Immigrants and Inequality in Public Schools
Amy Ellen Schwartz Professor of Public Policy, Education and Economics and Leanna Stiefel Professor of Economics and Education Policy
Institute for Education and Social Policy Wagner Graduate School and Steinhardt School New York University Prepared for Project on Social Inequality and Educational Disadvantage September 24, 2009
We thank Elizabeth Debraggio for her excellent research assistance and Richard Murnane, Rebecca Blank, and Richard Freeman for helpful comments on an earlier outline.
I. Introduction While there is considerable discussion about the impact of immigration and immigrants on the economy as a whole, there is much less that examines the impact on schools and student performance. In 2007, immigrants were almost 13% of the U.S. population and over 5% of five to seventeen year olds (see Table 1). Thus, understanding whether, and in what way, immigrants change the context of schooling and student outcomes is critical to policy makers’ goals of improving academic performance and lessening educational disparities among American youth. This paper begins to fill the gap in knowledge on the impact of immigrants in education by examining the relationship between immigrant communities and school resources and the impact of immigrants on their native- and foreign-born schoolmates. Much of the previous work looking at the impact of immigrants on the nativeborn focuses on labor markets. While this work has generally found that an increase in the immigrant population in a community has a negative impact on the wages of lowskilled native-born workers, evidence on the size of this impact is mixed. Some authors, such as Rachel Friedberg and Jennifer Hunt (1995) and David Card (2001), find that immigration has a small, negative effect on employment rates and wages of low-skilled native-born workers. More recent work by George Borjas (2003), however, finds a larger negative impact of immigration on the wages of low-skilled native-born workers, especially those without high school diplomas. At the same time Gianmarco Ottaviano and Giovanni Peri (2005) argue that the small negative impact of immigration on lowskilled native-born wages and employment is outweighed by increased demand for higher-level workers and improved physical capital. 1
A smaller literature examines the effects of immigrants on the academic performance of fellow immigrants and native-born classmates. Peter Jensen and Astrid Wurtz (2008) conclude that higher immigrant concentration in Danish schools has a negative impact on the reading scores of immigrant and native students, although controlling for sorting across neighborhoods yields a more modest negative impact on native Danes and no effect on immigrants. Using data from Israel, Eric Gould, Victor Lavy, and M. Daniele Paserman (2008) show negative short- and long-term effects of high immigrant concentrations on the native-born, but the impact on fellow immigrant students is less clear. Jane Friesen and Brian Krauth (2008) find that the effect varies by immigrant group and attribute this variation to differences in human capital and cultural norms across immigrant groups. They do not explore the impact of immigrant concentration on the native-born. In this paper, we examine evidence from New York City (NYC) on the distribution of resources across communities and schools and the consequences, both for immigrants and for their native-born schoolmates, for academic performance. Because NYC is the largest and perhaps the most diverse K-12 public school system in the country, it provides an excellent setting to study immigrants and inequality in public schools. Compared to the U.S. average of 12.6% foreign-born population and 5% foreign-born 5-17 year old population, New York City has 28.3% and 9% respectively. Only Miami and Los Angeles have larger foreign-born populations in total or of 5-17 year olds. (See Table 1.) Additionally, nearly 17 % of NYC’s public elementary and
2
middle school students are immigrants, originating from over 200 countries and speaking over 160 languages and dialects.1 We begin by analyzing the determinants of school resources and the extent to which the presence of immigrants - either in the community or in the school - shape those resources. Resources are broadly defined to include spending, teachers per pupil, teacher characteristics such as experience and education level, and composition of schools (peers). We ask whether schools in immigrant communities differ in the resources they receive compared to schools serving other populations. Then, we examine the performance of immigrant and native-born students and, in particular, the extent to which these vary with the presence of immigrants in schools We use detailed, longitudinal data from administrative sources on test scores, demographics, and educational needs; school resources, size, characteristics, etc; and characteristics of school communities in New York City. In the next section (II), we discuss immigrants and school resources, including a literature review, description of data, and analysis of results. In Section III we turn to immigrants and school performance, again reviewing existing literature, describing the data and analyzing the results. In Section IV, we conclude. II. Immigrants and School Resources Although there is little in the way of explicit policy or procedure that directly relates school resources to immigrants, there are a number of channels through which the 1
Immigrant students disproportionately attend public schools in NYC where a large fraction of K-12 students attend private schools, thus accounting for their higher proportion of elementary and middle school students. According to the American Community Survey 3-Year estimates, roughly 22 percent of NYC’s K-8 students attend private school, as do 19 percent of 9-12 graders. These figures are larger than other major cities. See Table 3.
3
presence (or characteristics) of immigrants may influence the resources of public schools. To be clear, there are some federal funds provided to school districts serving recent immigrants and there are a small number of high schools designed specifically to serve the needs of immigrant students. At the elementary and middle school level, however, the intra-district distribution of resources is determined by the interplay of funding formulae, student needs, teacher and principal decisions, community demands, and a host of social, political and institutional factors. Further, different processes govern the distribution of financial resources (i.e., expenditures), teachers, including both the number (determining the pupil-teacher ratio) and their characteristics (education, experience, etc.), and the students peers. According to NYC school district policies, school resources are distributed according to formulas that allocate resources based on student and school characteristics. There are formulas for allocating teacher, principal and other staff resources, for textbook allowances, and so on. As an example, more teachers with particular licensing or qualifications are allocated to schools with higher proportions of students with special needs – say limited English proficient students. Thus, the presence of immigrants may result in more resources for schools if they are disproportionately eligible for help gaining English proficiency or if they are disproportionately poor. Alternatively, more immigrants may result in fewer resources if they are less likely to participate in special education. That said, the characteristics of the teaching staff also reflect the choices and preferences of teachers over schools and neighborhoods and school preferences over teachers. As an example, desirable schools in convenient locations may be more able to choose more educated, experienced teachers to fill vacant slots, resulting in differences in
4
spending on teachers across schools. In principle, then, teaching resources may be driven by the characteristics of the students, the school and the neighborhood as well. In urban settings, elementary and middle school attendance is largely determined by a student’s residence. School district maps demarcate geographic boundaries of a catchment zone (or attendance zone) a school serves, such that all students residing within that zone are eligible to enroll in that school. The implication is that the characteristics of the student body – which form the peer group for any student in the school – are significantly a reflection of the characteristics of the catchment area.
