Abstract

We use matched birth records and longitudinal student records in Florida to investigate whether first-, second-, and third-generation Asian and Hispanic immigrants have different educational success (measured by test scores, disciplinary problems, truancy, high school graduation, and college readiness). We find that, for both Asian and Hispanic students, early-arriving first generation immigrants perform better than do second-generation immigrants, who perform better than third-generation immigrants. The earlier the arrival, the better the students tend to perform. There is therefore a general pattern of successively reduced achievement in the generations following the generation that immigrated to the United States.

1.  Introduction

Over the last five decades, the United States has experienced the second largest wave of immigration in its history. Triggered by the enactment of the Immigration and Nationality Act of 1965, this immigration flow gained further momentum in the 1990s and changed the demographic composition nationwide. The last thirty years have averaged more than one million documented immigrants per year, and, unlike the earlier waves of immigration, the majority of these immigrants have come from Latin America and Asia. Immigrants and children of immigrants currently account for nearly a quarter of all school-aged children in the United States, and are projected to account for one third by 2050 (Passel 2011). How this new wave of immigrant youth fare in the U.S. public education system, therefore, has significant short-term and long-term welfare implications.

This paper presents the first use of matched administrative school and natality records to offer the most comprehensive analysis to date of the educational experiences of immigrants in the U.S. public school system. These data permit, for the first time, the study of a wide range of student outcomes across first-, second-, and third-generation immigrants in the United States, including a detailed breakdown of outcomes by the timing of first-generation students’ entry into the United States. The rich data we use from Florida administrative records provide the opportunity to follow hundreds of thousands of immigrant children longitudinally through school. In particular, we investigate across immigrant generations the gaps in student achievement, misbehavior, truancy, persistence in high school, and college readiness upon graduation, and explore how these gaps evolve across grades. By considering a variety of outcomes—both traditional academic outcomes as well as nonacademic outcomes—we are able to observe how successive generations of immigrants fare in terms of both human capital accumulation and social factors that could relate to future life chances and success. We then examine the extent to which these gaps are driven by observable education production-function inputs that are typically out of the control of the school system (such as socioeconomic differences and potentially malleable factors such as differences in school quality).

Theoretically, the implications of the prior literature on immigration is ambiguous as to how well recent immigrants would fare in the U.S. public school system compared to their second- and third-generation peers. On one hand, the straight-line assimilation theory, which was introduced by Park (1914) and later extended by Gordon (1964), predicts improved outcomes across immigrant generations, primarily driven by the wealth and education accumulation of second and higher generations.1

On the other hand, many have argued that this theoretical framework, which was developed based on the immigrant experiences at the turn of the 19th century, may not be applicable to the new wave of immigration because of fundamental differences in transportation and communication technologies, and the cultural, linguistic, and socioeconomic diversity of communities in which these new immigrants settle.2 For instance, Portes and Zhou (1993) propose the segmented assimilation theory, arguing that immigrants may take divergent assimilation paths depending on the attributes of host communities they face upon arrival to the United States. These paths may include downward assimilation (which predicts deteriorating outcomes across immigrant generations) as well as selective acculturation (which predicts improving outcomes across generations while preserving the values of the immigrant community), and straight-line assimilation.3

Empirically, there has been a fair amount of research on the educational attainment of immigrants compared with natives in the United States,4 but the extant literature lacks a comprehensive analysis of educational outcomes and experiences of first-, second-, and third-generation immigrants while in school due to limitations in the existing data—in particular, due to the inability to identify immigrant generation in school records. These “intermediate” educational outcomes are important to understand because they could serve as early indicators of disparities in later human capital formation and labor market outcomes. For instance, Schwartz and Stiefel (2006) and Conger, Schwartz, and Stiefel (2007) examine disparities between first-generation immigrant students and their native peers in New York City public schools along reading and math test scores, as well as non-test outcomes including student attendance, mobility across schools, and participation in special education programs. They find that immigrants—especially those who enter the school system in earlier grades—outperform their native peers on reading and math tests, and have lower absence and special education participation rates.5 Similarly, in a difference-in-differences framework, Stiefel, Schwartz, and Conger (2010) isolate the effect of immigration on student outcomes from that of student mobility by comparing the educational outcomes of teen immigrants with native migrants who switch school districts in high school. Their findings indicate that teen immigrants outperform observationally comparable native migrants in terms of high school graduation and test scores.

While these studies provide valuable information about the educational experiences of first-generation immigrants in the public school system, they potentially mask important cross-generational differences as they fail to distinguish between second- and third- (and higher) generation immigrant students. Another strand of literature makes use of survey data that typically provide detailed information about the children's families to make comparisons between first-, second-, and third- (and higher) generation immigrants (Kao and Tienda 1995; Portes and MacLeod 1996; Kao 1999; Glick and White 2003; Cortes 2006). For instance, Kao and Tienda (1995) and Glick and White (2003) use the National Educational Longitudinal Survey of 1988 (NELS-88) to examine the cross-generational differences in course grades, test scores, and college aspirations in high school, and find that both first- and second-generation immigrant students earn higher grades, have higher test scores, and express higher educational aspirations than their third- (or higher) generation immigrant peers. Similarly, Galindo and Reardon (2006) and Reardon and Galindo (2009) use the Early Childhood Longitudinal Survey-Kindergarten Cohort (ECLS-K) to study these differences among Mexican immigrants in early years of schooling. They find that whereas recent immigrants (first- and second-generation) typically come from more disadvantaged backgrounds and have worse English skills when they enter the school system, they show the greatest achievement gains in elementary school. However, such studies that make use of survey data typically contain a limited number of student outcomes (and no high-stakes outcomes), no information about the schools these students attend, and are unable to follow individual student outcomes over time. Further, these studies commonly suffer from small sample sizes, which hinder subgroup analysis and further complicate understanding potential mechanisms through which the relationships may be operating. For instance, due to data limitations, Glick and White (2003) can only distinguish between “preschool immigrants” and all others; and Reardon and Galindo (2009) can only look at first-generation immigrants of Mexican origin who enter the United States before kindergarten, and neither of these studies has been able to examine Asian immigrants separately. Table 1 provides a summary of some of the earlier literature on generational differences in immigrant student performances and experiences in the United States.

Table 1.
Summary of Relevant Quantitative Research on Immigrant Student Performance in the United States
ArticleDataOutcomesComparison and dimensions of heterogeneityMain findings
Schwartz and Stiefel (2006)  Student-level school records in NYC Reading and math scores - Foreign-born versus native- Country of birth - Immigrants outperform natives- Disparities substantially explained by student, school characteristics 
Conger, Schwartz, and Stiefel (2007)  Student-level school records in NYC Attendance, school mobility, special education participation - Foreign-born versus native- Country of birth - Immigrants have higher rates of attendance, lower rates of special education participation, and higher rates of school mobility- Significant variation by country of origin 
Cortes (2006CILS, 1992—1993; CCD, 1992—1993 Reading and math scores - Foreign-born versus native- Age at arrival (less than 5 years, 5—9 years, 10 or more years in the U.S.) -Test score gap between foreign-born and native students decreases the longer immigrant children reside in the U.S.- Foreign born children in enclave schools perform as well as immigrant children that attend non-enclave schools 
Glick and White (2003HSB 1980, 1990; NELS-88, 1990, 1992 Dropout rates, reading and math scores - First-, second-, and third-generation immigrants- Age at arrival (preschool versus recent arrivals) - Immigrants in 1980s perform worse than native-born; in 1990s, they perform better in levels- In changes, immigrants about on par with native-born- Dropout rates same for immigrants and native-born 
Kao and Tienda (1995NELS-88 Math and reading scores, college aspirations - First-, second-, and third-generation immigrants - Both first- and second-generation immigrants outperform native-born 
Kao (1999NELS-88 High school GPA, math and reading scores - First-, second-, and third-generation immigrants - First- and second-generation immigrants do better than same-race native-born, and as well as or better than white native-born 
Portes and MacLeod (1996Children of Immigrants; Miami, Ft. Lauderdale, San Diego, 1996 Math and reading scores - Second-generation immigrants - Parents’ socioeconomic status, length of U.S. residence, and hours spent on homework significantly affects the second-generation immigrant students’ academic performance 
Reardon and Galindo (2009ECLS-K, KG class of 1998—99 Reading and math scores - First-, second-, and third-generation Mexican immigrants - Compared to non-Hispanic white students, Hispanic students enter kindergarten with much lower average math and reading skills-Students with Mexican and Central American origins, particularly first- and second-generation immigrants, have the lowest math and reading skill levels at kindergarten entry but show the greatest achievement gains in the early years of schooling 
Hull (2017Student-level school records matched to birth records in North Carolina Reading and math scores - First-, second-, and third-generation Hispanic immigrants - First- and second-generation Hispanic immigrants drive the improvement in Hispanic test scores (compared to white students) between grades 3 and 8 
ArticleDataOutcomesComparison and dimensions of heterogeneityMain findings
Schwartz and Stiefel (2006)  Student-level school records in NYC Reading and math scores - Foreign-born versus native- Country of birth - Immigrants outperform natives- Disparities substantially explained by student, school characteristics 
Conger, Schwartz, and Stiefel (2007)  Student-level school records in NYC Attendance, school mobility, special education participation - Foreign-born versus native- Country of birth - Immigrants have higher rates of attendance, lower rates of special education participation, and higher rates of school mobility- Significant variation by country of origin 
Cortes (2006CILS, 1992—1993; CCD, 1992—1993 Reading and math scores - Foreign-born versus native- Age at arrival (less than 5 years, 5—9 years, 10 or more years in the U.S.) -Test score gap between foreign-born and native students decreases the longer immigrant children reside in the U.S.- Foreign born children in enclave schools perform as well as immigrant children that attend non-enclave schools 
Glick and White (2003HSB 1980, 1990; NELS-88, 1990, 1992 Dropout rates, reading and math scores - First-, second-, and third-generation immigrants- Age at arrival (preschool versus recent arrivals) - Immigrants in 1980s perform worse than native-born; in 1990s, they perform better in levels- In changes, immigrants about on par with native-born- Dropout rates same for immigrants and native-born 
Kao and Tienda (1995NELS-88 Math and reading scores, college aspirations - First-, second-, and third-generation immigrants - Both first- and second-generation immigrants outperform native-born 
Kao (1999NELS-88 High school GPA, math and reading scores - First-, second-, and third-generation immigrants - First- and second-generation immigrants do better than same-race native-born, and as well as or better than white native-born 
Portes and MacLeod (1996Children of Immigrants; Miami, Ft. Lauderdale, San Diego, 1996 Math and reading scores - Second-generation immigrants - Parents’ socioeconomic status, length of U.S. residence, and hours spent on homework significantly affects the second-generation immigrant students’ academic performance 
Reardon and Galindo (2009ECLS-K, KG class of 1998—99 Reading and math scores - First-, second-, and third-generation Mexican immigrants - Compared to non-Hispanic white students, Hispanic students enter kindergarten with much lower average math and reading skills-Students with Mexican and Central American origins, particularly first- and second-generation immigrants, have the lowest math and reading skill levels at kindergarten entry but show the greatest achievement gains in the early years of schooling 
Hull (2017Student-level school records matched to birth records in North Carolina Reading and math scores - First-, second-, and third-generation Hispanic immigrants - First- and second-generation Hispanic immigrants drive the improvement in Hispanic test scores (compared to white students) between grades 3 and 8 

Notes: CILS = Children of Immigrants Longitudinal Survey; CCD = Common Core of Data; HSB = High School and Beyond; NELS = National Educational Longitudinal Study; ECLS-K = Early Childhood Longitudinal Program, Kindergarten Cohort; GPA = grade point average.

