
on the sign and magnitudes of the nativity gap—favoring the foreign-born in

D
o
w
N
o
UN
D
e
D

F
R
o
M
H

T
T

:
/
/

D
io
R
e
C
T
.

io
T
.

e
D
tu
e
D
P
UN
R
T
io
C
e
–
P
D

F
/

1
1
1
7
1
6
9
2
9
8
1
e
D
P
2
0
0
6
1
1
1
7
P
D

B
sì
G
tu
e
S
T

o
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3

IS THERE A NATIVITY GAP?

the ﬁfth grade, but showing largely insigniﬁcant results in the eighth—the
value-added results indicate consistently higher value added among immi-
grants. There is a considerable variation in the size of the advantage, ranging
from 0.037 standard deviations in the eighth grade reading (1997–98) A .148
in eighth-grade math (2000–1). Finalmente, the estimates from the fully speci-
ﬁed production functions are similarly consistent—immigrants outperform
native-born students, ceteris paribus—and the range of estimates is narrower.
The estimate of the nativity gap ranges from a low of 0.030 in eighth-grade
reading in 2000–1 to a high of 0.068 in math in that same grade and year.

È interessante notare, in the ﬁfth grade, the size of the unadjusted gap is reduced by
as much as one-half as adjustments are made for differences in students and
schools. In the eighth grade, the disparity emerges with control variables, hav-
ing been obscured by compositional differences and other differences driving
disparities in performance.

Is the Production Function Different for Foreign-Born Students?

Our analyses thus far have constrained the coefﬁcients (or marginal effects)
of variables in the production function to be the same for native-born and
foreign-born students. There are, Tuttavia, many reasons to suspect that there
are productivity differences across groups: immigrants might well respond
differently than native-born students to demographic or educational character-
istics or school resources. In particular, given the signiﬁcance of the language
variables in the production function results and the correlation between im-
migrant status and these variables, immigrants might respond differently to
English language programs.

As described above, we estimated a set of models in which we allowed the
coefﬁcients to differ between the native- and foreign-born students. To pre-
serve space, Tavolo 6 shows key coefﬁcients for ﬁfth and eighth grade reading
in 1997–98.19 The coefﬁcients on the foreign-born interactions capture the
difference in responsiveness between native-born and foreign-born students.
The complete set of interactions on all variables in the production function, COME
well as several selected sets, is statistically signiﬁcant, as shown by the F statis-
tics at the bottom of the table. Also while the coefﬁcients of the foreign-born
in the production function differ statistically from those of the native-born,
most differences between groups are substantively small. Così, Per esempio,
the results indicate that the prior-year test score is a somewhat less important
predictor for the foreign-born than for the native-born—the coefﬁcient for
native-born ﬁfth-grade students is 0.676, and the foreign-born coefﬁcient is
0.028 inferiore.

19. Complete results are available from the authors.

EDUCATION FINANCE AND POLICY

D
o
w
N
o
UN
D
e
D

F
R
o
M
H

T
T

:
/
/

D
io
R
e
C
T
.

io
T
.

e
D
tu
e
D
P
UN
R
T
io
C
e
–
P
D

F
/

1
1
1
7
1
6
9
2
9
8
1
e
D
P
2
0
0
6
1
1
1
7
P
D

B
sì
G
tu
e
S
T

o
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3

Amy Ellen Schwar tz and Leanna Stiefel

Tavolo 6 Selected Reading Test Production Function Coefﬁcients, 1997–98, Interaction Model

Fifth Grade

Eighth Grade

Prior score

Asian and other

Black

Hispanic

Language other than English

Took LAB

LAB percentile

LEP

Constant

FB*Prior Score

Foreign-born

FB*Asian and Other

FB*Black

FB*Hispanic

FB*Language other than English

FB*Took LAB

FB*LAB percentile

FB*LEP

Observations

R-squared

F-stat for interactions = 0

0.676∗∗∗
(0.007)

0.078∗∗∗
(0.015)

−0.131∗∗∗
(0.013)

−0.118∗∗∗
(0.013)

0.029∗∗∗
(0.009)

−0.840∗∗∗
(0.068)

0.013∗∗∗
(0.001)

0.146∗∗∗
(0.050)

0.173∗∗
(0.086)

−0.028∗∗
(0.013)

−0.125
(0.190)

−0.110∗∗∗
(0.030)

0.018
(0.033)

0.004
(0.030)

0.118∗∗∗
(0.024)

−0.387∗∗∗
(0.118)

0.005∗∗∗
(0.002)

0.078
(0.085)

64,971

0.58

8.20∗∗∗

F-stat for demographic interactions = 0

F-stat for demographics and educational = 0

79.95∗∗∗

1284.12∗∗∗

0.719∗∗∗
(0.007)