Thus,
neighborhoods with large immigrant populations are likely to be served by schools educating large numbers of immigrant students; neighborhoods with high poverty rates are likely to be served by schools educating large shares of poor students, and so on. Of course, the school population may differ from the population of the catchment zone for many reasons – including, for example, differences in age profiles or fertility across groups (say, immigrants versus native-born), differential reliance on private schooling or non-zoned magnet schools. Importantly, since our analyses define immigrant students as those that are foreign-born, the native-born children of foreign-born parents will not be counted as foreign-born.2 In the end, whether – and how much – the characteristics of the neighborhood population shape the school population and the peers available in school is an empirical question that we consider below. Note that if immigrants are distributed completely evenly across schools – that is, if there is no segregation between the foreign and the native born – there will be no correlation between immigrant representation and school resources. Differences emerge
2
The Urban Institute’s Children of Immigrants Data indicates that 84% of immigrant parents have nativeborn children. See Table 4.
5
due, in part, to the concentration of immigrants across schools or to differences in their characteristics, countries of origin, etc. Finally, we should note that school resources are also a reflection of school policies– about articulation, grade span, mission, etc. – and, more generally the political economy of city finances and expenditures. As an example, middle class neighborhoods with high voter turnout may be better able to garner resources for their school through special appropriations, or by voting for resource allocation processes that favor their own schools. In this case, if immigrant communities have less political power (due to lower voter turnout, if not a lower citizenship rate) then the rules of the game may direct fewer resources to their schools. Alternatively, communities may differ in the mix of public services they “demand,” leading to differences in services provided. Taken together, these imply that school resources – including the number and characteristics of the teaching staff and school spending – may reflect the characteristics of both the school and the community and, ultimately, may be interconnected as schools shape residential location decisions and the composition of the community. A. Previous Research on the Intradistrict Distribution of Resources While a significant body of literature explores the distribution of resources across school districts, studies that examine intradistrict resource distribution are relatively scarce. That said, disparities in intradistrict resource disparities may be greater than disparities across districts (see for example, Linda Hertert, 1995; Sarah Burke, 1999). Existing evidence on intradistrict resource allocation, while somewhat mixed, suggests that schools with higher percentages of poor and minority students receive more money and have a higher number of teachers per student than other schools in the same district 6
(see for example, Ross Rubenstein, Amy Schwartz, Leanna Stiefel & Hella Bel Hadj Amor, 2007). Teachers in these schools, however, are also more likely to be inexperienced and receive lower salaries (see for example, Hamilton Lankford, Susanna Loeb & James Wyckoff, 2002; Patrice Iatarola & Leanna Stiefel, 2003; Charles Clotfelter, Helen Ladd & Jacob Vigdor, 2005; Rubenstein et al., 2007). Research on the relationship between intradistrict resource allocation and immigrant students is particularly limited. Amy Schwartz and Alec Gershberg (2001) explore the relationship between resources and characteristics of the public schools attended by immigrant students and find that immigrant children, on average, attend schools with fewer resources. More recently, Amy Schwartz and Leanna Stiefel (2004) find that school resources decrease as the percentage of immigrant students increases, but this relationship is largely driven by differences in the educational needs of immigrant students. B. Data In order to assess the effect communities have on school composition and resources, we combine data on NYC school catchment areas (attendance zones) and NYC public elementary schools. Census data were obtained for 1990 and 2000 from the Neighborhood Change Database (NCDB), which compiles social, demographic, economic and housing data for all census tracts over time. School zone aggregate level data were created using tract level statistics, reallocated and reweighted to reflect the representation of each tract in the zone. This provided us with a profile of the zone’s population – including statistics detailing age, race/ethnicity, nativity, poverty, average income, and so forth. As shown in Table 2a, during the 1990s school zones increased in 7
shares of foreign-born by 6.2 percent, from 27 percent in 1990 to over 33 percent in 2000. We also see increases in the representation of Hispanics and Asians, and an over 6 percent decrease in the representation of whites. We match this zone level data to school level data from the NYC Department of Education for 1990 and 2000 using unique school identification codes. These data provide information on the composition of the student body, as well as on school characteristics and resources. While information on immigrant status, resource room participation, expenditures and teacher characteristics are not available in the 1990 data, we do have statistics on the racial/ethnic composition of the student body, enrollment, student poverty, and English language ability. As such, while our panel model is more constrained than the model we use for 2000 cross section analysis, we are still able to conduct meaningful estimations of zone and school characteristics on school resources. We observe 628 elementary schools in 1990 and 633 elementary schools in 2000; of these, 577 appear in both years. These schools vary widely – in terms of size (ranging from 25 students to 2,200) and composition. While schools on average are roughly 20 percent white, 35 percent black, 36 percent Hispanic, and 9 percent Asian, there are schools reporting no students of a particular race and others that are over 90 percent single race. Further, Table 2b shows that while some schools have no LEP students or students eligible for free lunch, others are majority LEP or composed entirely of poor students. Finally, we note that in 2000 the average share of foreign-born students in elementary schools is roughly 7 percent, however, there are schools reporting no foreignborn students and others where over one-quarter of the student body is foreign-born. C. Empirical Strategy 8
Although the theory suggests a variety of factors may influence school resources, which matter and how much are empirical questions. We use data on New York City elementary schools and their zones to gain insight into the relationship between schools resources and the characteristics of the school and neighborhoods. Notice, however, that the complex interplay of allocation formulae, preference of teachers, principals and community, and, ultimately, city politics means that disentangling the direction of causality between the underlying determinants and observed outcomes is quite challenging. It is difficult to distinguish supply from demand factors and the reduced form expenditure equations that we estimate are not designed to identify underlying structural parameters. Further, identifying whether the observed relationship between the fraction of immigrant students and school resources or characteristics reflects the impact of inflows of immigrants on those features or the tendency of immigrants to enroll in schools with these features is difficult. Thus, our regression models may be best viewed as descriptive, capturing “what is,” although we make an effort to gain some insight into causal relationships. To do so, we estimate a series of regression models linking measures of school-level resources on a set of variables capturing the characteristics of the school and its zone of the general form: (1)
Yit = Β0 + Β1Schoolit + Β2Zoneit + Β3 Si + eit,
where i indexes schools and t indexes time. Y is a school level resource measure (such as per pupil expenditure or teacher-pupil ratio) or composition of students in a school; School is a set of variables capturing the characteristics of students, such as the percentage of students who are poor, the percentage of students who are immigrants, and