In this study, we make use of a unique dataset from Florida, in which student records are matched with birth certificate data for all children born between 1992 and 2002, in order to overcome the limitations of the prior literature and address a set of heretofore unanswered questions. Linking school records to birth records allows us to take advantage of the benefits of both administrative data and survey data in an environment in which we can observe outcomes for the entire population, and Florida, to date, is the only major immigrant-receiving state in the nation where it has proven feasible to match birth and school records. There are several distinct advantages that are realized as a consequence: First, as with survey data, our ability to match birth and school records allows us to identify second-generation immigrants, but unlike studies using survey data, we are able to break down the analysis along several important dimensions, such as age at entry into the school system.6 The birth–school match also enables us to contrast whether estimated cross-generational differences in educational outcomes are different depending on whether children are identified based on their stated ethnic identity or their parents’ stated ethnic identity, permitting us to reduce the likelihood that our results are driven by the self-identification bias identified by Duncan and Trejo (2016).

Second, using our administrative data, we can examine not only high-stakes outcomes like student test scores, high school graduation, and high school course-taking, but also noncognitive outcomes, such as disciplinary incidents and truancy. Given our emerging understanding of the strong importance of both cognitive and noncognitive skills (see, e.g., Heckman and Rubinstein 2001; Heckman, Stixrud, and Urzua 2006), a study that analyzes the emergence of different skill sets, across generations and across entry ages for first-generation immigrants, has significant importance for policy and educational practice.

Finally, our data allow us to take a closer look at the achievement gaps by race/ethnicity and how these gaps evolve over grades—questions that were addressed in Clotfelter, Ladd, and Vigdor (2009). In particular, using student-level data from North Carolina, Clotfelter, Ladd, and Vigdor find that Hispanic and Asian students gain on white students in reading and math tests as they progress between grades 3 and 8. Because we can follow approximately two million individual students over more than a decade, including hundreds of thousands of foreign-born children and children of immigrants, we are able to go one step further and explore whether these relative gains vary across first-, second-, and third-generation Hispanic and Asian students.7

We find a general pattern of deteriorating educational outcomes across successive immigrant generations. In particular, we find that first-generation immigrants—beyond a transition period—perform better in reading and math tests than do second-generation immigrants, and second-generation immigrants perform better than third-generation immigrants. We also find that recent immigrants are significantly more likely to graduate from high school than more established generations, and are better prepared for college upon high school graduation. This pattern also holds true for student misbehavior and truancy, and for both Asian and Hispanic immigrants, and remains unchanged even after controlling for observed student, family, and school attributes, or whether we use the student's or the mother's reported racial/ethnic identity. We also find the relative gains in achievement between grades 3 and 8 among Hispanics reported in Clotfelter, Ladd, and Vigdor (2009) are almost exclusively driven by first-generation immigrants. On the other hand, both first- and second-generation Asian immigrants gain considerably on white students in reading and math tests between third and eighth grades. Although our analysis does not directly test the validity of different immigrant assimilation theories (as we are unable to look at the differences in educational outcomes between immigrant children and their parents), these findings present evidence that immigrants in this new wave follow a distinct assimilation path than what is predicted by straight-line assimilation theory.

A possible mechanism behind the observed decline in student outcomes across immigrant generations is immigrant optimism/motivation fading across generations. We find indirect evidence supporting this hypothesis. In particular, we find significant cross-generational differences in the revealed educational preferences of Hispanic immigrants in elementary and high school, with second- and third-generation immigrants more likely to “undershoot” compared with observationally similar recent immigrants, presenting evidence of educational aspirations dissipating across generations.

Finally, from a public policy perspective, our study highlights the need for more parental information in administrative school records. Our findings reveal significant differences in the educational outcomes and experiences of second- and third-generation immigrants in the public school system—a distinction that has been typically missed in prior studies using school records—as these data do not contain information on parental place of birth. Such information could enable researchers and policy makers to better identify students in need and design better-targeted interventions to improve their outcomes.

2.  Data and Empirical Strategy

Data

To address our research questions, we merge together two large datasets from the state of Florida: student-level administrative data and birth records. School records provide detailed information on all students in grades K–12 enrolled in Florida public schools between the 2002–03 and 2011–12 school years. This information includes: performance on the high-stakes Florida Comprehensive Assessment Test (FCAT) in reading and math of all students between grades 3 and 10; courses taken in elementary, middle, and high school; student demographics, poverty, measures of English proficiency (limited English proficiency indicator and language spoken at home, as reported by parents); attendance; disciplinary incidents; and the country of birth for all students (this last variable enables us to identify first-generation immigrants in the population, as well as their countries of origin).

We merge these school records with the birth records of all children born in Florida between 1992 and 2002. The birth records identify maternal country of birth,8 permitting us to differentiate between second-generation immigrants and higher-order generations (which, for ease of explanation, we will call third-generation immigrants).9 Birth records also record maternal self-reported ethnicity, allowing us to reduce the likelihood that our results are driven by bias associated with cross-generational differences in self-reported ethnicity discussed by Duncan and Trejo (2016). Furthermore, birth records offer background information not typically seen in school records including maternal education and marital status. Because the birth–school match is based on full name, exact date of birth, and social security numbers, the birth and school records are virtually perfectly matched (Figlio et al. 2014).10 Because we can determine second- versus third-generation immigrant status only for native Florida children born between 1992 and 2002, we restrict the data for non-Florida-born children to those years of birth as well.

Florida is an ideal location in this context both because of its outstanding student longitudinal data system and also because of its position as one of the major destinations for the recent wave of immigration. Currently, foreign born individuals constitute one fifth of the population in Florida, and 9.2 percent of all foreign-born individuals and 8.5 percent of all foreign-born children in the United States, reside in Florida. Further, the composition of these immigrants closely resembles the new wave of immigrants nationwide: 85 percent of the current immigrant population in Florida was born in Latin America or Asia. In the 2009–10 school year, one tenth of all Florida K–12 public school students were born outside of the United States, and another quarter of Florida's students had foreign-born parents. Therefore, Florida can be regarded as a mirror for the future demographics of the public school system in the United States.

One limitation of using data from a single state is that the immigrants residing in that state may differ systematically from those living elsewhere in the country. We investigate this possibility directly using data from the American Community Survey, pooled across the years 2006–2014. We find that, indeed, Hispanic and Asian immigrants into Florida are positively selected (driven largely by the fact that as a fraction of total immigrants, Mexicans are relatively underrepresented in Florida versus the United States as a whole). However, the degree of positive selection is not particularly great across generations: First-generation Asian and Hispanic immigrants in Florida have 7.2 percentage point higher rates of maternal college education than their counterparts outside of Florida.11 For second-generation immigrants, the difference is 5.7 percentage points, and for third-plus–generation immigrants in Florida, the difference is 6.8 percentage points.12 Because of the apparent positive selection into Florida, driven largely by compositional differences, we stratify our analysis, to the extent possible, by country of origin to increase the external validity of our findings.

Figure 1 presents the distribution of immigrant generations by major race/ethnicities in Florida. In this graph, the first category represents the foreign-born students, the second category identifies students born in Florida to foreign-born mothers, the third category represents Florida-born students with U.S.-born mothers, and the fourth category denotes students born in another state in the United States. Overall, 71 percent of children in Florida public schools were born either outside of the United States or in Florida—a necessary condition for understanding second- versus third-generation immigrant status. Of those for whom it is possible to divine immigrant generation, about 14 percent are first-generation immigrants, 21 percent are second-generation, and 66 percent are third-generation (or longer ago) immigrants. The distribution among Hispanic and Asian students, on the other hand, reflect the recent wave of immigration into the United States, with the majority of Hispanic and Asian students being first- or second-generation immigrants. For instance, among Hispanics born either in Florida or abroad, 33 percent of the students are foreign-born and 47 percent have an immigrant mother. Among Asians born either in Florida or abroad, 43 percent of the students are foreign-born and 54 percent have an immigrant mother. One fifth of Hispanics and 3 percent of Asians for whom we can determine generation status are in the third or higher generation. In our main analysis, we exclude the students born in another state as we cannot identify their immigrant generation.13

Figure 1.

Student Distribution in Four Major Race/Ethnicity Categories by Immigrant Generation

Figure 1.