0.038∗∗
(0.016)

−0.112∗∗∗
(0.018)

−0.101∗∗∗
(0.015)

−0.006
(0.010)

−0.772∗∗∗
(0.106)

0.014∗∗∗
(0.002)

0.304∗∗∗
(0.086)

0.919∗∗∗
(0.094)

−0.058∗∗∗
(0.010)

0.172
(0.214)

−0.133∗∗∗
(0.030)

−0.025
(0.032)

−0.084∗∗∗
(0.028)

0.129∗∗∗
(0.024)

−0.535∗∗∗
(0.138)

0.010∗∗∗
(0.003)

0.150
(0.114)

57,465

0.60

12.86∗∗∗

84.36∗∗

1229.41∗∗∗

Notes: aRobust standard errors, adjusted for within-school clustering, in parentheses. bAll interac-
tion variables are the foreign-born variable interacted with the full set of variables as in table 4.
cDemographic interactions are foreign-born interacted with free lunch, reduced-price lunch, female,
age, Asian, black, Hispanic only. dDemographics plus educational variables exclude the cohort
variables from the “all interactions” list only.

∗signiﬁcant at 10%; ∗∗signiﬁcant at 5%; ∗∗∗signiﬁcant at 1%.

D
o
w
N
o
UN
D
e
D

F
R
o
M
H

T
T

:
/
/

D
io
R
e
C
T
.

io
T
.

e
D
tu
e
D
P
UN
R
T
io
C
e
–
P
D

F
/

1
1
1
7
1
6
9
2
9
8
1
e
D
P
2
0
0
6
1
1
1
7
P
D

B
sì
G
tu
e
S
T

o
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3

IS THERE A NATIVITY GAP?

Language and race/ethnicity are two particularly interesting groups of vari-
ables. Recall that students who experience a language other than English at
home are eligible for the LAB, and if they score lower than 40 percent on
the LAB, they are eligible for services to address limited English proﬁciency.
For all tests and grades shown, foreign-born students exposed to a language
other than English at home do better than comparable native-born students,
but those foreign-born students who take the LAB do worse than native-born
students who take the LAB. Each additional percentage score on the LAB test
adds more to the test score for foreign-born than for native-born students.
È interessante notare, in the presence of these controls, the impact of LEP eligibility
is not signiﬁcantly different for the foreign-born than the native-born; questo è,
foreign-born students who are eligible for LEP services do no better or worse
than their native-born peers. The results for language effects are consistent
across tests and grades, although the coefﬁcients are not always statistically
signiﬁcant at the 5 percent or better level.

Equally interesting are the race and ethnicity results. Foreign-born Asian
students do worse than their native-born ethnic peers, while no signiﬁcant
difference emerges between the foreign- and native-born blacks. Results are
inconsistent for Hispanics. Finalmente, the coefﬁcient on the foreign-born dummy
variable becomes insigniﬁcant in these speciﬁcations, indicating that nativity
by itself no longer matters.

In summary, some differences in marginal effects emerge between native-
and foreign-born students, suggesting that some of the disparities in per-
formance reﬂect differences in the impact of underlying factors. That said,
the magnitudes of the differences in the coefﬁcients are by and large rela-
tively small, except for the language variables. In questo caso, the results sug-
gest that additional work exploring the acquisition of English language skills
among immigrants and its implications for their academic performance is
warranted.

Regional Analyses

As is well known, the large size of the immigrant population in New
York City is matched with an astonishing diversity. Così, the summary
foreign-born statistic may mask important disparities within the immi-
grant community. Per esempio, students from the Dominican Republic are
quite different than students from the former Soviet Union, and pooling
them as we have done may mask important differences in their academic
performance.

Tavolo 7 shows that disparities in performance (both unadjusted and value-
added performance) in 1997–98 across region groups vary widely in both

EDUCATION FINANCE AND POLICY

D
o
w
N
o
UN
D
e
D

F
R
o
M
H

T
T

:
/
/

D
io
R
e
C
T
.

io
T
.

e
D
tu
e
D
P
UN
R
T
io
C
e
–
P
D

F
/

1
1
1
7
1
6
9
2
9
8
1
e
D
P
2
0
0
6
1
1
1
7
P
D

B
sì
G
tu
e
S
T

o
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3

Amy Ellen Schwar tz and Leanna Stiefel

Tavolo 7 Regression Coefﬁcients, Mean Difference in Level and Value-Added Reading
Performance by Region, 1997–98

FIFTH GRADERS

EIGHTH GRADERS

Level

Russia

Eastern Europe

Western Europe

China

East Asia

South Asia

West Asia

Africa

Dominican Republic

Caribbean

Guyana

Latin America

Constant

0.679∗∗∗
(0.057)