9
the percentage of students served by programs that impose higher costs on schools such as percent limited English proficient and a set of variables indicating the borough in which the school is located, Si is a school fixed effect, Zoneit is a vector of variables capturing the characteristics of the catchment area, B1 through B3are coefficients to be estimated and eit is an error term with the usual properties. We estimate a series of alternative models, including different sets of variables, in an effort to ‘tease out’ the key relationships. Before investigating the resources, we consider differences in the composition of students in schools. We begin by examining the representation of foreign born students in school. Are immigrants evenly distributed across schools or are there “immigrant schools”? To what extent do immigrant neighborhoods have immigrant schools? Does the relationship between neighborhoods and schools depend upon the other characteristics of the population – say, their race or education? What determines the percentage poor or LEP? To answer these, we match data on the schools for the 19992000 school year to data on the catchment zones from the 2000 Census. We next turn to models of the characteristics of the teachers and expenditures. A first set of regressions uses 2000 data on schools and zones. A second set of regressions explores the determinants of pupil-teacher ratios using data for 1990 and 2000 (data on other school resources were not available for 1990). This allows us to use school fixed effects, which means the coefficients are identified by changes in variables between 1990 and 2000. D. Results As shown in Figure 1, New York City’s elementary schools vary considerably in 10
the representation of immigrant students. While some have few immigrants, in others, more than one fifth of the students are foreign-born. At the same time, there is wide variation in the representation of immigrants in the catchment zones. As shown in Figure 2, although all zones have at least some immigrants (that is, there are no zones in the ‘zero percent’ group), some have few immigrants, and there are zones in which more than two thirds of the residents are foreign-born. As expected, the representation of immigrants in schools and their catchment zones are related. As shown in Figure 3, the percentage of immigrants in schools increases with the percentage in the zone. To be specific, the percentage of foreign born in the neighborhood schools increases by 0.27 points as the percentage of immigrants in the zone increased by one percent. (See column (1) of Table 5a.) That said, this relationship is not very “tight” - there is considerable variation in the percentage foreign-born in the school that is unexplained by the variation in the percentage in the zone. What else matters? As shown in column (2) of Table 5a, we introduce a set of variables capturing the characteristics of the population in the school zone, as well as a set of indicator variables for each of New York City’s five boroughs. The results are interesting. Other things equal, neighborhoods with higher percentages of blacks (or Hispanics) have schools with lower percentages of immigrants. On the other hand, the percentage of immigrants is increasing in neighborhood poverty and education. The key relationship between the representation of immigrants in the school and the zone is, however, unchanged. In column (3) we explore the possibility that this relationship is non-linear by adding the square of the percentage foreign-born. As shown, we find some evidence that the link between the school and neighborhood immigrant community is
11
tighter at higher levels. The implication is that immigrants are, indeed, somewhat concentrated in schools and the representation of immigrants in the community shapes the student body in the neighborhood school. Further, as shown in Figure 4, comparing the composition of catchments zones in 1990 and 2000 reveals that the representation of immigrants increased in most zones. Thus, the impact of immigration on the composition of schools is distributed unevenly and understanding the impact on the native-born requires examining the impact on resources. How do other student characteristics vary with the composition of the community? As shown in Table 5b, we investigate the percentage of students who are LEP, poor, white, black, Hispanic and Asian. Interestingly, the results suggest that larger shares of immigrants in the catchment area mean a higher proportion of students in the public school are LEP, poor and Asian, a smaller share are black or white, but there is no significant relationship to percentage Hispanic. As shown in Table 6, we investigate four measures of school resources - the pupil-teacher ratio ( (in columns (1) and (2)); the percentage of teachers with five or more years of experiences (in columns (3) and (4)), the percentage of teachers with master’s degrees in columns (5) and (6)) and expenditures per pupil (in columns (7) and (8)). For each, the first regression includes only zone variables; the second adds variables capturing the characteristics of the students. To begin, we find that the pupil-teacher ratio – a rough measure of class size -increases with the percentage of foreign-born in the catchment zone, holding other zone characteristics constant. The size of the effect is not large (.04) but suggests that having
12
a ten percentage point higher representation of immigrants means having .4 more pupils, on average, in each class. (The average pupil-teacher ratio is roughly 16.) Interestingly, other population characteristics matter, too. The pupil-teacher ratio increases with the percentage Hispanic and percentage Asian, but decreases with the percentage speaking no English and the percentage below poverty level. Adding school variables changes the magnitudes of the results, indicating that some of the relationship is driven by the characteristics of the school, rather than the community. As an example, pupil-teacher ratios increase more with increases in the percentage of the students who are foreign-born in the school than the zone population. The negative relationship with limited English proficiency, poverty and special education are consistent with explicit funding formulae and policies that are intended to compensate for costly educational needs. Turning to teacher experience, however, we find no relationship between experience and the representation of the foreign-born in either school or community, although the percentage of teachers with Masters degrees is consistently increasing in the percentage of the foreign-born in both the neighborhood and the school. Other variables have little impact, except for race variables. Teacher experience and education are decreasing in the percentage black and Hispanic and increasing in the percentage Asian. Does the greater education of the teachers balance their smaller number to even up resources? Not quite. Expenditures models (in columns (7) and (8)) show that spending per pupil decreases with the percentage of immigrants, driven primarily by the immigrants in the school rather than the community. At the same time, spending is increasing in the percentage of students who are limited English proficient, poor and/or in special education and in these models, the impact of race is substantially muted. Taken
13
together, these regressions suggest that immigrants may depress school resources, with consequences both for their own education and the education of their native-born school mates. That said, it is possible that this relationship isn’t causal – and, particularly, that omitted variable bias taints the regressions. To investigate this, we use data for the 1990 and 2000 school year and matching Census data and estimate a model of the pupilteacher ratio that includes school fixed effects. In doing so, we hope to ameliorate the impact of any omitted time invariant variables. Unfortunately, we are unable to obtain data on the immigrants in schools for 1990 and so this variable is excluded. As shown in Table 7, models estimated using only the 1990 data are quite similar to those estimated for 2000. The pooled model finds no statistically significant zone variables. That is, only the school characteristics seem to matter and, particularly, the size of the school. III.