Student Distribution in Four Major Race/Ethnicity Categories by Immigrant Generation

Table 2 presents the distribution of country of birth among first-generation Hispanic and Asian immigrants, and portrays the diversity of the immigrant population in Florida. For instance, over 90 percent of the first-generation Hispanic immigrants in the sample were born in fourteen countries, with the majority coming from Cuba (22 percent), Mexico (17 percent), Colombia (12 percent), and Puerto Rico (10 percent). In any analysis of Hispanic immigrants, especially in a location such as Florida, there is the question of how to characterize those who were born (or whose parents were born) in Puerto Rico and Cuba. Ten percent (23 percent) of first-generation Hispanic immigrants and 11 percent (20 percent) of second-generation Hispanic immigrants in Florida are of Puerto Rican (Cuban) origin, so this is a meaningful decision in the present analysis. Immigrants from both countries tend to be more advantaged socioeconomically and contextually/linguistically than other Latino groups (Galindo and Reardon 2006) and have consistently outperformed Hispanic immigrants from other countries (ACT 2007), and, of course, all Puerto Ricans are themselves U.S. citizens, so their status as “immigrants” in the United States is somewhat different from immigrants from other places. In our main analysis, we include first- and second-generation Puerto Rican and Cuban immigrants. That said, all of our results reported below are virtually unchanged when we exclude these students from the analysis.14

Table 2.
Country of Birth among First-Generation Asian and Hispanic Immigrants
Hispanic StudentsAsian Students
Country of BirthPercent of First Generation HispanicsNumber of Unique StudentsCountry of BirthPercent of First Generation AsiansNumber of Unique Students
Cuba 22.18 41,964 Philippines 16.04 3,914 
Mexico 17.32 32,766 China 13.42 3,275 
Colombia 12.02 22,744 India 13.22 3,225 
Puerto Rico 10.86 20,541 South Korea 9.32 2,273 
Venezuela 7.31 13,827 Vietnam 8.02 1,956 
Dominican Republic 3.93 7,438 Pakistan 4.36 1,064 
Honduras 3.79 7,164 Japan 3.25 793 
Peru 3.74 7,082 Bangladesh 3.04 741 
Argentina 3.18 6,012 Thailand 2.50 610 
Ecuador 2.17 4,111 Canada 2.48 605 
Guatemala 2.01 3,804 Guyana 1.87 456 
Nicaragua 1.75 3,304 Burma 1.31 319 
Brazil 1.37 2,586 Taiwan 0.99 241 
Spain 1.35 2,549 Other 20.19 4,926 
Other 7.04 13,314    
Hispanic StudentsAsian Students
Country of BirthPercent of First Generation HispanicsNumber of Unique StudentsCountry of BirthPercent of First Generation AsiansNumber of Unique Students
Cuba 22.18 41,964 Philippines 16.04 3,914 
Mexico 17.32 32,766 China 13.42 3,275 
Colombia 12.02 22,744 India 13.22 3,225 
Puerto Rico 10.86 20,541 South Korea 9.32 2,273 
Venezuela 7.31 13,827 Vietnam 8.02 1,956 
Dominican Republic 3.93 7,438 Pakistan 4.36 1,064 
Honduras 3.79 7,164 Japan 3.25 793 
Peru 3.74 7,082 Bangladesh 3.04 741 
Argentina 3.18 6,012 Thailand 2.50 610 
Ecuador 2.17 4,111 Canada 2.48 605 
Guatemala 2.01 3,804 Guyana 1.87 456 
Nicaragua 1.75 3,304 Burma 1.31 319 
Brazil 1.37 2,586 Taiwan 0.99 241 
Spain 1.35 2,549 Other 20.19 4,926 
Other 7.04 13,314    

Notes: The distribution of country of birth among first-generation immigrants is obtained using the set of students who were born outside of the United States between 1992 and 2002, and subsequently attended eighth grade in a Florida public school between 2002—03 and 2011—12 school years.

Empirical Framework

We are interested in five distinct student outcomes: (1) FCAT scores in reading and math standardized to zero mean and unit variance at the grade-year level; (2) disciplinary problems15 (as measured by whether the student was involved in a disciplinary incident that led to an in-school or out-of-school suspension); (3) attendance (percent absent days); (4) on-time high school graduation (whether the student received a standard high school diploma within four years after entering the ninth grade for the first time); and (5) college readiness, as measured by whether the student took a college credit–bearing course16 before high school graduation. All these K–12 outcomes and experiences have been shown to be predictive of future human capital formation and labor market success for students. For example, there is considerable evidence indicating that cognitive skills measured by test scores and high school graduation are directly related to individual earnings and productivity.17 Similarly, a number of studies have shown strong correlation between misbehavior and truancy in middle and high school and postsecondary success (e.g., Rumberger 1995; Rumberger and Larson 1998; Balfanz, Herzog, and Mac Iver 2007; Balfanz 2009). Finally, in a recent study, Smith, Hurwitz, and Avery (2017) find a significant positive effect of taking and passing Advanced Placement courses in high school on receiving a bachelor's degree in four years. Therefore, examining the cross-generational differences in these K–12 outcomes is not only important to better understand the disparities among immigrants while they are in the public school system, but could also speak to inequalities in future earnings and well-being.

Along these five dimensions, we compare first-, second-, and third-generation Hispanic and Asian immigrants, with third-generation white students (that is, white students whose mothers were born in the United States) as the baseline group.18 In the case of student test scores, disciplinary incidents, and attendance, we measure these outcomes in the eighth grade so it is more reasonable to assume that the estimated differences between early entering first generation and later generations can be interpreted as relatively permanent. This restriction leaves us with seven birth cohorts of students who are old enough to reach the eighth grade by the end of our sampling frame (students born between 1992 and 1998). For the high school outcomes, we are limited to those students born between 1992 and 1994, because of the years of data availability.

We execute our analysis by dividing individuals into eleven groups—nine groups of first-generation immigrants who first arrive in the school system in each grade from eighth (latest-arriving immigrants) to kindergarten (earliest-arriving immigrants);19 one group of second-generation immigrants whose mothers were born outside of the United States; and one group of third-generation “immigrants” whose mothers were born in the United States but who identify as having either an Asian identity or a Hispanic identity. Because of the possibility of bias in self-identification of Asians and Hispanics (described by Duncan and Trejo 2016), we identify the set of second- and third-generation immigrants based either on child's ethnic identification in school or mother's ethnic identification at the time of the child's birth.

We present four basic sets of analyses. In the first set of models, we show raw comparisons across the generations, and for first-generation students, across age at entry, controlling only for disability status, age at eighth grade, year, and gender. In the second set of models, we include a set of control variables: Because immigrants might attend systematically different sets of schools depending on generation, we control for school by cohort fixed effects20 so that we are directly comparing immigrants from different generations attending the same schools. In alternative models, we control for elementary school fixed effects as these are likely to provide a more granular control for neighborhood quality. We also control for a measure of low-income status (free or reduced-price lunch [FRPL] eligibility) during the eighth grade (for test scores, disciplinary incidents, and truancy) or during the ninth grade (for high school graduation and course-taking), or during the first year we observe the student in the school system because eligibility for recent immigrants might change as they integrate into the society. In the third set of models, we control for a measure of language minority (whether or not English is spoken at home) to gauge the extent to which language accumulation explains the differences in student outcomes across immigrant generations.21 It is important, however, to note that there is little common support between first-generation immigrants (especially Hispanics) and third-generation white students along this dimension, and controlling for this measure might lead to misleading gaps in student outcomes between these two groups. Therefore, we introduce this measure separately from school fixed effects and low-income status. In the fourth set of models, we further control for a set of family background characteristics observed on the birth certificate—maternal education and marital status. Our objective in presenting these multiple sets of models is to show the degree to which cross-generational differences, and differences across first generation students depending on age at entry, are sensitive to whether or not we control for variables that can be correlated with immigrant generation or perhaps age at entry. We cluster all standard errors at the school-by-year level.

In addition to studying cross-generational differences in eighth-grade levels of outcomes, we also follow individual students over time. We examine how these gaps evolve between third and eighth grades using early-entering first-, second-, and third-generation immigrants.

3.  Results

Tables 3 and 4 present the descriptive statistics for our five outcomes of interest and student/family attributes, comparing third-generation white students with Hispanic students by generation in table 3, and third-generation white students with Asian students by generation in table 4. For ease of tabular comparison, we present descriptive statistics just for early entering (those who entered the public school system by the third grade) versus late entering first generation students. These raw statistics reveal that outcomes differ, sometimes dramatically, across generations, yet the same general pattern is observed for both Asian and Hispanic immigrants, with worsening educational outcomes as we move from recent immigrants to more established immigrant generations beyond a transitional period for first-generation immigrants. For instance, foreign-born Hispanic students who enter the school system after the third grade score roughly 90 percent of a standard deviation worse in eighth-grade reading tests (roughly 0.6σ worse in math) than third-generation white students, are 16 percentage points (30 percent) less likely to graduate from high school, and are 13 percentage points (25 percent) less likely to take a college credit-bearing course before graduating from high school. In stark contrast, the Hispanic gap in test scores shrinks to about 0.2σ in reading and math for early entering first generation Hispanics, and goes up to 0.28σ for second-generation Hispanics, and further increases to about 0.3σ for third-generation (or higher) Hispanics. Similarly, recent Hispanic immigrants (early entering first- and second-generation) have more favorable disciplinary outcomes and attendance records, are significantly more likely to graduate from high school, and are more likely to take an AP, IB, or AICE course before graduating from high school than third-generation immigrants. The same patterns are observed for Asians, with more favorable outcomes than third-generation whites along all five dimensions.22