0.379∗∗∗
(0.075)

0.377∗∗∗
(0.072)

0.641∗∗∗
(0.080)

0.494∗∗∗
(0.044)

0.326∗∗∗
(0.041)

0.342∗∗∗
(0.072)

−0.014
(0.068)

−0.234∗∗∗
(0.038)

−0.037
(0.027)

−0.230∗∗∗
(0.043)

−0.089∗∗
(0.036)

−0.016
(0.019)

0.295∗∗∗
(0.033)

0.237∗∗∗
(0.045)

0.226∗∗∗
(0.054)

0.334∗∗∗
(0.047)

0.275∗∗∗
(0.031)

0.222∗∗∗
(0.027)

0.261∗∗∗
(0.046)

0.157∗∗∗
(0.051)

−0.022
(0.025)

0.079∗∗∗
(0.019)

−0.030
(0.027)

0.048∗∗
(0.022)

−0.461∗∗∗
(0.025)

0.667∗∗∗
(0.066)

0.344∗∗∗
(0.052)

0.212∗∗∗
(0.060)

0.459∗∗∗
(0.075)

0.431∗∗∗
(0.046)

0.208∗∗∗
(0.064)

0.110∗
(0.058)

−0.036
(0.072)

−0.418∗∗∗
(0.046)

−0.135∗∗∗
(0.034)

−0.313∗∗∗
(0.052)

−0.243∗∗∗
(0.036)

0.001
(0.029)

0.292∗∗∗
(0.032)

0.224∗∗∗
(0.034)

0.114∗∗∗
(0.043)

0.168∗∗∗
(0.035)

0.186∗∗∗
(0.026)

0.074∗
(0.039)

0.069
(0.043)

0.051
(0.048)

−0.111∗∗∗
(0.027)

0.004
(0.018)

−0.074∗∗
(0.030)

−0.064∗∗∗
(0.021)

−0.500∗∗∗
(0.028)

Observations

64971

57465

R-squared

0.02

0.54

0.03

0.58

Notes: aRobust standard errors, adjusted for within-school clustering, in paren-
theses. bModels include regional dummy variables. See appendix for list of coun-
tries in each region.

∗signiﬁcant at 10%; ∗∗signiﬁcant at 5%; ∗∗∗signiﬁcant at 1%.

ﬁfth and eighth grades.20 On raw scores, students from Russia outperform
native-born students by almost 0.7 standard deviations in reading in both
grades, while students from the Dominican Republic on average earn con-
siderably lower scores than native-born students—between 0.2 E 0.4 stan-
dard deviations lower. Value-added scores diminish the differences, although

20. Results for math and for 2000–1 in both subjects are qualitatively similar and available from the

authors.

D
o
w
N
o
UN
D
e
D

F
R
o
M
H

T
T

:
/
/

D
io
R
e
C
T
.

io
T
.

e
D
tu
e
D
P
UN
R
T
io
C
e
–
P
D

F
/

1
1
1
7
1
6
9
2
9
8
1
e
D
P
2
0
0
6
1
1
1
7
P
D

B
sì
G
tu
e
S
T

o
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3

IS THERE A NATIVITY GAP?

Russians still score approximately 0.31 standard deviations higher than native-
born students, while students from the Dominican Republic show −0.1 to no
signiﬁcant differences.

Notice, Tuttavia, that there are also considerable differences in the char-
acteristics of students across regions, as shown in Table 8. While only 56.8
percent of ﬁfth grade Russian immigrants are free-lunch-eligible, compared
A 94.7 percent of Dominican students. Allo stesso modo, only 2 percent of Russians
are LEP-classiﬁed, while one-quarter of all Dominican ﬁfth graders were LEP-
classiﬁed in 1997–98. There are also substantial differences in tenure in New
York City public schools. The average African immigrant has attended New
York’s public schools for less than three years, but the average West Asian
student attended for more than four years. Results are similar in the eighth
grade and in other years.21