Immigrants and School Outcomes The centerpiece of our work on school outcomes is a set of education production
function (EPF) models estimated using student-level longitudinal data. After a brief review of the copious literature on education production functions in general, we discuss ways in which immigrant concentrations, in particular, might affect the performance of native-born and immigrant students. We then describe the data used, the models we estimate, and results. A. Brief Literature Review A significant body of research focuses on demographic differentials in student achievement. These studies often draw on education production function models to 14
explain differences in academic performance across students and schools (see Eric Hanushek, 1986; Petra Todd and Kenneth Wolpin, 2003). Studies using this model consistently highlight the importance of family characteristics to student achievement (e.g., Meredith Phillips et al., 1998; Petra Todd and Kenneth Wolpin, 2007). In particular, many of these studies seek to explain achievement differences across racial and/or ethnic groups (e.g., Phillips et al., 1998; Todd and Wolpin, 2007; Roland Fryer and Steven Levitt, 2004, 2006). Using data from the Texas Schools Project data, Rivkin et al. (2005) conclude that teacher quality has a strong influence on student achievement, yet achievement is not related to traditional measures of teacher quality such as experience and education. Also using TSP data, Hanushek and colleagues found segregation (2006), student mobility (2003), and special education (2002) also matter to student outcomes. While a large body of literature examines differential achievement across race and ethnicity, very few quantitative studies explore differences in performance between native- and foreign-born students. Amy Schwartz and Leanna Stiefel (2006) investigate this “nativity gap” using an education production function model and conclude that disparities in academic performance across foreign- and native-born students are largely explained by student and school characteristics. Others examine the performance of immigrants by generation, generally finding that the first generation foreign-born perform better than native-born students, but that the second generation do not always (See Grace Kao, 1999; Grace Kao and Marta Tienda, 1995; Glick and White, 2003). B. How Might Immigrants Affect Performance of Immigrants and Native-Born Students?
15
While the EFP literature develops a framework for understanding variation in student performance in general, our interest in this paper is discovering how the presence of immigrants, in particular, affects their own performance and that of their native-born classmates. Because so little work has been done in this area, we briefly discuss the ways in which such presence might matter. There are two general mechanisms through which concentrations of immigrant students could affect student performance – through school and community resources and through their peers (which could be considered a kind of resource). Variations in the proportions of immigrants might affect the level or kind of resources that schools are allocated or can garner. As shown previously, in NYC, schools in disproportionately immigrant catchment zones and schools with higher proportions of immigrants have larger pupil-teacher ratios and better educated teachers (see Table 6). In addition to such between school differences, there might be impacts due to the allocation of resources within schools. For example, resources might be channeled to educate immigrants in special language classes or to mount special programs to help acculturation, or, alternatively be used in programs that benefit native-born more than foreign-born students. More generally, resources in neighborhoods devoted to youth (after-school services or libraries etc.) might differ depending on the presence of immigrants and/or might be devoted disproportionately to immigrants (native-born) at the expense of the native-born (immigrants). These latter two mechanisms (within school allocations and outside school services) are noted here but not pursued empirically, although we come back to these possibilities in the discussion below.
16
As noted earlier, immigrants might affect academic performance through their interactions as peers. Of course, immigrants might be better or worse peers for nativeborn and/or for foreign-born students. The popular image of immigrants is that they are particularly motivated students with high academic aspirations who receive encouragement from their families and communities. These traits could boost peer performance by direct inter-student “tutoring” or some other way. If, on the other hand, immigrants have poor English or other academic skills, or limited family and community resources, they may be poor peers. To some extent, then, the impact of the immigrants may depend on their own characteristics and, perhaps, their interactions with other students. The difficulty in estimating causal effects of peers (including racial peers) is well known and not reviewed here (see Hanushek, Kain and Rivkin, 2009 for a good review of the literature and some evidence). Although, we do not separately estimate such causal effects in our models, in discussion of our empirical results, we call on the possibility of this effect in explaining some findings. C. Data We use administrative data from the NYC Department of Education on all 3rd through 8th grade students in NYC public schools between 1997 and 2002. We restrict our sample to include students who enrolled in at least two years, which allows for student fixed effects in estimations, and who have at least one standardized test score (in reading and/or math). Our final sample has 665,408 unique students, yielding a total of 2,265,062 student observations. The data include detailed demographic and academic information for each student, including birthplace, race, gender, English language ability, language spoken at 17
home, poverty (measured by free lunch eligibility), resource room participation, and scores on standardized reading and/or math tests. We define a student as foreign-born if his or her birthplace is not in the United States. Of the students in our sample, 17.77% are foreign-born. Further, as shown in the top panel of Table 8, the foreign-born are nearly three times as likely to be Asian and 16 percent less likely to be black, relative to their native-born peers. In 2000, foreign-born students are roughly four times as likely to be LEP, over twice as likely to live in a home where English is not spoken, and half as likely to receive resource room services. Also, we see that immigrants have higher z-scores than the native-born in both math and reading. This is consistent with previous research showing that foreign-born students outperform than native-born peers. As we are interested in exploring whether not only the presence, but also the composition of foreign-born students in schools affects resources and performance, we examine student summary statistics by nativity and race (shown in Table 9). While not surprising given past research, the statistics provide strong evidence that school resources and composition vary significantly – particularly between race groups. For all races, the foreign-born students, on average, attend larger schools; however there are large disparities between races in the education and experience of their teachers. The percentages of teachers who are licensed, who have been teaching for at least five years, and who have Masters degrees are all roughly ten percentage points higher for whites and Asians than for blacks and Hispanics, regardless of nativity (panel 2). Further, in the third panel, we note a gap of approximately seven percentage points in the exposure to foreign-born students in schools between native- and foreign-
18
born children overall. This gap is largest among whites: foreign-born whites attend schools where over a quarter of the students are foreign-born, whereas native-born whites attend schools that are 87 percent native. Exposure to students of the same race group is also highly evident: overall, schools attended by native-born students are roughly 16 percent white, 37 percent black, 36 percent Hispanic and 10 percent Asian, yet whites, blacks, and Hispanics all attend schools where over 50 percent of the students are of their race. Moreover while not a majority, Asian students attend schools that are almost a third Asian, which is still very racially concentrated particularly given the smaller representation of Asians in the overall student population. (Similar trends in racial exposure indices are also evident among the foreign-born.) There are also marked differences in the composition of the foreign-born students across race groups (panel 4). Whites attend schools where the foreign-born student population is under 10 percent LEP, over 70 percent white and Asian, and roughly 60 percent poor. While more likely to be LEP and poor, Asian students also attend school with a foreign-born population that is majority white and Asian. Black and Hispanic students though, attend school with foreign-born students who are over three-quarters black and Hispanic, over 85 percent poor, and, among Hispanics, roughly a third of whom are LEP. The significant disparities in the composition of schools and the foreign-born student population offer the opportunity to gain insight into the impact of immigrants on performance and the underlying reason. Finally, we note that the students in one’s grade or class may matter in a different way than the students in one’s school: the composition
19
of students in a school may drive differences in resource allocation, but one’s classmates may affect his or her motivation or learning environment. We explore this below. D. Models In principle, student academic performance in a given period is a function over past periods of family background (such as income, education of parents, numbers of siblings), education services (such as special education or language learning services), school resources (such as teacher quality or class sizes) including peers, and student attributes that contribute to student performance (such as ability, motivation, perseverance, health, family support and so on). Thus, an ideal model would include information on all of these over time, allowing us to disentangle the effects of current peers, resources, etc. on student performance. Clearly neither we (nor anyone else) have the full set of data to estimate such a model, but there are two broad ways to approach estimating. The most common is to use current-year resources and education experiences and family characteristics along with test scores from previous periods to “capture” and control for the past history of a student. A second method takes advantage of longitudinal data by including a student “fixed effect” or dummy variable for each student (previous test scores can also be included), thereby controlling all time-invariant characteristics of that student that may be creating bias in the model. Using student fixed effects controls for student natural abilities, unique schooling past, invariant family supports and so on, but means that coefficients are not estimated for any individual characteristics, such as race or gender, that do not change over time.3 We use both of these methods -- a “value added” specification with past test scores to examine how
3
Estimates of these coefficients can be formed using a random effects specification.