Table 3.
Cross-Generational Differences in Student Outcomes and Characteristics: Hispanic Immigrants
White Third-GenerationHispanic First-Generation Late-EnteringHispanic First-Generation Early-EnteringHispanic Second-GenerationHispanic Third-Generation
FCAT reading score − eighth grade 0.265 −0.603 0.0750 −0.0118 −0.0159 
 (0.909) (1.168) (0.964) (0.917) (0.926) 
FCAT math score − eighth grade 0.264 −0.383 0.0865 0.0159 −0.0117 
 (0.884) (1.110) (0.950) (0.899) (0.918) 
Disciplinary incident − eighth grade 0.202 0.191 0.200 0.204 0.246 
 (0.401) (0.393) (0.400) (0.403) (0.430) 
% Absent days − eighth grade 0.0636 0.0569 0.0545 0.0562 0.0676 
 (0.0768) (0.0733) (0.0670) (0.0722) (0.0823) 
Graduated from HS 0.591 0.431 0.535 0.562 0.522 
 (0.492) (0.495) (0.499) (0.496) (0.500) 
Took an advanced course in HS 0.508 0.376 0.532 0.483 0.470 
 (0.500) (0.484) (0.499) (0.500) (0.499) 
FRPL eligible 0.315 0.750 0.739 0.698 0.585 
 (0.464) (0.433) (0.439) (0.459) (0.493) 
LEP − eighth grade 0.000265 0.559 0.0932 0.0369 0.0213 
 (0.0163) (0.497) (0.291) (0.189) (0.144) 
LEP − ever 0.00338 0.829 0.839 0.597 0.272 
 (0.0581) (0.377) (0.368) (0.490) (0.445) 
English nonnative 0.0102 0.944 0.953 0.838 0.477 
 (0.100) (0.231) (0.211) (0.369) (0.499) 
Special education 0.132 0.0513 0.112 0.134 0.148 
 (0.339) (0.221) (0.316) (0.341) (0.355) 
Mother − less than HS 0.159   0.339 0.303 
 (0.366)   (0.473) (0.460) 
Mother − HS graduate 0.386   0.362 0.350 
 (0.487)   (0.480) (0.477) 
Mother − Some college 0.258   0.183 0.233 
 (0.438)   (0.387) (0.423) 
Mother − Bachelor's or more 0.194   0.114 0.111 
 (0.396)   (0.317) (0.314) 
Mother married 0.773   0.675 0.612 
 (0.419)   (0.468) (0.487) 
White Third-GenerationHispanic First-Generation Late-EnteringHispanic First-Generation Early-EnteringHispanic Second-GenerationHispanic Third-Generation
FCAT reading score − eighth grade 0.265 −0.603 0.0750 −0.0118 −0.0159 
 (0.909) (1.168) (0.964) (0.917) (0.926) 
FCAT math score − eighth grade 0.264 −0.383 0.0865 0.0159 −0.0117 
 (0.884) (1.110) (0.950) (0.899) (0.918) 
Disciplinary incident − eighth grade 0.202 0.191 0.200 0.204 0.246 
 (0.401) (0.393) (0.400) (0.403) (0.430) 
% Absent days − eighth grade 0.0636 0.0569 0.0545 0.0562 0.0676 
 (0.0768) (0.0733) (0.0670) (0.0722) (0.0823) 
Graduated from HS 0.591 0.431 0.535 0.562 0.522 
 (0.492) (0.495) (0.499) (0.496) (0.500) 
Took an advanced course in HS 0.508 0.376 0.532 0.483 0.470 
 (0.500) (0.484) (0.499) (0.500) (0.499) 
FRPL eligible 0.315 0.750 0.739 0.698 0.585 
 (0.464) (0.433) (0.439) (0.459) (0.493) 
LEP − eighth grade 0.000265 0.559 0.0932 0.0369 0.0213 
 (0.0163) (0.497) (0.291) (0.189) (0.144) 
LEP − ever 0.00338 0.829 0.839 0.597 0.272 
 (0.0581) (0.377) (0.368) (0.490) (0.445) 
English nonnative 0.0102 0.944 0.953 0.838 0.477 
 (0.100) (0.231) (0.211) (0.369) (0.499) 
Special education 0.132 0.0513 0.112 0.134 0.148 
 (0.339) (0.221) (0.316) (0.341) (0.355) 
Mother − less than HS 0.159   0.339 0.303 
 (0.366)   (0.473) (0.460) 
Mother − HS graduate 0.386   0.362 0.350 
 (0.487)   (0.480) (0.477) 
Mother − Some college 0.258   0.183 0.233 
 (0.438)   (0.387) (0.423) 
Mother − Bachelor's or more 0.194   0.114 0.111 
 (0.396)   (0.317) (0.314) 
Mother married 0.773   0.675 0.612 
 (0.419)   (0.468) (0.487) 

Notes: Standard deviations are given in parentheses. See online appendix table A.1 for the numbers of unique students in each category given in columns. FCAT = Florida Comprehensive Assessment Test; HS = high school; FRPL = free or reduced-price lunch; LEP = limited English proficient.

Table 4.
Cross-Generational Differences in Eighth-Grade Student Outcomes and Characteristics: Asian Immigrants
White Third-Asian First-GenerationAsian First-GenerationAsian Second-Asian Third-
GenerationLate-EnteringEarly-EnteringGenerationGeneration
FCAT reading score − eighth grade 0.265 0.0366 0.674 0.593 0.363 
 (0.909) (1.167) (0.960) (0.921) (0.885) 
FCAT math score − eighth grade 0.264 0.475 0.837 0.780 0.396 
 (0.884) (1.046) (0.945) (0.914) (0.882) 
Disciplinary incident − eighth grade 0.202 0.0691 0.0825 0.0859 0.151 
 (0.401) (0.254) (0.275) (0.280) (0.358) 
% Absent days − eighth grade 0.0636 0.0314 0.0311 0.0291 0.0513 
 (0.0768) (0.0571) (0.0471) (0.0470) (0.0712) 
Graduated from HS 0.591 0.575 0.708 0.758 0.628 
 (0.492) (0.494) (0.455) (0.429) (0.484) 
Took an advanced course in HS 0.508 0.749 0.811 0.805 0.602 
 (0.500) (0.433) (0.391) (0.396) (0.490) 
FRPL eligible 0.315 0.413 0.357 0.334 0.351 
 (0.464) (0.492) (0.479) (0.472) (0.478) 
LEP − eighth grade 0.000265 0.370 0.0374 0.0152 0.00167 
 (0.0163) (0.483) (0.190) (0.122) (0.0408) 
LEP − ever 0.00338 0.561 0.475 0.303 0.0501 
 (0.0581) (0.496) (0.499) (0.460) (0.218) 
English nonnative 0.0102 0.725 0.662 0.568 0.112 
 (0.100) (0.446) (0.473) (0.495) (0.315) 
Special education 0.132 0.0251 0.0490 0.0570 0.0860 
 (0.339) (0.156) (0.216) (0.232) (0.280) 
Mother − Less than HS 0.159   0.138 0.179 
 (0.366)   (0.345) (0.383) 
Mother − HS graduate 0.386   0.316 0.311 
 (0.487)   (0.465) (0.463) 
Mother − Some college 0.258   0.198 0.298 
 (0.438)   (0.398) (0.458) 
Mother − Bachelor's or more 0.194   0.338 0.208 
 (0.396)   (0.473) (0.406) 
Mother married 0.773   0.886 0.693 
 (0.419)   (0.318) (0.462) 
White Third-Asian First-GenerationAsian First-GenerationAsian Second-Asian Third-
GenerationLate-EnteringEarly-EnteringGenerationGeneration
FCAT reading score − eighth grade 0.265 0.0366 0.674 0.593 0.363 
 (0.909) (1.167) (0.960) (0.921) (0.885) 
FCAT math score − eighth grade 0.264 0.475 0.837 0.780 0.396 
 (0.884) (1.046) (0.945) (0.914) (0.882) 
Disciplinary incident − eighth grade 0.202 0.0691 0.0825 0.0859 0.151 
 (0.401) (0.254) (0.275) (0.280) (0.358) 
% Absent days − eighth grade 0.0636 0.0314 0.0311 0.0291 0.0513 
 (0.0768) (0.0571) (0.0471) (0.0470) (0.0712) 
Graduated from HS 0.591 0.575 0.708 0.758 0.628 
 (0.492) (0.494) (0.455) (0.429) (0.484) 
Took an advanced course in HS 0.508 0.749 0.811 0.805 0.602 
 (0.500) (0.433) (0.391) (0.396) (0.490) 
FRPL eligible 0.315 0.413 0.357 0.334 0.351 
 (0.464) (0.492) (0.479) (0.472) (0.478) 
LEP − eighth grade 0.000265 0.370 0.0374 0.0152 0.00167 
 (0.0163) (0.483) (0.190) (0.122) (0.0408) 
LEP − ever 0.00338 0.561 0.475 0.303 0.0501 
 (0.0581) (0.496) (0.499) (0.460) (0.218) 
English nonnative 0.0102 0.725 0.662 0.568 0.112 
 (0.100) (0.446) (0.473) (0.495) (0.315) 
Special education 0.132 0.0251 0.0490 0.0570 0.0860 
 (0.339) (0.156) (0.216) (0.232) (0.280) 
Mother − Less than HS 0.159   0.138 0.179 
 (0.366)   (0.345) (0.383) 
Mother − HS graduate 0.386   0.316 0.311 
 (0.487)   (0.465) (0.463) 
Mother − Some college 0.258   0.198 0.298 
 (0.438)   (0.398) (0.458) 
Mother − Bachelor's or more 0.194   0.338 0.208 
 (0.396)   (0.473) (0.406) 
Mother married 0.773   0.886 0.693 
 (0.419)   (0.318) (0.462) 

Notes: Standard deviations are given in parentheses. See online appendix table A.1 for the numbers of unique students in each category given in columns. FCAT = Florida Comprehensive Assessment Test; HS = high school; FRPL = free or reduced-price lunch; LEP = limited English proficient.

Interestingly, the findings reported in these two tables also reveal that recent immigrants outperform more established generations despite the evidence that immigrants accumulate wealth and education across generations, and that later generations have significantly better English skills than recent immigrants. For instance, among Hispanics, 75 percent of early entering first-generation immigrants are eligible for subsidized meals, compared to 70 percent among second-generation Hispanics, and 59 percent among third-generation Hispanic immigrants. Similarly, one third of second-generation Hispanic immigrants have mothers with less than a high school degree, compared to 30 percent for third-generation Hispanic immigrants. Eighty-four percent of the early entering first-generation Hispanics have been categorized as limited English proficient at least once since they entered the public school system, in stark contrast to 60 percent for second-generation, and 27 percent for third-generation Hispanics. Similar, yet considerably smaller, cross-generational differences in wealth, education, and English skills are observed among Asian immigrants.

Cross-generational Differences in Educational Outcomes

Figure 2 illustrates the patterns of eighth-grade reading and math test scores for Hispanic and Asian students, arrayed from left to right in terms of the amount of time spent in the United States (the top panel of online appendix figure A.2 repeats the same analysis for Hispanic immigrants, excluding Cuban and Puerto Rican immigrants). It's clear that whether or not we present raw differences or differences controlling for school fixed effects and background characteristics, the patterns of results are the same: For both Asian and Hispanic students, the latest arriving first-generation students fare the worst, and eighth-grade test scores rise essentially monotonically with the amount of time spent in the United States, peaking for those who began kindergarten in Florida schools. Second-generation students, either Asian or Hispanic, perform worse than the earliest-arriving first generation students, and third-generation students tend to perform worse still. For Asians, there is essentially no difference between the models in which we control for school fixed effects and background characteristics and those in which we control for just disability, year, age, and gender. For Hispanic students, it makes a difference regarding level of performance whether or not we control for these variables—across the board, Hispanic students are estimated to perform relatively better (and for the earliest-arriving first-generation immigrants, exceed third-generation white students’ performance) in models where we control for a variety of factors than in cases in which we don't control for these variables. But the control variables simply shift the relationship relative to third-generation white students; they do not alter the relationship across generations for Hispanic students virtually at all.23

Figure 2.