To what extent do these underlying differences explain the regional dis-
parities in performance? Tavolo 9 shows the results of our regression analyses
for ﬁfth and eighth grade reading in 1997–98 and 2000–1.22 As before, IL
disparities among immigrants from different regions, and for each region
compared to native-born students, are considerably dampened by regression
analysis. Clearly many of the interregional disparities are driven by sociodemo-
graphic, educational, and cohort characteristics captured in the regressions.
To be speciﬁc, more than three-quarters of the 0.679 gap in level of ﬁfth-grade
reading performance in 1997–98 between Russians and native-born students
seems to be explained by differences captured elsewhere in the regression.
Similarly dramatic declines obtain for several other regions—Eastern Europe
drops from 0.379 A 0.097; Western Europe from 0.377 A 0.121; China from
0.641 A 0.150; and East Asia, South Asia, and West Asia show the same
pattern. È interessante notare, while the unadjusted results show immigrants from
some regions earning high scores and others falling below the native-born,
the education production function disparities are positive, with Guyana the
only consistent exception. Così, the results make clear that the earlier result
pointing to a nativity gap that favors the foreign-born is not driven merely by
the performance of a single group, as some might have expected. Invece, stu-
dents from several world regions outperform their native-born peers, spanning
countries with exceptional educational systems and highly educated popula-
tions as well as countries with inadequate educational systems and low literacy
rates. Allo stesso tempo, there is no single immigrant experience, and future
work is warranted that examines the causes of these divergent educational
experiences.

21. The 2000–1 results are available from the authors.
22. The 2000–1 results and math results are similar.

EDUCATION FINANCE AND POLICY

D
o
w
N
o
UN
D
e
D

F
R
o
M
H

T
T

:
/
/

D
io
R
e
C
T
.

io
T
.

e
D
tu
e
D
P
UN
R
T
io
C
e
–
P
D

F
/

1
1
1
7
1
6
9
2
9
8
1
e
D
P
2
0
0
6
1
1
1
7
P
D

B
sì
G
tu
e
S
T

o
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3

Amy Ellen Schwar tz and Leanna Stiefel

Tavolo 8 Characteristics of Students by Region, 1997–98, Fifth and Eighth Graders

Percent Percent
Free-

Number of Lunch
Students

Eligible Eligible

Reduced-
Percent
Price Lunch Percent Special
LEP

Percent Public

Education Female Schools

Years in
NYC

Region

Fifth Grade

Africa

Caribbean

China

Dominican
Republic

East Asia

Eastern Europe

Guyana

206

1,911

459

1,409

471

329

729

Latin America

1,296

Russia

1,127

South Asia

West Asia

Western Europe

638

219

252

All foreign-born

9,046

All native-born

55,925

Eighth Grade

Africa

Caribbean

China

Dominican
Republic

East Asia

Eastern Europe

Guyana

224

2,890

678

1,667

665

382

956

Latin America

1,784

Russia

1,230

South Asia

West Asia

Western Europe

693

257

266

All foreign-born 11,692

All native-born

45,773

77.7

83.5

69.1

94.7

45.2

59.0

84.1

86.1

56.8

71.3

68.0

56.3

76.6

73.5

70.1

74.7

66.4

92.2

46.6

61.3

77.2

84.1

49.8

70.9

68.5

59.4

72.9

66.8

7.3

6.4

10.2

2.3

21.0

14.9

8.2

5.6

10.6

10.2

7.3

12.7

8.1

7.0

8.9

7.4

12.4

2.1

17.4

12.6

7.6

5.5

13.4

11.4

5.4

10.2

8.3

7.9

6.3

1.0

6.5

25.1

2.8

6.1

0.0

18.5

2.0

5.8

3.2

3.6

8.2

4.3

6.3

2.0

10.8

35.5

7.4

6.0

0.5

21.2

2.0

8.1

5.1

4.1

10.9

3.1

4.9

5.5

3.3

5.3

4.0

3.0

6.6

6.9

4.7

5.8

8.2

6.7

5.5

9.7

2.2

4.6

3.7

4.4

2.4

3.4

4.0

5.3

2.5

2.0

4.7

6.4

4.0

8.6

51.0

54.0

50.1

49.7

50.5

51.7

52.8

46.6

49.1

45.9

51.1

46.0

50.2

51.1

49.6

52.7

49.0

49.2

52.3

52.9

53.2

46.9

48.9

47.0

37.0

55.6

50.1

50.9

2.7

3.1

4.1

4.4

3.9

3.7

3.1

4.4

3.9

4.0

4.3

3.8

4.9

3.4

4.2

6.0

5.3

5.2

4.3

6.3

4.7

5.5

5.9

5.7

5.2

7.7

D
o
w
N
o
UN
D
e
D

F
R
o
M
H

T
T

:
/
/

D
io
R
e
C
T
.

io
T
.

e
D
tu
e
D
P
UN
R
T
io
C
e
–
P
D

F
/

1
1
1
7
1
6
9
2
9
8
1
e
D
P
2
0
0
6
1
1
1
7
P
D

B
sì
G
tu
e
S
T

o
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3

IS THERE A NATIVITY GAP?