20
student fixed characteristics affect scores and then a fixed effects model to more fully control for student unobserved characteristics. We estimate a series of education production function models of the following general form:
(2) Testijt = β0 + β1FB i*%FBijt + β2NB i*%FBijt + β3 Socio ijt + β4Educ ijt + β5Y ijt + β6Testi,j,t-1 + ∫i + εijt where i, j, and t index student, school and year, respectively, Test is the student’s normalized score on a citywide math or reading test,4 FB (NB)is an indicator that takes a value of one if the student is born in a country outside (within) the United States, Socio is a vector of variables capturing the student’s poverty status (measured by eligibility for free or reduced price lunch), gender, age and race, Educ is a vector of variables capturing the student’s educational characteristics including English language abilities and participation in part-time special education programs, Y, as before, measures school resources, and Testi,j,t-1 and ∫i. are the lagged test score and student fixed effect, respectively. All models include dummies for year and grade and are estimated with robust standard errors, adjusted for within-school clusters. The particular focus of these models is the coefficient on the interactions between FB (NB) and the percent of immigrants in a school. The sign and size of this coefficient indicates how a one percentage point increase in the percent of immigrants in a school affects the test score, on average, of a NB (FB) student, controlling for socioeconomic, educational, school resource and other unobserved but time invariant individual attributes 4
A normalized test score is constructed for each test, grade and year so that the mean over all students is zero and the standard deviation is one.
21
of students. The percentage of immigrants is measured as a proportion of all students in a school (or grade) and for subgroups of students and of immigrants (poor, LEP, and by race). These subgroups are included in order to explore how the concentration of fellow students and immigrants affect performance. Finally, because reducing racial test score disparities are so important to U.S. policy makers, we estimate models separately by race to investigate subgroup differences. E. Results i.
Basic Models – The Effects of Concentrations of Immigrants on Nativeborn and Foreign-born Performance Table 10 displays the basic EFP results with three different specifications. In
columns (1) and (2), we include a one-year lagged test score to create a value added model (VAM). This specification allows us to retain student characteristics that are nontime variant, such as race and gender (as well as ones that are time-variant, such as resource room (part time special education) participation by year). Note that the coefficients on the student characteristic variables are consistent with the past empirical literature. Ceteris paribus, females (compared to males) perform worse in math but better in reading, while black and Hispanic students perform worse than white students, but Asians perform better, and poor students perform worse than non-poor ones. Students whose home language is not English perform better, but conditional on this, LEP students perform worse. Finally prior test scores have a large influence on current performance, with coefficients of approximately 0.7. Most importantly for this paper, in these first models, as the percentage of 22
foreign-born in a school increases, the performance of both native and foreign-born students declines, although this effect is statistically significant only for the foreign-born on reading and small (-0.002). As described earlier, an important concern in interpreting these coefficients, however, is the possibility that there are unobserved differences between students that bias the estimates. Thus, the models in columns (3) through (6), and in the rest of the paper, we use student fixed effects to control for unobserved ability and other timeinvariant individual characteristics. We consider these models to be strongest in terms of isolating the unbiased effects of concentrations of immigrants on performance. 5 Results in columns (3) and (4) show negative impacts for both the native and foreign-born of increasing percentages of immigrants, but statistically significant only for the native-born (-0.004 in both reading and math). Note, however, this impact may not be substantively important. A typical native-born student attended school with slightly more than 14.5% foreign-born students. For such a student, the effect of the concentration of foreign-born, ceteris paribus, is lower performance of .06 standard deviations (-0.004 * 14.5), which is a small effect. As a benchmark, the raw test score difference between white and black students in New York City is around 0.7 standard deviations and the regression adjusted gap is around 0.1 standard deviations. Thus, while the effect of a higher concentration of immigrants is negative for the native-born, the size of the effect is quite small. In these models, the effect on the foreign-born of the concentration of immigrants is also negative, albeit statistically insignificant.
5
Note that the models in columns (3) and (4) were also estimated as VAMs - with lagged test scores - with almost identical results for the coefficients on concentration of foreign-born (results available from authors).
23
Columns (5) and (6) add an interaction with the percentage of immigrants in a student’s grade (while retaining the percent immigrants in the school). Interestingly, the magnitude of the effect of the concentration in the school doubles and becomes significant for both the native and foreign-born. Further, the effect of higher concentrations of immigrants in a student’s grade is positive and statistically significant. What could explain these results? First, although we include controls for teacher-pupil ratios and other school resources (which vary across schools with the concentrations of immigrants) there may be some intra-school differences in resources that are not observed in our data. It could be, for example, that at the school level, immigrants divert resources to administration or other non-instructional programs, leaving less for direct classroom teaching. Conversely, at the grade level, measures may capture the impact of immigrants as classroom peers and immigrants may provide “good” peers that help both the native-born and the foreignborn to succeed. Thus, conditional on the percentage of immigrants in a school and their effect on school (or instructional) resources, a higher proportion of immigrants in one’s grade could be good for achievement. The full effect of an increase in immigrants on a student is the combined effect of the change in the grade and school composition. That is, the overall effect for particular students could be positive or negative, although if those concentrations are about equal, then on average, it is negative.6 ii.
How does the Composition of the Immigrants Matter?
6
There is some evidence from other work that the concentrations of immigrants vary by grade and year (Ingrid Ellen, Katherine O’Regan and Dylan Conger, 2008). We do not pursue this line of inquiry into the differences between school and grade concentrations in this paper, due to data and time restrictions, but we do note that effects at both levels appear to be present and that the negative school-level effect is larger than the positive grade-level effect.
