Abstract
Research on generational differences in immigrant youths’ academic achievement has yielded conflicting findings. This meta-analysis reconciles discrepant findings by testing meta-analytic moderators. Fifty-three studies provided 74 comparisons on academic outcomes. First- and second-generation youths did not significantly differ on academic achievement (Hedges’s g = .01), and second-generation students performed slightly better than third-or-later–generation peers (g = .12). Moderation tests indicated that second-generation immigrants outperformed first-generation students on standardized tests (g = .20) and earned better grades than third-or-later–generation peers (g = .20). Immigrant advantage was stronger for Asian, low-socioeconomic, and community samples. Immigrant advantage may be overestimated in studies that use self-reported rather than school-reported achievement. Together, our results suggest a small, heterogeneous second-generation immigrant advantage that varies by immigrant population and study characteristics.
Immigrant youths comprise foreign-born (first generation) and native-born children with at least one foreign-born parent (second generation). They are among the fastest growing segments in the American education system. Between 1970 and 2011, the percentage of school-age children who were first- or second-generation immigrants more than quadrupled (Morse, 2005; U.S. Census Bureau, 2011a). Currently, immigrants make up 25% of the U.S. population younger than age 18 (U.S. Census Bureau, 2011a). They are increasingly settling in areas outside of traditional immigrant communities and entering public school systems that are often inadequately prepared to meet their needs (Morse, 2005). Immigrant youths are highly diverse in their English proficiency, language and culture of origin, parental educational and socioeconomic background, and other factors associated with academic achievement.
Risk factors for low achievement are disproportionately likely to affect immigrant youth, including limited English proficiency, high rates of poverty (U.S. Census Bureau, 2011b), living in segregated, low–socioeconomic status (SES) neighborhoods, and experiences of racism and discrimination (Orfield & Easton, 1996). Immigrant parents are less familiar with the American education system (Pong, Hao, & Gardner, 2005). Many immigrant youths, especially those living in poor, inner-city neighborhoods, are served by school systems that are underfunded, overcrowded, and understaffed, with high teacher turnover and school violence (García-Coll & Magnuson, 1997). In sum, family, school, and community factors that undermine student learning disproportionally affect immigrant youths.
To our knowledge, there has been no comprehensive review of extant findings on the achievement of immigrant youths. A coherent understanding of immigrant achievement would not only guide schools and policymakers in meeting the unique educational needs of this rapidly growing population but also advance theories explaining the integration process. The current review is the first meta-analysis to summarize this literature. By testing meta-analytic moderators, we reconcile conflicting results and evaluate conflicting theories. These moderators include the characteristics of achievement measures (school grades vs. standardized test scores, and academic subject), the characteristics of immigrants (race-ethnicity, SES, and age), and the study methodology (data collected from school records vs. self-reports, national vs. community samples, and the year of data collection). Below, we briefly review major integration theories and the implications of each theory for achievement differences between first-, second-, and third-or-later–generation immigrants, as well as implications for meta-analytic moderators.
Immigrant Risk or Immigrant Advantage?
Early research on immigrant achievement emphasized the challenges faced by immigrant youths, a framework that other researchers have called the immigrant risk paradigm (Crosnoe & Fuligni, 2012). Traditional assimilation theory (Gordon, 1964) suggests that the integration of immigrant groups into the mainstream is largely inevitable. Gordon posited that integration occurs linearly, with subsequent generations experience increasing socioeconomic and academic success as they adopt Anglo-American cultural values. In support of this theory, data from High School and Beyond, a national survey, indicated that first- and second-generation 10th graders had lower standardized test scores than did their third-or-later–generation peers (Glick & White, 2003).
As more data accumulated, however, findings suggested that many immigrants exhibited better health, behavioral, and educational outcomes than native youths, despite exposure to a greater number of risk factors (Georgiades, Boyle, & Duku, 2007). This finding surfaced so frequently that the term immigrant paradox was coined to describe the phenomenon. Indeed, Fuligni (1997) found that foreign-born students of Latino, East Asian, Filipino, and European descent received higher grades in both mathematics and English than their U.S.-born peers. Explanations for the paradox centered on the unique space occupied by immigrant youths, allowing them to draw on strengths from both cultures (Kasinitz, Mollenkopf, Waters, & Holdaway, 2008). Immigrant students could gain an advantage by selectively integrating, adapting to their school environment while retaining values and beliefs from their culture of origin (Rumbaut, 1990). Immigrant parents are thought to be especially optimistic about their children’s prospects for success, and immigrant children correspondingly exert more effort in school (Kao & Tienda, 1995). Aware of their parents’ sacrifices to afford them educational opportunities, immigrant youths may be driven to succeed by a sense of family obligation (Fuligni, 1997).