Tavolo 9 Regional Regression Coefﬁcients, Education Production Functions, Fifth- and Eighth-
Grade Reading, 1997–98 and 2000–1

FIFTH GRADERS

EIGHTH GRADERS

1997–98

2000–1

1997–98

2000–1

Russia

Eastern Europe

Western Europe

China

East Asia

South Asia

West Asia

Africa

Dominican Republic

Caribbean

Guyana

Latin America

Constant

0.123∗∗∗
(0.026)

0.097∗∗∗
(0.036)

0.121∗∗
(0.047)

0.150∗∗∗
(0.040)

0.082∗∗∗
(0.030)

0.055∗∗
(0.026)

0.107∗∗
(0.046)

0.089∗
(0.050)

0.120∗∗∗
(0.019)

0.018
(0.017)

−0.135∗∗∗
(0.028)

0.087∗∗∗
(0.019)

0.055
(0.077)

0.069∗∗
(0.032)

0.052
(0.041)

0.065
(0.041)

0.159∗∗∗
(0.033)

0.033
(0.036)

−0.015
(0.024)

0.092∗
(0.049)

0.181∗∗∗
(0.049)

0.069∗∗∗
(0.022)

−0.028
(0.022)

0.004
(0.027)

0.068∗∗∗
(0.022)

0.086
(0.074)

0.169∗∗∗
(0.033)

0.133∗∗∗
(0.036)

0.065
(0.041)

0.107∗∗∗
(0.032)

0.102∗∗∗
(0.029)

−0.011
(0.037)

−0.025
(0.037)

0.055
(0.048)

0.060∗∗∗
(0.020)

0.004
(0.017)

−0.114∗∗∗
(0.032)

0.025
(0.018)

0.839∗∗∗
(0.083)

0.178∗∗∗
(0.030)

0.098∗∗∗
(0.036)

0.062
(0.044)

0.077∗
(0.042)

−0.014
(0.030)

0.033
(0.032)

−0.086∗∗
(0.038)

0.113∗∗
(0.046)

0.085∗∗∗
(0.015)

−0.032∗∗
(0.015)

−0.086∗∗∗
(0.031)

−0.005
(0.015)

1.015∗∗∗
(0.091)

Observations

64,971

71,141

57,465

57,152

R-squared

0.60

0.52

0.61

0.62

Notes: aRobust standard errors, adjusted for within-school clustering, in parentheses.
bThe model includes controls for free-lunch eligibility, reduced-price lunch eligibility,
genere, age, ethnicity/race, English proﬁciency, Language Assessment Battery scores,
special education status, last year’s reading and math scores, and teacher qualiﬁca-
zioni. Cohort dummies control for the numbers of years in NYC public schools. IL
models include school ﬁxed effects.

∗signiﬁcant at 10%, ∗∗signiﬁcant at 5%; ∗∗∗signiﬁcant at 1%.

7. CONCLUSIONS
Our analyses of the academic performance of native-born and foreign-born
students in New York City public schools suggest that immigrant students
by and large perform better than native-born students in reading and math,
even though the size of the nativity gap is diminished if one controls for
the characteristics of students and schools. These results, estimated for a
representative elementary school grade and a representative middle school

EDUCATION FINANCE AND POLICY

D
o
w
N
o
UN
D
e
D

F
R
o
M
H

T
T

:
/
/

D
io
R
e
C
T
.

io
T
.

e
D
tu
e
D
P
UN
R
T
io
C
e
–
P
D

F
/

1
1
1
7
1
6
9
2
9
8
1
e
D
P
2
0
0
6
1
1
1
7
P
D

B
sì
G
tu
e
S
T

o
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3

Amy Ellen Schwar tz and Leanna Stiefel

grade, are remarkably robust. Inoltre, while foreign-born students have
different production functions than do native-born students, the differences
are substantively small, except perhaps for the impacts of language-skill-related
variables, where the foreign-born do seem to be affected differently.

The implications of our results for school policy are intriguing. How dif-
ferent are immigrants from native-born students? Not as much, perhaps, COME
raw score differentials suggest. And to the extent that there are differences,
immigrant students typically perform better. Thus immigrants may not pose
problems for the four areas of concern—equity, peers, accountability, or future
labor market success. Infatti, our results suggest that differences between im-
migrants and foreign-born are driven by the same factors that drive differences
among native-born students, although responsiveness may be somewhat dif-
ferent. As an example, immigrants who live in homes where languages other
than English are spoken do better than similarly situated native-born students,
while immigrants who are assessed for English language proﬁciency do worse,
ceteris paribus.