24
Do higher concentrations of foreign-born who are learning English, poor, or of various races all have a similar effect on the native-born? And do some of these groups of foreign-born also affect immigrant students as well? To answer these questions, we add variables capturing the composition of the student body – percentage LEP, percentage poor, percentage black, percentage Hispanic and percentage Asian.7 In addition, these models include interactions with the NB and FB for percentages of foreign-born students who are LEP, poor and black, Hispanic or Asian. Thus we are able to determine if the school’s composition of students in general and then of the foreign-born specifically affect performance. We begin with poverty and LEP and then add the race variables. All models include student fixed effects and controls from the basic models in Table 10. As shown in columns (1) and (2), increases in the percentage LEP students have a very small negative effect on all students in math, while increases in the percentage poor students have an equally small but positive effect in both math and reading. For the native-born students, specifically, we find the same small negative effect of foreign-born (-0.004) as in previous models and a very slight, small counterbalancing positive effect of increases in poverty for math only (0.0005). The LEP composition of the foreign-born has no effect on the native-born indicating that once LEP students are accounted for in general, the percentage of foreign-born who are LEP makes no difference to the native-born. For the foreign-born students, conversely, the effects of higher percentages of foreign-born are slightly positive but very small in math (0.001), while the effects of higher concentrations of LEP and poor foreign-born affect both math and reading 7
White is the reference category.
25
negatively (between -0.001 and -0.002). These coefficients, though, are very small. Their negative effect could be an indication that foreign-born students learn (slightly) better when they go to school with native-born students. Columns (3) and (4) add racial composition of the school to the variables in the model. The addition does not much change the coefficients on percentages of LEP or poor in general or in the foreign-born population. The overall racial composition of the school affects performance of all students in isolated and small ways – negatively in math as the percentage of blacks increases (-0.001) and positively in reading as the percent Asian increases (0.001). In terms of the racial composition of the foreign-born specifically, there are some small, isolated negative effects on the native-born of higher proportions of black foreign-born (reading -0.001), Hispanic foreign-born (reading 0.001), and Asian (reading and math -0.001). On the foreign-born students, the small isolated effects work in the positive direction. iii.
Does the Impact Vary Across Races? We estimate models of the effect of concentration of foreign-born students on the
native-born and foreign-born test scores of black, Hispanic, Asian and white students. These results are shown in Table 12. For native-born students, the effects of increasing percentages of foreign-born are negative, but interestingly the effects are considerably smaller for Asian and white (not larger than -0.002) than for black and Hispanic students (between -0.004 and -0.005). Thus black and Hispanic students, who are the lowest scoring on average of the four race/ethnicity groups, may be the most affected by increasing immigrant shares. Moreover, from Table 9, we see that the average black native-born student attends school 26
with slightly less than the overall average of foreign-born students for native-born students (12.3% versus 14.5%), while the average Hispanic native-born student attends school with a larger percentage (15.9% vs. 14.5%). For foreign-born students, there are almost no statistically significant effects in the subgroup regressions. The one exception is for Hispanics, in reading, where the coefficient is -0.002. Combined with the result for native-born Hispanics, it appears that Hispanics in general are more disadvantaged than other races by higher concentrations of immigrant students. IV. Conclusions The New York City public schools educate an extraordinarily large and diverse population of immigrant students in a wide array of schools differing in resources, organizational features and students. The broad variation in the schools, students, and communities offers an excellent opportunity to examine whether - and how – resource availability is affected by differences across communities of immigrants and, subsequently, how schools’ ability to education native-born and foreign-born students is affected by those resource differences. Our findings suggest that community composition helps shape the student body – communities with higher percentages of immigrants have schools with larger immigrant shares. (See SIED project, conceptual model, Figure 1, path a.) Additionally, more concentrated immigrant communities have schools with slightly high teacher-pupil ratios, percentages of teachers with masters, and lower spending per pupil. Less predictably, we find the percentage of immigrant students in a school has a small, but significant, negative effect on native-born performance, which persists even when controlling for 27
differences across school in resources, the composition of the student body, and the composition of the foreign-born population specifically (percentage LEP, poor, black, Hispanic and Asian). (See SIED project, conceptual model, Figure 1, path d.) Estimates of the effect of the percentage of immigrants at the grade level, however, are positive, suggesting that the presence of immigrants may affect within school resources in a way that hurts performance but may serve as quality peers in the classroom. Finally, subgroup analysis also suggests that black and Hispanic native-born students may be more adversely affected by increasing percentages of immigrants than their white or Asian peers. Past research on the performance of immigrants and native-born students in NYC and elsewhere has focused on differences between the groups, generally finding that the foreign-born (first generation) perform better than the native-born, especially in the context of a value-added production function model. Some researchers argue this means the foreign-born are high quality peers, but our results indicate that a more nuanced situation exists. School-wide, higher concentrations of immigrants affect the native-born negatively, but at the grade level, this effect is reversed. For foreign-born students, school-wide concentrations of immigrants are not as strong or uniform a negative influence on test scores. We proposed possible explanations for these findings in our discussion of results, but there remains much important work to do in order to understand why this happens, especially for the native-born students. In future work, looking inside the school seems to be a promising path. We would benefit from knowledge on how immigrants are distributed across grade levels and whether that distribution matters. Additionally, does it mater from where immigrants
28
originate – do concentrations of foreign-born students from different parts of the world have different effects and why? If they do, is it because they enter the school system predominantly at different grade levels or with different levels of English language proficiency or some other reason? In another vein, how are the foreign-born distributed across classrooms at the grade level – are they evenly distributed or segregated and how does that distribution affect the quality of teachers in the classroom available to foreignand native-born students? Finally, a longer perspective on student “lives” in school could be helpful. For this study, we include students in grades three through eight because these are the grades when students are tested in NYC (and nationally since the No Child Left Behind Act) and thus have performance outcomes. But how might immigrant concentrations and compositions in grades one and two influence performance later on? Unexpected results, as in this paper, present opportunities to further understand the ways in which neighborhoods affect school composition and how that composition affects student outcomes. We look forward to the results of such work.