A variation of the immigrant paradox model was proposed by Kao and Tienda (1995), who, in analyses of national survey data, found evidence of an educational advantage of the second-generation over first- and third-or-later–generation students. They hypothesized that many recent immigrants expect upward mobility for themselves and their children, whereas later generations may be more disillusioned with that prospect, a phenomenon they termed immigrant optimism. Kao and Tienda (1995) further posited that second-generation students would have a unique advantage over first- and third-or-later–generation youths, owing to their mastery of the English language coupled with immigrant optimism.
We evaluated these various integration theories by comparing the achievement of first- and second-generation students, and the achievement of second- and third-or-later–generation students. We opted to summarize mean differences between generations so that we could detect a second-generation advantage, as proposed by immigrant optimism theory. In addition to comparing generations of immigrants, we sought to shed light on the sources of inconsistencies in the literature. Past investigations have provided evidence of both immigrant risk and immigrant advantage, suggesting a need to move beyond main effect explanations to a consideration of between-study differences that may explain conflicting results.
Characteristics of Academic Achievement as Moderators of Study Findings
To begin with, discrepant findings across studies may reflect differences in how academic achievement is assessed. The most commonly used indicators of achievement include students’ grades and standardized test scores. Both grades and test scores are designed to assess a student’s academic knowledge and skills, and the correlation between them tends to be moderate to high (e.g., Kuncel, Credé, & Thomas, 2005). As with all assessments, however, each indicator has unique method variance. Grades and test scores differ in the degree to which they are predicted by behavioral, attitudinal, and personality factors, including academic effort, educational values and aspirations, and conscientiousness (Duckworth & Seligman, 2006; Lekholm & Cliffordson, 2008; Noftle & Robins, 2007).
Although school districts are increasingly using standards-based grading, course grades have typically incorporated indicators of engagement, including attending class, being on time and prepared, completing homework assignments and long-term projects on time, participating in class discussion, and pursuing extra-credit opportunities (Duckworth & Seligman, 2006; Stricker, Rock, & Burton, 1993). To excel in courses, students need to sustain attention and effort despite distractions, boredom, and fatigue over the course of a semester (Duckworth & Seligman, 2006). Teacher subjectivity, too, is likely to play a role in most course grades. Teachers may be somewhat more likely to give engaged students the benefit of the doubt when they are on the cusp between grades (Malouf, 2008; Perlmutter, 2004). Indeed, engagement correlates more strongly with grade point average (GPA) than test scores (Marsh, Trautwein, Lüdte, Köller, & Baumert, 2005; Marsh & Yeung, 1998).
More recent immigrants exhibit lower rates of social-behavioral problems, invest more effort into their course work, and have more positive attitudes toward school (Fuligni, 1997; Suárez-Orozco & Suárez-Orozco, 1995). Compared to later generations of same-race immigrants, first-generation immigrants, on average, are perceived as more hardworking (Fuligni, 1997). To summarize, the type of indicator used to measure academic performance may moderate generational differences, with school grades favoring more recently immigrated youths and standardized test scores favoring students from later immigrant generations.
Generational differences could also vary based on the academic subjects included in measures of achievement. First-generation immigrants likely have less mastery over the English language than later generation peers (Larson, 2004), leading to an immigrant disadvantage in English subject achievement. On the other hand, achievement in mathematics may be less affected by English proficiency.
Characteristics of Immigrants as Moderators of Study Findings
We also considered the characteristics of immigrant youths as meta-analytic moderators, including their race-ethnicity, SES, and age. The focus on immigrant characteristics is consistent with segmented assimilation theory (Portes & Zhou, 1993), which posits that integration trajectories are greatly influenced by immigrants’ modes of incorporation. These refer to contextual factors such as individual and family resources, characteristics of the receiving community, and the host country’s economic and political climate. As an example, the presence of an existing successful ethnic enclave will likely extend opportunities and resources to new immigrants to achieve financial success even without full integration into Anglo-American society. By the same token, the presence of a disenfranchised subculture, as is often found in inner-city communities, may increase the likelihood of new immigrant youths associating with marginalized native youths and sharing their negative outcomes. Alba and Nee (2003), in their new assimilation theory, asserted that there is substantial variation within and across ethnic groups in the amount and forms of capital that they bring to a new context, influencing trajectories of intergenerational change. These theories highlight the potential role of immigrant characteristics in understanding heterogeneous patterns of intergenerational change.