To some degree, these results are good news for New York City pub-
lic schools. Immigrants make up over 17 percent of elementary and middle
school children, and New York City’s role as a port of entry seems likely to
continue. The city has trouble helping many of its students reach adequate
educational levels, and the job would be even more difﬁcult if immigrant chil-
dren needed help beyond what the native-born children need. But this study
indicates that this may not be the case. Perhaps this is because the city already
supplies immigrants with sufﬁcient resources where they are needed. Infatti,
Schwartz and Stiefel (2004) provide evidence that resources are distributed on
the basis of educational characteristics of students and neutrally with respect
to immigrant status. Perhaps immigrants perform better because immigrant
families are particularly motivated to “make it” in America and, to do so, Essi
place a premium on education. Alternatively, immigrant students may, SU
average, come to the United States with better schooling backgrounds than
their native-born peers. To the extent that prior-test performance inadequately
captures academic preparation in the regression equation, our foreign-born
variables may reﬂect those differences. Whatever the reasons, there is little
evidence here that immigrant students are faring particularly poorly in city
schools.

Some will wonder whether these results merely reﬂect the overall poor per-
formance of New York City public school children. Perhaps the foreign-born
advantage is not hard-won at all but, Piuttosto, the result of low performance
by the native-born students. So how well do New York City students perform
compared to other U.S. students? Recent NAEP results from ﬁve large urban
districts (Lutkus et al. 2003) reveal that New York City (and Houston) scored

D
o
w
N
o
UN
D
e
D

F
R
o
M
H

T
T

:
/
/

D
io
R
e
C
T
.

io
T
.

e
D
tu
e
D
P
UN
R
T
io
C
e
–
P
D

F
/

1
1
1
7
1
6
9
2
9
8
1
e
D
P
2
0
0
6
1
1
1
7
P
D

B
sì
G
tu
e
S
T

o
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3

IS THERE A NATIVITY GAP?

higher in fourth-grade reading than the three other urban districts, compa-
rable to all “central cities,” although lower than the national average.23 Thus,
while urban students do not perform as well as students in suburban districts,
New York City students do not do poorly compared to other urban districts.
Perhaps our immigrants are part of the reason.

In sum, our results suggest that nativity itself explains little of the dispari-
ties in performance across students. Invece, much of the disparity is explained
by the same set of variables that explain differences in performance of native-
born students—English language proﬁciency, prior performance, genere, E
the characteristics of their schooling. This means that it may be best to channel
new resources toward “old” persistent problems of socioeconomic and racial
disparities in performance, which do continue to disadvantage poor, black,
and Hispanic students, whether immigrant or native-born.

For helpful comments, we thank Dylan Conger, William Duncombe, Christopher
Jepsen, an anonymous referee, David Figlio, seminar participants at New York Uni-
versity, University of Texas–Dallas, McMaster University, and participants at sessions
of the American Education Finance Association 2004 meetings and the Association
for Public Policy and Management 2004 meetings. The Spencer Foundation provided
generous support for the project, and Luis Chalico and Charles Parekh were excellent
research assistants. All responsibility for analyses and conclusions is ours.

REFERENCES
Berg, Nathan, and John Kain. 2003. The academic performance of immigrant and
native-born school children in Texas. Unpublished paper, University of Texas–Dallas.

Bleakley, Hoyt, and Aimee Chin. 2004. Language skills and earnings: Evidence from
childhood immigrants. Review of Economics and Statistics 86, 481–96.

Cook, M. D., and William N. Evans. 2000. Families or schools? Explaining the conver-
gence in white and black academic performance. Journal of Labor Economics 18:729–54.

Demie, Feysia. 2001. Ethnic and gender differences in educational achievement and
implications for school improvement strategies. Educational Research 43:91–106.

Ellen, Ingrid G., Katherine O’Regan, Amy Ellen Schwartz, and Leanna Stiefel. 2002.
Immigrant children and urban schools: Evidence from New York City on segregation
and its consequences for schooling. In Brookings—Wharton Papers on Urban Affairs,
edited by William G. Gale and Janet Rothenberg Pack. Washington, DC: Brookings
Institution Press, pag. 183–205.

Fryer, Roland G., and Steven D. Levitt. 2004. Understanding the black-white test score
gap in the ﬁrst two years of school. Review of Economics and Statistics 86:551–60.

23. The other cities are Atlanta, Chicago, e Los Angeles. For New York, only fourth-grade results are

given.

EDUCATION FINANCE AND POLICY

D
o
w
N
o
UN
D
e
D

F
R
o
M
H

T
T

:
/
/

D
io
R
e
C
T
.

io
T
.

e
D
tu
e
D
P
UN
R
T
io
C
e
–
P
D

F
/

1
1
1
7
1
6
9
2
9
8
1
e
D
P
2
0
0
6
1
1
1
7
P
D

B
sì
G
tu
e
S
T

o
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3

Amy Ellen Schwar tz and Leanna Stiefel

Fryer, Roland G., and Steven D. Levitt. 2005. The black-white test score gap through
third grade. NBER Working Paper No. 11049.