29
References Behrman, Jere R. and Steven G. Craig. “The Distribution of Public Services: An Exploration of Local Governmental Preferences.” American Economic Review 77 (1987): 37-49. Bergstrom, Theodore C. and Robert Goodman. “Private Demands for Public Goods.” American Economic Review 63, no. 3 (1973): 280-296. Betts, Julian. R. and Magnus Lofstrom. “The Educational Attainment of Immigrants: Trends and Implications,” NBER Working Paper # 6757, 1998. Borjas, George J. “The Labor Demand Curve is Downward Sloping: Reexamining the Impact of Immigration on the Labor Market.” The Quarterly Journal of Economics 118, no. 4 (2003): 1335-1374. ---. “Do Foreign Students Crowd Out Native Students from Graduate Programs?” NBER Working Paper #10349, 2004. Burke, Sarah M. “An Analysis of Resource Inequality at the State, District, and School Levels.” Journal of Education Finance 24, no. 4 (1999): 435-458. Butcher, Kristen F. and Anne Morrison Piehl. “Recent immigrants: Unexpected implications for crime and incarceration.” Industrial and Labor Relations Review 51, no. 4 (1998): 654-679. Card, David. “Immigrant Inflows, Native Outflows, and the Local Labor Market Impacts of Higher Immigration.” Journal of Labor Economics 19, no. 1 (2001): 22-64. --- and Ethan G. Lewis. “The diffusion of Mexican immigrants during the 1990s: Explanations and impacts.” NBER Working Paper #11552, 2005. Clotfelter, Charles T., Helen F. Ladd and Jacob L. Vigdor. “Teacher-Student Matching and the Assessment of Teacher Effectiveness.” The Journal of Human Resources 41, no. 4 (2006): 778-820. Coleman, James S. “Equality of Educational Opportunity.” U.S. Office of Education, Washington D.C., 1966. Cortes, Patricia. “The Effect of Low-Skilled Immigration on U.S. Prices: Evidence from CPI Data.” Journal of Political Economy 116, no. 3 (2008): 381-422. Ellen, Ingrid G., Katherine O’Regan, and Dylan Conger. “Immigration and Urban Schools: The Dynamics of Demographic Change in the Nation’s Largest School District,” Education and Urban Society 41, no. 3 (2009): 295-316.
30
Friedberg, Rachel M. and Jennifer Hunt. “The Impact of Immigrants on Host Country Wages, Employment and Growth.” Journal of Economic Perspectives 9, no. 2 (1995): 23-44. Friesen, Jane and Brian Krauth. “Enclaves, peer effects, and student learning outcomes in British Columbia.” Unpublished paper, 2008. Fryer Jr., Roland G. and Steven D. Levitt. “Understanding the Black-White Test Score Gap in the First Two Years of School.” The Review of Economics and Statistics 86, no. 2 (2004): 447-464. --- and ---. “The Black-White Test Score Gap Through Third Grade.” American Law and Economics Review 8, no. 2 (2006): 249-281. Glick, Jennifer E. and Michael J. White. “The Academic Trajectories of Immigrant Youths: Analysis Within and Across Cohorts,” Demography, 40, no. 4 (2003): 759-783. Gould, Eric D., Victor Lavy and M. Daniele Paserman. “Does Immigration Affect the Long-Term Educational Outcomes of Natives? Quasi-Experimental Evidence.” IZA Discussion Paper #1883, 2005. Hanushek, Eric A. “The Economics of Schooling Production and Efficiency in Public Schools.” Journal of Economic Literature 24, no. 3 (1986): 1141–1177. ---, John F. Kain, and Steven G. Rivkin. “Inferring program effects for special populations: Does special education raise achievement for students with disabilities?” Review of Economics and Statistics 84, no. 4 (2002): 584-599. ---, --- and ---. “Disruption versus Tiebout improvement: The costs and benefits of switching schools.” Journal of Public Economics 88 (2003): 1721-1746. ---, --- and ---. “Teachers, schools and academic achievement.” Econometrica 73, no. 2 (2005): 417-458. ---, --- and ---. “New Evidence about Brown v. Board of Education: The Complex Effects of School Racial Composition on Achievement.” Journal of Labor Economics 27, no. 3 (2009): 349-383. Hertert, Linda. “Does equal funding for districts mean equal funding for classroom students? Evidence from California.” in Where does the money go? Resource allocation in elementary and secondary schools eds. Lawrence O. Picus and James L. Wattenburger (Thousand Oaks, CA: Corwin Press, 1995) 71. Hoxby, Caroline. “Do Immigrants Crowd Disadvantaged American Natives Out of Higher Education?” in Help or Hindrance? The Economic Implications of Immigration for African–Americans, eds. Daniel S. Hamermesh and Frank D. Bean (New York: Russell Sage Foundation, 1998), 282. 31
Iatarola, Patrice and Leanna Stiefel. “Intra-district equity of public education resources and performance.” Economics of Education Review 22, no. 1 (2003): 69-78. Jensen, Peter and Astrid Wurtz Rasmussen. “Immigrant and Native Children’s Cognitive Outcomes and the Effect of Ethnic Concentration in Danish Schools.” Study paper no. 20, University Press of Southern Denmark, 2008. Kao, Grace. “Psychological Well-Being and Educational Achievement Among Immigrant Youth,” in Hernandez Donald J., ed., Children of Immigrants: Health, Adjustment, and Public Assistance, National Academy Press, Washington D.C. 1999. --- and Marta Tienda. “Optimism and Achievement: The Educational Performance of Immigrant Youth,” Social Science Quarterly, 76, no. 1 (1995): 1-19. Lankford, Hamilton, Susanna Loeb and James Wyckoff. “Teacher Sorting and the Plight of Urban Schools: A Descriptive Analysis.” Educational Evaluation and Policy Analysis 24, no. 1 (2002): 37-62. Neymotin, Florence. “Immigration and its Effect on the College-Going Outcomes of Natives.” Economics of Education Review 28, no. 5 (2009): 538–550. Ottaviano, Gianmarco I. P. and Giovanni Peri. “Rethinking the Gains from Immigration: Theory and Evidence from the U.S.” NBER Working Paper # 11672, 2005. --- and ---. “Rethinking the Effects of Immigration on Wages.” NBER Working Paper #12497, 2006. Ousey, Graham C. and Charis E. Kubrin. “Exploring the Connection Between Immigration and Violent Crime Rates in U.S. Cities, 1980-2000.” Social Problems 56, no. 3 (2009): 447-473. Peri, Giovanni and Chad Sparber. “Task Specialization, Immigration, and Wages.” Center for Research and Analysis of Migration (CReAM) Discussion Paper Series # 00802, 2008. Rubenstein, Ross, Amy Ellen Schwartz, Leanna Stiefel and Hella Bel Hadj Amor. “From districts to schools: The distribution of resources across schools in big city school districts.” Economics of Education Review 26, no. 5 (2007): 532-545. Schwartz, Amy Ellen, Leanna Stiefel and Dae Yeop Kim. “The Impact of School Reform on Student Performance: Evidence from the New York Network for School Renewal Project.” Journal of Human Resources 39, no. 2 (2004): 500-522. --- and Alec Ian Gershberg. “Immigrants and Education: Evidence from New York City.” 32
in Proceedings from the Annual Meeting of the National Tax Association (Washington, D.C., National Tax Association, 2001). --- and Leanna Stiefel. “Immigrants and the Distribution of Resources within an Urban School District.” Educational Evaluation and Policy Analysis 26, no. 4 (2004): 303-328. --- and ---. “Is there a Nativity Gap? The Achievement of New York City Elementary and Middle School Immigrant Students.” Education Finance and Policy 1, no. 1 (2006): 17-49. Todd, Petra and Kenneth Wolpin. “Towards a Unified Approach for Modeling the Production Function for Cognitive Achievement.” Economic Journal 113 (2003): 3-33. --- and ---. “The Production of Cognitive Achievement in Children: Home, School, and Racial Test Score Gaps.” Journal of Human Capital 1, no. 1 (2007): 91-136.