The contexts of reception may vary across racial-ethnic groups. Ethnographies of Asian immigrants indicate that the academic achievement of these students is often well supported at the family, peer, and community levels (Caplan, Choy, & Whitmore, 1991; Zhou & Bankston, 1998). Meanwhile, Latino and Black children face greater risk for academic failure, as they are more likely to reside in neighborhoods with high rates of delinquency and community violence, face discrimination and racism, struggle against negative stereotypes regarding their academic ability, and encounter peer pressure for antischool attitudes (Pong & Hao, 2007). In addition, Latino parents, on average, arrive in the United States with lower levels of educational attainment and less English proficiency than their non-Latino counterparts (Crosnoe & Fuligni, 2012; Pong & Landale, 2012). These factors may explain why immigrant youths from Latin America have less successful achievement outcomes than students from other groups (Fernandez-Kelly & Portes, 2008; Pong & Landale, 2012). Ethnographies indicate that Black immigrant youths originating from the Caribbean Basin may experience a struggle between traditional family values and a peer culture that sometimes pulls students toward antisocial activities (Portes & Zhou, 1993). Some evidence suggests that immigrant Blacks outperform native Blacks in school (Thomas, 2009). As a proxy for modes of incorporation, we examined race as a moderator of generational differences in achievement. We expected the immigrant advantage to be strongest among Asian immigrant groups and weakest among Latino groups.
Community SES is another important predictor of immigrant youths’ outcomes (Kroneberg, 2008). Schools serving affluent communities likely have more resources than schools serving impoverished neighborhoods. SES might also affect achievement through social norms and community standards for education (Kroneberg, 2008). When the average parental education and occupation level is high, parental expectations for youth achievement are also likely to be high, and school and community attitudes tend to support education (Kroneberg, 2008). In economically distressed neighborhoods, youths who face unequal opportunities may rebel against dominant social institutions, including school (Schwartz, Kelly, & Duong, 2013). Ample evidence exists linking high community SES to student achievement (for a review, see Sirin, 2005).
The intersection of SES with the immigrant experience can be complex. Some authors have found that generational differences are magnified after controlling for parental income and education, which indicates that some immigrant groups are performing better than would be expected given their socioeconomic background (Fuligni, 1997). This could be attributed to high achievement norms among immigrants, even in low-SES communities (Zhou & Bankston, 1998). For instance, among some Asian immigrants, proeducation values may have been carried over from their countries of origin, which often have highly competitive education systems (Zhou & Kim, 2006). We expected that immigrant values would be especially protective in low-SES communities where achievement norms are otherwise low, and thus the immigrant advantage would be more pronounced in low-SES than in high-SES samples.
It is also possible that normative development (indexed by age) interacts with the immigrant experience. Academic motivation typically declines as students progress in school (Fredricks & Eccles, 2002). As classroom curricula increase in difficulty, attitudes and effort become more important determinants of success (Graham & Taylor, 2002). Middle and high school students are at greater risk for school disengagement than are elementary school students (Fredricks & Eccles, 2002). Immigrant youths’ strong achievement orientation may be especially protective against the academic disengagement of older students.
The empirical evidence regarding the impact of age on the immigrant advantage, however, has been scarce and inconsistent. Some studies suggest a growth in the immigrant advantage with age. For instance, Palacios, Guttmannova, and Chase-Lansdale (2008) showed that first- and second-generation students begin school with higher reading scores and showed faster rates of growth. Other studies suggest no age differences. Glick and White (2003) found that the trajectory of growth in reading and mathematics scores from 10th to 12th grade was similar across immigrant generational status in two national surveys. Similarly, Fuligni (1997) documented achievement advantages for immigrant youths that were consistent across students in the 6th, 8th, and 10th grades. Given the conflicting evidence, we did not formulate strong hypotheses but explored the age of samples as a potential meta-analytic moderator.
Methodological Characteristics as Moderators of Study Findings
Methodological aspects of the study can also affect results. For instance, whether studies collected data about academic achievement using school records or self-reports can have important implications. In a meta-analysis, Kuncel et al. (2005) showed that students tend to inflate their achievement when self-reporting. Moreover, the degree of inflation is not consistent across all students. The validity of self-reported achievement was greater for White than non-White students and for higher achieving than lower achieving students. Studies included in the current meta-analysis that rely on self-reported achievement, then, may show greater heterogeneity in findings that mask true intergenerational differences.