Cross-Generational Differences in Eighth-Grade Test Scores among Hispanic and Asian Students, Compared with Third-Generation White Students, by Grade of Entry into the School System for the First-Generation Immigrants

Notes: The “no controls,” “with controls,” and “full controls” specifications are obtained using models with only disability, gender, age controls, and cohort fixed effects (no controls), controls for school by cohort fixed effects and school lunch eligibility (with controls), and also maternal characteristics (full controls). The first-generation indicators replaced by a series of first-generation indicators based on grade of entry. Standard errors clustered at the school by year level. Capped spikes represent the 95 percent confidence interval for each coefficient. The regressions include 19,286 unique first-generation Hispanic students, 21,998 unique second-generation Hispanic students, 10,533 unique third-generation Hispanic students, 2,375 unique first-generation Asian students, 2,534 unique second-generation Asian students, 331 unique third-generation Asian students, and 76,664 unique third-generation white students.

Figure 2.

Cross-Generational Differences in Eighth-Grade Test Scores among Hispanic and Asian Students, Compared with Third-Generation White Students, by Grade of Entry into the School System for the First-Generation Immigrants

Notes: The “no controls,” “with controls,” and “full controls” specifications are obtained using models with only disability, gender, age controls, and cohort fixed effects (no controls), controls for school by cohort fixed effects and school lunch eligibility (with controls), and also maternal characteristics (full controls). The first-generation indicators replaced by a series of first-generation indicators based on grade of entry. Standard errors clustered at the school by year level. Capped spikes represent the 95 percent confidence interval for each coefficient. The regressions include 19,286 unique first-generation Hispanic students, 21,998 unique second-generation Hispanic students, 10,533 unique third-generation Hispanic students, 2,375 unique first-generation Asian students, 2,534 unique second-generation Asian students, 331 unique third-generation Asian students, and 76,664 unique third-generation white students.

As discussed by Duncan and Trejo (2016), this pattern of declining test scores across successive immigrant generations might be indicative of bias associated with cross-generational differences in self-reported ethnicity. To investigate this possibility, we replicate the analysis in figure 2 using the maternal racial identity reported in birth records to identify race/ethnicity for second- and third-generation immigrants, using third-generation white students with non-Hispanic and non-Asian mothers as the baseline group. If the decline in test scores is indeed driven by negative selection into later generations, one would expect the achievement gaps between early entering first-generation immigrants and more established generations to decrease when Hispanic and Asian identity is defined based on mothers’ self-reported race/ethnicity.24

Figure 3 presents the cross-generational differences obtained using our baseline model where the first set of estimates (labeled “Student Hispanic” and “Student Asian”) use the self-reported race/ethnicity of the student, whereas the second set of estimates (labeled “Mother Hispanic” and “Mother Asian”) use the self-reported race/ethnicity of the mother to identify Hispanic and Asian second- and third-generation immigrants. In both cases, first-generation immigrants are classified based on their self-reported racial/ethnic identity, and third-generation white students are used as the baseline group.

Figure 3.

Cross-Generational Differences in Eighth-Grade Test Scores Compared with Third Generation White Students, by Grade of Entry into the School System for the First-Generation Immigrants, Student Racial/Ethnic Identification versus Mother Racial/Ethnic Identification

Notes: The estimates are obtained using the model that controls for disability, gender, age controls, and cohort fixed effects. “Student Hispanic” and “Student Asian” specifications use the self-reported race/ethnicity whereas “Mother Hispanic” and “Mother Asian” specifications use the self-reported race/ethnicity of the mother to identify second- and third-generation Asian and Hispanic immigrants. Standard errors clustered at the school by year level. Capped spikes represent the 95 percent confidence interval for each coefficient. The regressions include 19,286 unique first-generation Hispanic students, 28,870 unique second-generation students with Hispanic mothers, 9,823 unique third-generation students with Hispanic mothers, 2,375 unique first-generation Asian students, 2,051 unique second-generation students with Asian mothers, 233 unique third-generation students with Asian mothers, 21,998 unique second-generation Hispanic students, 10,533 unique third-generation Hispanic students, 2,375 unique second-generation Asian students, 331 unique third-generation Asian students, and 79,052 unique third-generation white students with non-Hispanic and non-Asian mothers.

Figure 3.

Cross-Generational Differences in Eighth-Grade Test Scores Compared with Third Generation White Students, by Grade of Entry into the School System for the First-Generation Immigrants, Student Racial/Ethnic Identification versus Mother Racial/Ethnic Identification

Notes: The estimates are obtained using the model that controls for disability, gender, age controls, and cohort fixed effects. “Student Hispanic” and “Student Asian” specifications use the self-reported race/ethnicity whereas “Mother Hispanic” and “Mother Asian” specifications use the self-reported race/ethnicity of the mother to identify second- and third-generation Asian and Hispanic immigrants. Standard errors clustered at the school by year level. Capped spikes represent the 95 percent confidence interval for each coefficient. The regressions include 19,286 unique first-generation Hispanic students, 28,870 unique second-generation students with Hispanic mothers, 9,823 unique third-generation students with Hispanic mothers, 2,375 unique first-generation Asian students, 2,051 unique second-generation students with Asian mothers, 233 unique third-generation students with Asian mothers, 21,998 unique second-generation Hispanic students, 10,533 unique third-generation Hispanic students, 2,375 unique second-generation Asian students, 331 unique third-generation Asian students, and 79,052 unique third-generation white students with non-Hispanic and non-Asian mothers.

Among both Hispanics and Asians we find that the general patterns in the cross-generational differences in test scores remain unchanged whether we use the student's or the mother's self-reported racial/ethnic identity to classify second- and third-generation Hispanic or Asian students. In figure A.3 in the online appendix we repeat the same analysis for Hispanic students, excluding students of Cuban or Puerto Rican origin, and show that the conclusions remain unchanged. To investigate whether changes in the national origin composition of Hispanic and Asian students across generations account for any of the observed cross-generational differences, we break down the Hispanic and Asian analyses for first- and second-generation by country of origin (Puerto Ricans, Cubans, Mexicans, and all others for Hispanics; Filipinos, Chinese, Japanese, and all others for Asians)25 in online appendix figures A.4–A.7, and show that second-generation immigrants perform either worse than, or comparable to, early-entering first-generation immigrants.

In online appendix table A.1, we present the results in which we move from the first to the second model sequentially. Column 2 controls for school by cohort fixed effects, column 3 introduces low-income status in the eighth grade, column 4 replaces low-income status in the eighth-grade FRPL eligibility with the status during the first year we observe the student, column 5 introduces an indicator for whether English is the main language spoken at home, column 6 introduces maternal background factors excluding native language spoken, and column 7 reintroduces native language. Two findings are worth highlighting when making the step-by-step controls. First, the observed differences in English language skills, wealth, and education between Hispanic students and third-generation white students explain a significant portion of the Hispanic achievement gaps (except for the late-entering first generation immigrants), yet the extent of their role in explaining these gaps diminishes across generations.26 Second, the acquisition of English language, wealth, and education across immigrant generations explains very little of the cross-generational differences in educational outcomes among Hispanic and Asian students. Table A.2 in the online appendix presents the same set of analyses in models, excluding Cuban and Puerto Rican students, and demonstrates the patterns remain the same regardless of our treatment of these student groups. In online appendix table A.3, we replicate the same analysis from table A.1 using elementary school (attended in the fifth grade) fixed effects instead of the middle school attended in the eighth grade. Although the achievement gap between late-entering first-generation immigrants and others declines (as immigrants who entered the school system after elementary school are excluded from this analysis), and differences in elementary schools attended seem to explain more of the eighth-grade achievement gap between white students and Hispanic/Asian students (especially for more established generations) compared to middle schools, similar patterns emerge when it comes to how achievement gaps change as we introduce new covariates into the model.

We next turn to attendance and behavioral outcomes, as reported in figure 4 (and online appendix table A.4, and the bottom panel of online appendix figure A.2 and table A.5 for specifications, excluding Cuban and Puerto Rican students). Again, with these outcomes the differences across generations (as well as when compared with third-generation white students) are very similar whether or not we control for a variety of background factors and school-by-cohort fixed effects. For the most part, there is not much of a pattern in disciplinary incidents and absenteeism across timing of first-generation entry, and second-generation Asian and Hispanic students have disciplinary and absenteeism outcomes similar to first-generation students. For third-generation students, however, outcomes tend to be worse for both Hispanic and Asian students, with often markedly increased rates of both disciplinary outcomes and absenteeism relative to first- and second-generation immigrants. Third-generation Hispanic absenteeism rates are very similar to third-generation white absenteeism rates (and third-generation Hispanic disciplinary rates exceed third-generation white student rates), while third-generation Asian disciplinary rates are very similar to third-generation white disciplinary rates.

Figure 4.

Cross-Generational Differences in Disciplinary Incidents and Absences among Hispanic and Asian Students, Compared with Third-Generation White Students, by Grade of Entry into the School System for the First-Generation Immigrants

Notes: The “no controls,” “with controls,” and “full controls” specifications are obtained using models with only disability, gender, age controls, and cohort fixed effects (no controls), controls for school by cohort fixed effects and school lunch eligibility during the eighth grade (with controls), and also maternal characteristics (full controls). The first-generation indicators replaced by a series of first-generation indicators based on grade of entry. Standard errors clustered at the school by year level. Capped spikes represent the 95 percent confidence interval for each coefficient. The regressions include 20,756 unique first-generation Hispanic students, 22,457 unique second-generation Hispanic students, 10,874 unique third-generation Hispanic students, 2,552 unique first-generation Asian students, 2,572 unique second-generation Asian students, 342 unique third-generation Asian students, and 79,729 unique third-generation white students.

Figure 4.

Cross-Generational Differences in Disciplinary Incidents and Absences among Hispanic and Asian Students, Compared with Third-Generation White Students, by Grade of Entry into the School System for the First-Generation Immigrants

Notes: The “no controls,” “with controls,” and “full controls” specifications are obtained using models with only disability, gender, age controls, and cohort fixed effects (no controls), controls for school by cohort fixed effects and school lunch eligibility during the eighth grade (with controls), and also maternal characteristics (full controls). The first-generation indicators replaced by a series of first-generation indicators based on grade of entry. Standard errors clustered at the school by year level. Capped spikes represent the 95 percent confidence interval for each coefficient. The regressions include 20,756 unique first-generation Hispanic students, 22,457 unique second-generation Hispanic students, 10,874 unique third-generation Hispanic students, 2,552 unique first-generation Asian students, 2,572 unique second-generation Asian students, 342 unique third-generation Asian students, and 79,729 unique third-generation white students.