Gandara, Patricia, Russell Rumberger, Julie Maxwell-Jolly, and Rebecca Callahan.
2003. English learners in California schools: Unequal resources, unequal out-
comes. Education Policy Analysis Archives 11 (36). Accessed 27 Luglio 2005. Available
epaa.asu.edu/epaa/v11n36/.

Glick, Jennifer E., and Michael J. White. 2003. The academic trajectories of immigrant
youths: Analysis within and across cohorts. Demography 40:759–83.

Hanushek, Erik A. 1986. The economics of schooling: Production and efﬁciency in
public schools. Journal of Economic Literature 24:1141–77.

Hedges, Larry V., and Amy Nowell. 1998. Black-white test score convergence since
1965. In The Black White Test Score Gap, edited by Christopher Jencks and Merrideth
Phillips. Washington, DC: Brookings Institution Press, pag. 149–81.

Hedges, Larry V., and Amy Nowell. 1999. Changes in the black-white gap in achieve-
ment test scores. Sociology of Education 72:111–35.

Hirschman, Charles. 2001. The educational enrollment of immigrant youth: A test of
the segmented-assimilation hypothesis. Demography 38:317–36.

Jacob, Brian A. 2004. Remedial education and student achievement: A regression-
discontinuity analysis. Review of Economics and Statistics 88:226–44.

Kao, Grace. 1999. Psychological well-being and educational achievement among
immigrant youth. In Children of immigrants: Health, adjustment, and public assistance,
edited by D. J. Hernandez. Washington, DC: National Academy Press, 410–77.

Kao, Grace, and Marta Tienda. 1995. Optimism and achievement: The educational
performance of immigrant youth. Social Science Quarterly 76:1–19.

Lutkus, UN. D., UN. W. Weiner, M. C. Daane, and Y. Jin. 2003. The nation’s report card:
Reading 2002, trial urban district assessment. Washington, DC: NOI. Department of Edu-
catione, NCES No. 2003-523.

Matsudaira, Jordan. 2004. Sinking or swimming? Evaluating the impact of English
immersion versus bilingual education on student achievement. Unpublished paper,
University of Michigan.

McEwan, Patrick J. 2003. Peer effects on student achievement: Evidence from Chile.
Economics of Education Review 22:131–42.

Portes, Alejandro, and MacLeod Dag. 1996. Educational progress of children of immi-
grants: The roles of class, ethnicity, and school context. Sociology of Education 69:255–
75.

Qin–Hilliard,Desiree Baolian, Erika Feinauer, and Blanca G. Quiroz. 2001. Introduc-
zione. Harvard Educational Review 71:v–x.

Reardon, Sean F. 2003. Sources of educational inequality: The growth of racial/ethnic
and socioeconomic test score gaps in kindergarten and ﬁrst grade. Unpublished Work-
ing Paper 03–05R, Population Research Institute, Pennsylvania State University.

D
o
w
N
o
UN
D
e
D

F
R
o
M
H

T
T

:
/
/

D
io
R
e
C
T
.

io
T
.

e
D
tu
e
D
P
UN
R
T
io
C
e
–
P
D

F
/

1
1
1
7
1
6
9
2
9
8
1
e
D
P
2
0
0
6
1
1
1
7
P
D

B
sì
G
tu
e
S
T

o
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3

IS THERE A NATIVITY GAP?

Rosen, Rae, Susan Wieler, and Joseph Pereira. 2005. New York City immigrants: IL
1990s wave. In Current Issues in Economics and Finance. New York: Federal Reserve
Bank of New York, pag. 1–4.

Schnepf, Sylke Viola. Giugno 2004. How different are immigrants? A cross-country and
cross-survey analysis on educational achievement. Institute for the Study of Labor,
Discussion Paper No. 1398.

Schwartz, Amy Ellen, and Leanna Stiefel. 2004. Immigrants and the distribution of
resources within an urban school district. Educational Evaluation and Policy Analysis
26:303–28.

Schwartz, Amy Ellen, Leanna Stiefel, and Dylan Conger. 2002. Alternative regional
classiﬁcations of New York City’s immigrant children. Unpublished paper, New York
Università. Accessed 11 Febbraio 2004. Available http://urban.nyu.edu/education/
regionpaper.pdf.

Singer, Audrey. 2004. The rise of new immigrant gateways. Living city census series.
Washington, DC: Brookings Institution Press.

Stiefel, Leanna, Amy Ellen Schwartz, and Dylan Conger. 2003. Language proﬁciency
and home languages in New York City elementary schools. Unpublished paper,
New York University. Accessed 11 Febbraio 2004. Available http://urban.nyu.edu/
education/nyclanguage.pdf.