33
Table 1: The foreign-born population of major metropolitan areas Total Population Total # foreign-born % foreign-born Miami 5,413,212 2,005,178 37.04% Los Angeles 12,875,587 4,488,563 34.86% New York 18,815,988 5,328,891 28.32% Houston 5,629,127 1,204,817 21.40% Washington DC 5,306,125 1,088,949 20.52% Dallas 6,144,489 1,092,361 17.78% Chicago 9,522,879 1,679,074 17.63% Phoenix 4,179,427 736,068 17.61% Boston 4,482,857 713,529 15.92% Philadelphia 5,827,962 508,977 8.73% Detroit 4,467,592 388,920 8.71% United States 301,621,159 38,059,694 12.62% Source: 2007 American Community Survey 1-Year Estimates, MSA level
34
5-17 year olds Total # foreign-born % foreign-born 880,718 121,182 13.76% 2,426,863 233,309 9.61% 3,189,054 288,978 9.06% 1,122,641 101,324 9.03% 920,332 78,026 8.48% 1,198,356 94,387 7.88% 1,771,832 109,966 6.21% 792,792 71,382 9.00% 731,199 43,479 5.95% 1,026,621 36,367 3.54% 836,553 30,848 3.69% 53,237,254 2,673,346 5.02%
Table 2a: NYC School Zones over time
immigrant white black Hispanic Asian poor over 65 under 18 HS plus avg income, in $1,000s (2000 $s) total population Observations
mean 27.29% 36.73% 31.30% 25.75% 5.72% 23.08% 12.34% 25.42% 62.71% 177.82 10,704
1990 min 2.16% 0% 0% 1.15% 0% 1.97% 2.49% 6.42% 28.52% 7.48 669 594
max 70.08% 98.47% 96.21% 87.87% 79.43% 65.61% 38.17% 48.01% 95.87% 2,307.52 85,639
mean 33.49% 30.70% 30.21% 28.86% 9.32% 23.97% 11.40% 26.06% 68.09% 187.32 12,584
2000 min 4.56% 0% 0% 1.32% 0% 2.24% 3.55% 6.50% 34.03% 6.19 1,430 588
max 73.66% 95.67% 94.50% 89.85% 86.53% 63.06% 30.33% 49.78% 97.72% 2,749.99 85,908
change 2000‐1990 6.20% ‐6.04% ‐1.09% 3.11% 3.60% 0.89% ‐0.94% 0.63% 5.38% 9.49 1,880
Source: Neighborhood Change Database, 1990 and 2000
Table 2b: Descriptive Statistics, NYC Elementary Schools, 1990 and 2000 1990 mean min max enrollment 757 25 2,008 white 22.2% 0% 96% black 36.0% 0% 98.4% Hispanic 34.4% 1.2% 100% Asian 7.4% 0% 91% LEP 12.7% 0% 49.4% poor 61.5% 0% 100% immigrant n/a n/a n/a Observations 628
mean 799 17.2% 34.3% 37.1% 11.3% 14.2% 74.3% 7.0%
Source: New York City Department of Education data, 1990 and 2000
35
2000 min 87 0% 0.1% 1.3% 0% 0% 6.7% 0% 633
max 2,200 94.7% 97.3% 98% 92.5% 54.8% 100% 26.7%
mean 778 19.7% 35.1% 35.8% 9.4% 13.5% 67.9% 7%
overall min 25 0% 0% 1.2% 0% 0% 0% 0% 1261
max 2,200 96% 98.4% 100% 92.5% 54.8% 100% 26.7%
Table 3: Enrollment in public and private schools, major US cities K‐8 enrollment 9‐12 enrollment public school private school public school private school Philadelphia 75.35% 24.65% 79.82% 20.18% New York City 77.92% 22.08% 81.29% 18.71% Boston 78.97% 21.03% 86.23% 13.77% Washington DC 80.15% 19.85% 81.56% 18.44% Chicago 85.21% 14.79% 85.02% 14.98% Los Angeles 87.77% 12.23% 89.57% 10.43% Dallas 88.13% 11.87% 88.55% 11.45% Miami 91.97% 8.03% 91.58% 8.42% Detroit 92.11% 7.89% 94.31% 5.69% Houston 92.41% 7.59% 90.97% 9.03% Phoenix 92.69% 7.31% 92.62% 7.38% United States 88.66% 11.34% 90.50% 9.50% Source: 2005‐2007 American Community Survey 3‐Year Estimates, Principle City
Table 4: Immigrant parents and their children, 2005‐2006 Immigrant Parents Immigrant Children Native‐born Children New York State 15.58% 84.42% United States 16.66% 83.34% Source: Urban Institute Children of Immigrants Data Tool
36
Table 5a: The determinants of foreign-born representation in schools, 2000 Dependent variable: % foreign-born students in the school
Zone Characteristics % foreign-born
(1)
(2)
(3)
0.271*** (0.010)
0.275*** (0.017)
-2.041*** (0.292)
-0.044*** (0.008) -0.043*** (0.015) -0.010 (0.022) 0.026 (0.043) -0.034 (0.046) 0.031 (0.044) 0.002 (0.001) 0.146*** (0.026) 0.102*** (0.029) -11.13*** (3.426)
0.201*** (0.040) 0.001** (0.001) -0.046*** (0.008) -0.040*** (0.015) -0.011 (0.022) 0.033 (0.043) -0.023 (0.046) 0.016 (0.044) 0.002* (0.001) 0.142*** (0.026) 0.010*** (0.029) -9.997*** (3.510)
N 633 0.581
Y 633 0.701
Y 633 0.703
% foreign-born squared % Black % Hispanic % Asian % over 65 % under age 18 % speaking no English average household income % below poverty level % with at least a HS degree Constant Borough fixed effects Observations R-squared
Robust standard errors, adjusted for within-zone clusters, in parentheses (*** p