Patterns like those seen regarding test scores re-emerge when considering on-time high school completion and advanced course-taking in high school, as presented in figure 5 (and online appendix table A.6, as well as figure A.8 and table A.7 for specifications, excluding Cuban and Puerto Rican students). Because of data limitations, for these outcomes we cannot differentiate among the earliest entrants—all we know is whether a student arrived in third grade or earlier. Nonetheless, the same inverse U-shaped pattern emerges for both Hispanic and Asian immigrants across generations and within the first generation. Among first-generation students, the probability of on-time graduation and the likelihood of taking advanced high school classes increase the earlier the student arrived in Florida public schools. Second-generation students fare similarly to the earliest arriving first-generation students, and then third-generation students fare worse than second-generation students. As with the pattern seen with test scores, it appears that higher-generation Hispanic and Asian students begin to fall back relative to their early-entering first-generation peers. We also inspect the possible role of selection into later generations along these two outcomes (not reported here), and find that these patterns persist whether we use students’ or mothers’ racial/ethnic identity to classify second- and third-generation Asian and Hispanic immigrants.

Figure 5.

Cross-Generational Differences in High School Graduation and Course-Taking among Hispanic and Asian Students, Compared with Third-Generation White Students, by Grade of Entry into the School System for the First-Generation Immigrants

Notes: The “no controls,” “with controls,” and “full controls” specifications are obtained using models with only disability, gender, age controls, and cohort fixed effects (no controls), controls for school by cohort fixed effects and school lunch eligibility during the ninth grade (with controls), and also maternal characteristics (full controls). The first-generation indicators replaced by a series of first-generation indicators based on grade of entry. Standard errors clustered at the school by year level. Capped spikes represent the 95 percent confidence interval for each coefficient. See online appendix table A.3 for number of observations in each group.

Figure 5.

Cross-Generational Differences in High School Graduation and Course-Taking among Hispanic and Asian Students, Compared with Third-Generation White Students, by Grade of Entry into the School System for the First-Generation Immigrants

Notes: The “no controls,” “with controls,” and “full controls” specifications are obtained using models with only disability, gender, age controls, and cohort fixed effects (no controls), controls for school by cohort fixed effects and school lunch eligibility during the ninth grade (with controls), and also maternal characteristics (full controls). The first-generation indicators replaced by a series of first-generation indicators based on grade of entry. Standard errors clustered at the school by year level. Capped spikes represent the 95 percent confidence interval for each coefficient. See online appendix table A.3 for number of observations in each group.

Cross-Generational Differences in Student Progress

Beyond cross-sectional comparisons in educational outcomes across immigrant generations, we also examine how these gaps in student outcomes evolve across grades. For this exercise, we construct a balanced panel of students who are observed in each grade from 3 through 8, and compare (regression-adjusted) patterns of test score growth across grades for first-, second-, and third-generation Hispanic and Asian students, relative to third-generation white peers.27 For this analysis, it is of course necessary that we concentrate only on early-entering first-generation immigrants, as we need to observe third-grade test scores for them.

Figure 6 presents these over-time patterns for the same students.28 (Similar patterns can be observed in the raw data, and are presented in figure A.9 in the online appendix.) As can be seen, second- and third-generation Hispanic students tend to maintain their same position relative to third-generation white students in both reading and math over the grades considered, whereas first-generation Hispanic students make rapid progress across the grades, eventually overtaking second- and third-generation Hispanic students (and, in the case of reading, third-generation white students as well, in the regression-adjusted specification). Third-generation Asian students maintain their position relative to third-generation white students, whereas both first- and second-generation Asian students improve in relative terms over the six-grade period, with first-generation students eventually pulling away from second-generation students as well. These patterns suggest that Clotfelter, Ladd, and Vigdor's (2009) findings that Asian and Hispanic students’ test scores in North Carolina grow more rapidly across the grades than do white students’ scores are driven by first- and (in the case of Asian students, second-) generation immigrants, and likely not by the more generationally established Asian- and Hispanic-identifying families. The top panel in figure A.10 in the online appendix presents the test score findings for Hispanic students (excluding Puerto Rican and Cuban students), and similar conclusions are reached.

Figure 6.

Cross-Generational Regression-Adjusted Trends in Student Test Scores over Grades among Hispanic and Asian Students, Compared with Third-Generation White Students in Grades 3 through 8, Conditional on Continuous Enrollment

Notes: The estimated cross-generational differences among Asian, Hispanic, compared with the third-generation white students are obtained using the covariates given in column 4, online appendix table A.1, estimated separately for each given grade. Standard errors clustered at the school by year level. Capped spikes represent the 95 percent confidence interval for each estimated coefficient. The regressions include 34,325 unique early-entering first-generation Hispanic students, 61,439 unique second-generation Hispanic students, 27,018 unique third-generation Hispanic students, 3,321 unique early-entering first-generation Asian students, 6,307 unique second-generation Asian students, 709 unique third-generation Asian students, and 218,830 unique third-generation white students.

Figure 6.

Cross-Generational Regression-Adjusted Trends in Student Test Scores over Grades among Hispanic and Asian Students, Compared with Third-Generation White Students in Grades 3 through 8, Conditional on Continuous Enrollment

Notes: The estimated cross-generational differences among Asian, Hispanic, compared with the third-generation white students are obtained using the covariates given in column 4, online appendix table A.1, estimated separately for each given grade. Standard errors clustered at the school by year level. Capped spikes represent the 95 percent confidence interval for each estimated coefficient. The regressions include 34,325 unique early-entering first-generation Hispanic students, 61,439 unique second-generation Hispanic students, 27,018 unique third-generation Hispanic students, 3,321 unique early-entering first-generation Asian students, 6,307 unique second-generation Asian students, 709 unique third-generation Asian students, and 218,830 unique third-generation white students.

One possible explanation for the rapid increase in test scores among first-generation immigrants is that they catch up with their peers (e.g., in their language skills) after they enter the school system. If this is the case, one would expect a more gradual rise in test scores for earliest-entering first-generation immigrants who have had more time to catch up before the third grade test. In online figure A.11, we repeat the same analysis restricting the early-entering first-generation immigrants to those who entered the school system in kindergarten. Earliest-arriving first-generation immigrants perform better than their second- and third-generation peers as early as the third grade, and improve their test scores in relation to their third-generation white peers, even though this improvement is more gradual compared with students who arrive in the United States during the first, second, or third grade.

We perform the same analysis looking at disciplinary incidents and truancy across the same grades; these results are presented in figure 7. In the case of disciplinary incidents, all Hispanic students tend to have similar or lower rates of disciplinary incidents to third-generation white students in the elementary grades, a relationship that holds in the middle grades for first- and second-generation Hispanic students. Third-generation Hispanic students, on the other hand, have more disciplinary incidents in the middle school grades. For Asian students, all generations have fewer disciplinary incidents in the elementary school grades, but in the middle grades the third-generation Asian students’ disciplinary issues are indistinguishable from those of third-generation white students, whereas first- and second-generation Asian students have substantially fewer disciplinary incidents in the middle school grades. Regarding absences, the relative patterns across the generations appear to be highly stable across grades for both Asian and Hispanic students. We report the unadjusted results in online appendix figure A.12, and repeat the same analysis excluding students of Cuban or Puerto Rican origin (reported in the bottom panel of figure A.10), and find similar patterns.

Figure 7.

Cross-Generational Regression-Adjusted Trends in Student Disciplinary Incidents and Absences over Grades among Hispanic and Asian Students, Compared with Third-Generation White Students in Grades 3 through 8, Conditional on Continuous Enrollment

Notes: The estimated cross-generational differences among Asian, Hispanic, compared with the third-generation white students are obtained using the covariates given in column 4, online appendix table A.1, estimated separately for each given grade. Standard errors clustered at the school by year level. Capped spikes represent the 95 percent confidence interval for each estimated coefficient. The regressions include 37,402 unique early entering first-generation Hispanic students, 63,251 unique second-generation Hispanic students, 28,406 unique third-generation Hispanic students, 3,694 unique early entering first-generation Asian students, 6,457 unique second-generation Asian students, 754 unique third-generation Asian students, and 234,095 unique third-generation white students.

Figure 7.

Cross-Generational Regression-Adjusted Trends in Student Disciplinary Incidents and Absences over Grades among Hispanic and Asian Students, Compared with Third-Generation White Students in Grades 3 through 8, Conditional on Continuous Enrollment

Notes: The estimated cross-generational differences among Asian, Hispanic, compared with the third-generation white students are obtained using the covariates given in column 4, online appendix table A.1, estimated separately for each given grade. Standard errors clustered at the school by year level. Capped spikes represent the 95 percent confidence interval for each estimated coefficient. The regressions include 37,402 unique early entering first-generation Hispanic students, 63,251 unique second-generation Hispanic students, 28,406 unique third-generation Hispanic students, 3,694 unique early entering first-generation Asian students, 6,457 unique second-generation Asian students, 754 unique third-generation Asian students, and 234,095 unique third-generation white students.

Other Possible Mechanisms behind the Cross-Generational Gaps

So far, beyond a transition period for first-generation immigrants after their initial entry into the school system, our findings have shown a steady decline in student performance across generations even after controlling for observed student, family, and school traits, with early-arriving first-generation immigrants academically outperforming those born in the United States, and those born in the United States to foreign-born mothers tending to academically outperform those born in the United States to native-born mothers. That said, there is not a clear-cut pattern regarding truancy and disciplinary outcomes—just regarding test scores, high school completion, and high school course-taking.

Aside from the aforementioned negative racial/ethnic selection into later generations, the decline in academic outcomes across successive immigrant generations might be driven by immigrant optimism/motivation fading across generations, with more established generations passing their leveled aspirations to their children. This decline in educational motivation might manifest itself in cross-generational differences in revealed educational preferences among observationally equivalent students. We investigate this possibility in the context of school choice behavior in middle school and advanced course-taking behavior in high school. For the former exercise, we focus on students in the last year of elementary school and examine whether first-generation immigrants are more likely to choose a high-performing middle school (schools that received a grade of “A” or “B” in Florida's school accountability system) in the following year, compared with their second- and third-generation peers in the same elementary school with similar achievement levels and socioeconomic status. In the latter exercise, we once again estimate the cross-generational gaps in college credit-bearing course-taking among high school graduates, yet, this time, we also account for the differences in eighth-grade test scores. In both cases, we exclude late-entering first-generation immigrants from the analysis, and use third-generation white students as the baseline group.