Suarez-Orozco, Carola, and Marcelo Suarez-Orozco. 1995. Transformations: Immigra-
zione, family life, and achievement motivation among Latino adolescents. Stanford, CA:
Stanford University Press.

Todd, Petra E., and Kenneth I. Wolpin. 2003. Towards a uniﬁed approach for modeling
the production function for cognitive achievement. Economic Journal 113:3–33.

Waters, Mary C. 1999. Black identities: West Indian immigrant dreams and American
realities. New York: Russell Sage Foundation.

APPENDIX. COUNTRIES IN REGION GROUPS

Russia: Armenia, Azerbaijan, Belarus, Georgia, Kazakhstan,
Kyrgyzstan, Moldova, Russia, Tajikistan, Turkmenistan,
Ukraine, Uzbekistan

East Europe: Albania, Bosnia & Herzegovina, Bulgaria, Croatia, Czech
Republic, Estonia, Hungary, Latvia, Lithuania, Macedonia,
Poland, Romania, Slovak Republic, Slovenia, Yugoslavia

West Europe: Australia, Austria, Belgium, Bermuda, Canada, Denmark,
Finland, France, Germany, Greece, Iceland, Ireland, Italy,
Luxembourg, Malta, Monaco, Netherlands, New Zealand,
Norway, Portugal, Spain, Sweden, Svizzera, United
Kingdom

China: China, Hong Kong, Taiwan

EDUCATION FINANCE AND POLICY

D
o
w
N
o
UN
D
e
D

F
R
o
M
H

T
T

:
/
/

D
io
R
e
C
T
.

io
T
.

e
D
tu
e
D
P
UN
R
T
io
C
e
–
P
D

F
/

1
1
1
7
1
6
9
2
9
8
1
e
D
P
2
0
0
6
1
1
1
7
P
D

B
sì
G
tu
e
S
T

o
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3

Amy Ellen Schwar tz and Leanna Stiefel

East Asia: Bhutan, Brunei Darussalam, Burma

(Myanmar),
Japan,
Cambodia, Fiji, French Polynesia, Indonesia,
North Korea, South Korea, Laos, Macao, Malaysia,
Maldives, Marshall Island, Micronesia, Mongolia, Nepal,
Papua New Guinea, Philippines, Samoa, Singapore,
Solomon Islands, Sri Lanka, Thailand, Vanuatu, Vietnam

South Asia: Bangladesh, India, Pakistan
West Asia: Afghanistan, Algeria, Bahrain, Cyprus, Egypt, Iran, Iraq,
Israel, Jordan, Kuwait, Lebanon, Libya, Morocco, Oman,
Qatar, Saudi Arabia, Syria, Tunisia, Turkey, United Arab
Emirates, Yemen

Africa: Angola, Benin, Botswana, Burkina Faso, Burundi,
Cameroon, Cape Verde, Central African Republic, Chad,
Comoros, Congo, Djibouti, Equatorial Guinea, Ethiopia,
Gabon, Gambia, Ghana, Guinea-Bissau, Guinea, Ivory
Coast, Kenya, Lesotho, Liberia, Madagascar, Malawi,
Mali, Mauritania, Mauritius, Mozambique, Namibia,
Niger, Nigeria, Rwanda, Sao Tome & Principe, Senegal,
Seychelles, Sierra Leone, Somalia, Republic of South
Africa, Sudan, Swaziland, Tanzania, Togo, Tonga, Uganda,
Zaire, Zambia, Zimbabwe

Dominican Republic: Dominican Republic

Caribbean: Antigua & Barbuda, Bahamas, Barbados, British Virgin
Islands, British West Indies, Cuba, Dominica, French
Antilles, French West Indies, Grenada, Guadeloupe, Haiti,
Jamaica, Nether Antilles, St. Kitts & Nevis, St. Lucia,
St. Vincent & Grenada, Trinidad & Tobago

Guyana: French Guiana, Guyana, Surinam

Latin America: Argentina, Belize, Bolivia, Brasile, Chile, Colombia, Costa
Rica, Ecuador, El Salvador, Guatemala, Honduras, Mexico,
Nicaragua, Panama, Paraguay, Peru, Uruguay, Venezuela

D
o
w
N
o
UN
D
e
D

F
R
o
M
H

T
T

:
/
/

D
io
R
e
C
T
.

io
T
.

e
D
tu
e
D
P
UN
R
T
io
C
e
–
P
D

F
/

1
1
1
7
1
6
9
2
9
8
1
e
D
P
2
0
0
6
1
1
1
7
P
D

B
sì
G
tu
e
S
T

o
N
0
7
S
e
P
e
M
B
e
R
2
0
2
3

49
Scarica il pdf