We find significant differences in the educational choices of first-, second-, and third-generation Hispanic students, with second- and third-generation Hispanic students more likely to “undershoot” compared with their foreign-born Hispanic peers. In particular, the estimates presented in the first two columns of table 5 reveal that first-generation Hispanic students are 4 percentage points (about 8 percent) more likely to choose a high-performing middle school, and 3 percentage points (about 6 percent) more likely to take advanced high school courses, compared with their observationally similar second-generation peers, who are, in turn, less likely to undershoot compared with their third-generation Hispanic peers (even though the gap shrinks considerably when we control for maternal attributes). We find similar, yet less precise, cross-generational differences among Asian immigrants (reported in the third and fourth columns of table 5).

Table 5.
Cross-Generational Differences in School Choice and Advanced Course-Taking in High School
Chose a High-Performing Middle School
Hispanic StudentsAsian Students
(I)(II)(I)(II)
Early-entering first-generation 0.010***  −0.017**  
 (0.003)  (0.006)  
Second-generation −0.029*** −0.037*** −0.020*** −0.018*** 
 (0.003) (0.003) (0.005) (0.005) 
Third-generation −0.038*** −0.041*** −0.047*** −0.044*** 
 (0.003) (0.003) (0.012) (0.012) 
Prior year school × Cohort FE Yes Yes Yes Yes 
FRPL eligible Yes Yes Yes Yes 
Prior year test scores Yes Yes Yes Yes 
Maternal characteristics No Yes No Yes 
N 246,792 176,871 20,356 13,307 
 Took an Advanced High School Course 
Early-entering first-generation 0.053***  0.139***  
 (0.006)  (0.013)  
Second-generation 0.019*** 0.026*** 0.133*** 0.130*** 
 (0.005) (0.005) (0.009) (0.009) 
Third-generation −0.009 −0.001 0.016 0.020 
 (0.006) (0.006) (0.027) (0.026) 
High school × Cohort FE Yes Yes Yes Yes 
Native language Yes Yes Yes Yes 
FRPL eligible Yes Yes Yes Yes 
Eighth-grade test scores Yes Yes Yes Yes 
Maternal characteristics No Yes No Yes 
N 37,543 26,629 4,063 2,802 
Chose a High-Performing Middle School
Hispanic StudentsAsian Students
(I)(II)(I)(II)
Early-entering first-generation 0.010***  −0.017**  
 (0.003)  (0.006)  
Second-generation −0.029*** −0.037*** −0.020*** −0.018*** 
 (0.003) (0.003) (0.005) (0.005) 
Third-generation −0.038*** −0.041*** −0.047*** −0.044*** 
 (0.003) (0.003) (0.012) (0.012) 
Prior year school × Cohort FE Yes Yes Yes Yes 
FRPL eligible Yes Yes Yes Yes 
Prior year test scores Yes Yes Yes Yes 
Maternal characteristics No Yes No Yes 
N 246,792 176,871 20,356 13,307 
 Took an Advanced High School Course 
Early-entering first-generation 0.053***  0.139***  
 (0.006)  (0.013)  
Second-generation 0.019*** 0.026*** 0.133*** 0.130*** 
 (0.005) (0.005) (0.009) (0.009) 
Third-generation −0.009 −0.001 0.016 0.020 
 (0.006) (0.006) (0.027) (0.026) 
High school × Cohort FE Yes Yes Yes Yes 
Native language Yes Yes Yes Yes 
FRPL eligible Yes Yes Yes Yes 
Eighth-grade test scores Yes Yes Yes Yes 
Maternal characteristics No Yes No Yes 
N 37,543 26,629 4,063 2,802 

Notes: Standard errors clustered at the prior year school level in the top panel, and clustered at the high school level in the bottom panel in parentheses. All models control for gender, age, special education status in addition to the covariates listed. FE = fixed effects; FRPL = free or reduced-price lunch.

**p < 0.01; ***p < 0.001.

Another interesting finding here is that, compared with their third-generation white peers with similar achievement levels, recent immigrants (first- and second-generation) are, in most cases, more likely to have greater family investment in education. For instance, both first- and second-generation Hispanic and Asian students are more likely to take advanced high school courses compared with their native white peers, holding course-availability constant. Hispanic first-generation immigrants are also more likely to choose a high-performing middle school at the end of elementary school, which might explain their relative test score improvement in middle school grades in relation to their third-generation white peers. The same is not true, however, for second-generation Hispanic students, or first- and second-generation Asian immigrants. Table A.8 in the online appendix repeats the school choice exercise restricting the first-generation immigrants to those who arrived in the United States before kindergarten, and similar conclusions are reached.

4.  Conclusion

This paper presents the first comprehensive look at the relative performance of early-arriving first-generation immigrants, late-arriving first-generation immigrants, second-generation immigrants, and third-generation immigrants, using population-level data in the United States. A clear pattern of results emerges, at least for the Florida public school students identified as Asian or Hispanic whom we study: We observe that whereas first-generation immigrants who arrive in Florida in fourth grade or later consistently perform at a level that is lower than other students, first-generation students who arrive in Florida in third grade or earlier generally perform better than any other generation of students who share their same ethnicity. Second-generation Asian and Hispanic students in turn tend to perform better than do third-generation students of the same ethnicity.

These results have potential implications for immigration policy and how one perceives the role immigrants play in American schools and beyond. They suggest that although it appears that newly arrived immigrant children perform very poorly and require considerable resources, not only do these children catch up very quickly to their native-born co-ethnic peers, but for those who arrive before the age of 9 years or so, they tend to exceed the performance of their native-born co-ethnic peers. And the U.S.-born children of foreign-born individuals, regardless of when they arrive in the United States, tend to outperform others from the same ethnicity whose families have lived longer in the country.

Our study also demonstrates the potential of administrative data systems that provide more information about parents and thus facilitate such analysis. That said, administrative datasets have important limitations that hinder the ability to better understand the mechanisms behind the observed differences in outcomes. For example, more research is needed (1) to explore the cultural and economic mechanisms behind these patterns of findings; and to understand why (2) third-generation Asian students no longer possess the relative academic advantages of first- and second-generation Asian students; (3) why third-generation Hispanic students appear to fall back academically relative to their first- and second-generation Hispanic peers; (4) why Asian immigrants, in general, perform better in school compared with other immigrants; and (5) why more established immigrant generations make arguably “suboptimal” educational choices compared with recent immigrants.

Notes

1. 

In particular, the straight-line assimilation theory predicts that immigrants will inevitably blend into the mainstream culture over time in four stages. First, immigrants will come into contact with the mainstream culture, followed by a period of conflict. The mainstream will eventually accommodate the immigrant group, who will then culturally and structurally assimilate into the native population. This assimilation is primarily driven by the wealth and education accumulation of second and higher generations, which helps eliminate discrimination by the native population (Kao and Tienda 1995).

2. 

For instance, Suarez-Orozco (2000) argues that recent immigrants are less likely to break ties with their origins due to the relative ease and accessibility of mass transportation and the new communication technologies. Further, these ties are refreshed with continuous migratory flow from their countries of origin. Additionally, unlike the immigrants of the last century, new immigrants face an increasingly diverse, multicultural, and multilingual mainstream. Finally, the declining demand for low-skilled labor in the U.S. workforce might inhibit upward social mobility for low-skilled new immigrants.

3. 

Further, Portes and Rumbaut (2001) examine the factors that might affect immigrants taking these distinct assimilation paths.

4. 

Another strand of literature examines the degree to which economic outcomes of immigrants converge to those of natives across generations; see Abramitzky and Boustan (2017) for a review.

5. 

They also show significant heterogeneity in immigrant achievement by country of origin. For instance, immigrant students from Central America and the Caribbean usually lag behind their native peers on standardized tests, while Asian immigrants tend to outperform.

6. 

That said, there are several studies using European data that examine educational outcomes by age at migration. See Hermansen (2017) for a summary of existing studies of the relationship between age at migration and education.

7. 

Using school records linked to birth records, Hull (2017) also examines the cross-generational differences in test score gains among Hispanic students in North Carolina and finds the observed trends in Hispanic test scores between grades 3 and 8 in Clotfelter, Ladd, and Vigdor (2009) are mainly driven by first-generation immigrants. Our study improves upon Hull's research in two important ways. First, because we use social security numbers (along with full names and exact date of birth) to link birth records to school records and our school records include country of birth (which enable us to identify first generation immigrants), we can identify immigrant generation for a significantly higher share of students (71 percent versus 19 percent) than Hull. Second, because of sample-size restrictions, Hull is unable to examine Asian immigrants.

8. 

During the time frame we examine in Florida, birth certificate applications listed the following options for the birth place of the mother: (1) United States; (2) Puerto Rico; (3) U.S. Virgin Islands; (4) Guam; (5) Canada; (6) Cuba; (7) Mexico; (8) Northern Marianas; (9) American Samoa; and (10) All others. Although the application form provided additional space to write down the birth place if it is not listed, either very few applicants opted to specify, or very few were recorded. Therefore, we can credibly identify the birth place of the mother only for the countries listed here.

9. 

Throughout the analysis, we identify second-generation immigrants as those who were born in the United States to foreign-born mothers. Another commonly used approach in this context is to define second-generation as those who were born in the United States with at least one immigrant parent. In this study, we cannot utilize this alternative definition as we do not observe father's birthplace in the birth records. Therefore, it is important to note that some students who would be identified as second-generation under the alternative are categorized as third-generation in our analysis. However, this should not be a great concern, because, based on the recent Census estimates, among the households with a native wife, only 3 percent have a foreign-born husband in the United States (see https://www2.census.gov/library/publications/2013/acs/acsbr11-16.pdf).

10. 

As Figlio et al. (2014) demonstrate, the match rate is essentially identical to that which would have been predicted based on the American Community Survey data. A high match rate is especially important when considering a Hispanic population, given the higher concentration of Hispanic surnames relative to names from other origins, as well as the uneven treatment of Hispanic mothers’ and fathers’ family names in children's administrative records.

11. 

These figures are based on children aged 5 to 16 years, and make use of individual weights.

12. 

We can also compare second-generation to third-plus–generation immigrants born in Florida who remain versus leave the state, and those born elsewhere in the United States who migrate to Florida versus those who do not. Second-generation Florida stayers have lower levels of maternal education than second-generation Florida leavers, in comparison with their third-plus–generation counterparts, but the patterns are reversed regarding second- versus third-plus–generation Florida in-migrants versus those who do not migrate to Florida.

13. 

To check the robustness of our findings to this restriction, we compare the nativity gaps in eighth-grade test scores using all U.S. born students in our sample and the nativity gaps obtained using only Florida-born second and higher generation students, and show that patterns remain unchanged. See figure A.1 in the online appendix, which can be accessed on Education Finance and Policy’s Web site at www.mitpressjournals.org/efp.

14. 

Results of analyses in which we exclude first- and second-generation Puerto Rican and Cuban immigrants are available in the online appendix.

15. 

These incidents can be triggered by a wide array of student misbehavior, ranging from disruptive behavior in the classroom to gang involvement. Based on the severity of the incident, teachers and principals have full discretion over the type of action taken, which may include corporal punishment, in-school or out-of-school suspension, placement in a different program, and expulsion.

16. 

Florida offers a wide array of high school acceleration programs for students to earn college credit while in high school, including Advanced Placement (AP), International Baccalaureate (IB), Advanced International Certificate of Education (AICE), and dual enrollment (DE) courses. In our main analysis, we focus on AP, IB, and AICE course-taking in math, science, English, and social studies because these courses are offered in the K–12 system and hence included in our data.

17. 

For an excellent review of the literature on the impact of cognitive skills and educational attainment on labor market outcomes and economic growth, see Hanushek (2009).

18. 

We use third-generation white students as the baseline group to assess the extent to which racial/ethnic gaps in educational outcomes documented in the extant literature between whites and Hispanics and Asians (e.g., Clotfelter, Ladd, and Vigdor 2009) are driven by recent Hispanic and Asian immigrants.

19. 

To identify students who enter the school system as early as kindergarten, we restrict the sample to students born in 1997 and 1998 for whom it is possible to track students all the way from kindergarten to the eighth grade in our sampling frame. In other analysis, we divide the first-generation immigrants into two groups—immigrants who enter the Florida public school system by the third grade (early entering) and those who enter after the third grade (late entering). In the latter analysis, we use all students born between 1992 and 1998 and subsequently attended a Florida public school.

20. 

For test scores, disciplinary incidents, and attendance, we use the year the student enters the eighth grade for the first time and the school attended during that year to define cohorts. For high school graduation, we use the year the student enters the ninth grade for the first time and the high school attended during that year.

21. 

We also show in the online appendix what happens as we control successively for each set of these control variables.

22. 

We also broke down this analysis by gender but did not find significantly different patterns for boys and girls. The results of this analysis are available from the authors upon request.

23. 

It is important to note that there is very little common support in the English native language indicator between first-generation immigrants (especially Hispanics) and their-third generation white peers. In particular, only 5 percent of first-generation Hispanic students report English as the language spoken at home, whereas 99 percent of third-generation white students speak English at home. Therefore, the identification of the native language “effect” on student outcomes in these regressions come primarily from the comparison of native and nonnative speakers of English among second- and third-generation Hispanic students.

24. 

With regard to Hispanic immigrants, changes between maternal ethnic identity and children's ethnic identity are the same across generations: Among children of mothers who self-identify as Hispanic, 73 percent of second-generation immigrants identify as Hispanic and 74 percent of third-generation immigrants identify as Hispanic. But there are large cross-generational differences with regard to Asians. Among children of mothers who self-identify as Asian, 70 percent of second-generation immigrants identify as Asian but only 34 percent of third-generation immigrants identify as Asian. This suggests that Asian immigrants have integrated across generations in a different manner than have Hispanic immigrants.

25. 

We are limited to these countries of origin because Puerto Ricans, Cubans, and Mexicans are the only Hispanic groups independently identified on the birth certificate under maternal place of birth, and Chinese, Japanese, and Filipinos are the only Asian groups identified on the birth certificate under maternal race.

26. 

It is also interesting to note that whether we control for FRPL eligibility in the eighth grade or FRPL eligibility during the first year after the student enters the school system makes very little difference.

27. 

Note that because this specification involves controlling for school-by-cohort fixed effects, we are comparing students to their peers at the same schools.

28. 

In this analysis, we look at the test score of the student the first time (s)he enters a grade. So, if the student is retained in a grade and we observe two test scores, we use the first one in our analysis. We exclude students who skip a grade in this analysis, but it is important to note that there are not that many students who skip a grade in our data and the results remain virtually unchanged when we include them.

Acknowledgments

We are grateful to Ran Abramitzky, Leah Boustan, Celeste Carruthers, Christina Felfe, Amy Schwartz, Steve Trejo, and seminar and conference participants and discussants at CESifo, AEFP, and APPAM for helpful comments and suggestions. We would also like to thank Florida Departments of Education and Health for providing the data used in this analysis. All opinions expressed in this paper are those of the authors and do not reflect the views of the Florida Departments of Education and Health or our funders.

REFERENCES

Abramitzky
,
Ran
, and
Leah Platt
Boustan
.
2017
.
Immigration in American economic history
.
Journal of Economic Literature
55
(
4
):
1311
1345
.
ACT
.
2007
.
State of college readiness for Latino students
.
Iowa City, IA
:
ACT
.
Balfanz
,
Robert
.
2009
.
Putting middle grades students on the graduation path
.
Baltimore, MD
:
Everyone Graduates Center Policy and Practice Brief
.
Balfanz
,
Robert
,
Liza
Herzog
, and
Douglas J. Mac
Iver
.
2007
.
Preventing student disengagement and keeping students on the graduation path in urban middle-grades schools: Early identification and effective interventions
.
Educational Psychologist
42
(
4
):
223
235
.
Clotfelter
,
Charles T.
,
Helen F.
Ladd
, and
Jacob L.
Vigdor
.
2009
.
The academic achievement gap in grades 3 to 8
.
Review of Economics and Statistics
91
(
2
):
398
419
.
Conger
,
Dylan
,
Amy E.
Schwartz
, and
Leanna
Stiefel
.
2007
.
Immigrant and native-born differences in school stability and special education: Evidence from New York City
.
International Migration Review
41
(
2
):
402
431
.
Cortes
,
Kalena E.
2006
.
The effects of age of arrival and enclave schools on the academic performance of immigrant children
.
Economics of Education Review
25
(
2
):
121
132
.
Duncan
,
Brian
, and
Stephen
Trejo
.
2016
.
The complexity of immigrant generations: Implications for assessing the socioeconomic integration of Hispanics and Asians
.
NBER Working Paper No. 21982
.
Figlio
,
David
,
Jonathan
Guryan
,
Krzysztof
Karbownik
, and
Jeffrey
Roth
.
2014
.
The effect of poor neonatal health on children's cognitive development
.
American Economic Review
104
(
12
):
3921
3955
.
Galindo
,
Claudia
, and
Sean F.
Reardon
.
2006
.
Latino students’ educational experiences and opportunities during kindergarten
.
Tempe, AZ
:
National Task Force on Early Childhood Education for Hispanics
.
Glick
,
Jennifer E.
, and
Michael J.
White
.
2003
.
The academic trajectories of immigrant youths: Analysis within and across cohorts
.
Demography
40
(
4
):
759
783
.
Gordon
,
Milton M.
1964
.
Assimilation in American life: The role of race, religion, and national origins
.
New York
:
Oxford University Press
.
Hanushek
,
Eric A.
2009
.
The economic value of education and cognitive skills
. In
Handbook of education policy research
,
edited by
Gary
Sykes
,
Barbara
Schneider
, and
David N.
Plank
, pp.
39
56
.
New York
:
Routledge
.
Heckman
,
James
, and
Yona
Rubinstein
.
2001
.
The importance of noncognitive skills: Lessons from the GED testing program
.
American Economic Review
91
(
2
):
145
149
.
Heckman
,
James
,
Jora
Stixrud
, and
Sergio
Urzua
.
2006
.
The effects of cognitive and noncognitive abilities on labor market outcomes and social behavior
.
Journal of Labor Economics
24
(
3
):
411
482
.
Hermansen
,
Are S.
2017
.
Age at arrival and life chances among childhood immigrants
.
Demography
54
(
1
):
201
229
.
Hull
,
Marie
.
2017
.
The academic progress of Hispanic immigrants
.
Economics of Education Review
57
:
91
110
.
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
, pp.
410
477
.
Washington, DC
:
National Academy Press
.
Kao
,
Grace
, and
Marta
Tienda
.
1995
.
Optimism and achievement: The educational performance of immigrant youth
.
Social Science Quarterly
76
(
1
):
1
19
.
Park
,
Robert E.
1914
.
Racial assimilation in secondary groups with particular reference to the Negro
.
American Journal of Sociology
19
(
5
):
606
623
.
Passel
,
Jeffrey S.
2011
.
Demography of immigrant youth: Past, present, and future
.
The Future of Children
21
(
1
):
19
41
.
Portes
,
Alejandro
, and
Dag
MacLeod
.
1996
.
Educational progress of children of immigrants: The roles of class, ethnicity, and school context
.
Sociology of Education
69
(
4
):
255
275
.
Portes
,
Alejandro
, and
Ruben G.
Rumbaut
.
2001
.
Legacies: The story of the immigrant second generation
.
New York
:
Russell Sage Foundation
.
Portes
,
Alejandro
, and
Min
Zhou
.
1993
.
The new second generation: Segmented assimilation and its variants
.
Annals of the American Academy of Political and Social Sciences
530
:
74
96
.
Reardon
,
Sean F.
, and
Claudia
Galindo
.
2009
.
The Hispanic‐white achievement gap in math and reading in the elementary grades
.
American Educational Association Journal
46
(
3
):
853
891
.
Rumberger
,
Russell W.
1995
.
Dropping out of middle school: A multilevel analysis of students and schools
.
American Educational Research Journal
32
(
3
):
583
625
.
Rumberger
,
Russell W.
, and
Katherine A.
Larson
.
1998
.
Student mobility and the increased risk of high school dropout
.
American Journal of Education
107
(
1
):
1
35
.
Schwartz
,
Amy Ellen
, and
Leanna
Stiefel
.
2006
.
Is there a nativity gap? New evidence on the academic performance of immigrant students
.
Education Finance and Policy
1
(
1
):
17
49
.
Smith
,
Jonathan
,
Michael
Hurwitz
, and
Christopher
Avery
.
2017
.
Giving college credit where it is due: Advanced placement exam scores and college outcomes
.
Journal of Labor Economics
35
(
1
):
67
147
.
Stiefel
,
Leanna
,
Amy E.
Schwartz
, and
Dylan
Conger
.
2010
.
Age of entry and the high school performance of immigrant youth
.
Journal of Urban Economics
67
(
3
):
303
314
.
Suarez-Orozco
,
Marcelo M.
2000
.
Everything you wanted to know about assimilation but were afraid to ask
.
Daedalus
129
(
4
):
1
30
.

Supplementary data