The second methodological moderator examined concerned sample selection. National and community samples differ in important ways. On the one hand, nationally representative samples often produce more accurate data than community samples, as a function of their larger sample size and their probability sampling (Yeager et al., 2011). Community samples, on the other hand, are drawn from select geographic regions and are not designed to be representative of all immigrants in the United States. These may provide more precise estimates because they represent demographically and geographically homogenous groups. Given these crucial differences, we examined sampling strategy (i.e., national vs. community samples) as a potential meta-analytic moderator.
Last, the meta-analytic technique affords us the unique capacity to consider cohort differences in study findings by testing year of data collection as a meta-analytic moderator. The number and composition of the U.S. immigrant population is continuously changing in response to immigration laws and world events. The dominant sending countries, for instance, have shifted in recent decades (for a review, see Wasem, 2013). Empirical comparisons of cohorts of immigrants are scarce. Indeed, we are aware of only one such study (Glick & White, 2003). Glick and White examined national survey data collected from a two independent cohorts of 10th graders in 1980 and 1990. After controlling for racial-ethnic background and SES, findings highlighted immigrant risk among the 1980 cohort but immigrant advantage in the 1990 cohort, in terms of achievement on standardized tests. We conducted exploratory analyses given the paucity of extant data about trends over time in immigrant achievement.
The Current Study
Our goal in the current project was to conduct the first quantitative review of the literature on academic outcomes among immigrant youths. First, we tested the immigrant risk and immigrant paradox models by comparing the achievement of first- and second-generation immigrant youths, and of second- and third-or-later–generation youths. Although we expected that effect sizes would be small and heterogeneous, we hypothesized that recently immigrated students would show slightly better outcomes than later generations. We also hypothesized that heterogeneity in our generational comparisons would be attributed to the characteristics of academic achievement. We expected that the immigrant advantage would be more pronounced for school grades than for standardized test scores, and for mathematics rather than English.
We also examined moderators related to the characteristics of immigrants. We expected that the immigrant advantage would be strongest among Asian immigrant groups and youths residing in low-SES communities. Another immigrant characteristic that we examined was age. Because data on this topic are scarce, we conducted exploratory analyses. Differences in findings across studies could also reflect study methodology. Specifically, we expected that effect sizes would vary based on whether academic data were collected via school records or self-report. We also examined the effects of sampling strategy, comparing studies that used nationally representative samples and studies of samples drawn from select communities. Finally, we considered the effects of year of data collection on study findings. By conducting the first meta-analysis of generational differences in academic achievement, we aimed to summarize the existing literature and reconcile conflicting findings, with the goal of moving forward education research on immigrant youths.
Method
Inclusion Criteria
We included all studies that (a) measured academic achievement, as indexed by school grades or standardized test scores, (b) coded generational status, and (c) provided a comparison between two generations of immigrants on at least one indicator of achievement. Because we focused on how immigrant children fare in the compulsory American educational system, eligible studies had to (d) include participants between 5 and 18 years of age, (e) be conducted in the United States, and (f) be available in the English language. We included studies described in published and unpublished manuscripts, including peer-reviewed journal articles, books and book chapters, government reports, dissertations and theses, and conference proceedings. Finally, studies had to (g) have data collected after 1965, when the Immigration and Nationality Act was passed. The Act abolished national quota systems that excluded Latin Americans, Asians, and Africans from immigrating to the United States, and dramatically changed immigration demographics.
Search Procedure
A bibliographic search was conducted using the following key terms: “immigrant,” “generational status,” “first-generation,” “second-generation,” “first generation,” “second generation,” with “academic achievement,” “academic performance,” “academic failure,” “academic outcome,” achievement test,” “test score,” “standardized test,” “educational performance,” “educational achievement,” “educational outcome,” “grades,” “scholastic outcome,” “scholastic performance,” and “scholastic achievement.” To generate these key terms, we first examined relevant manuscripts known to the authors and identified a preliminary set of search words. We then piloted the search using those words. We examined the results and added further search terms as appropriate. We repeated this iterative process until no new search terms emerged. The following databases were searched: Educational Resource Information Center, PsycArticles, PsycEXTRA, PsycINFO, Social Services Abstracts, Sociological Abstracts, Google Scholar, and ProQuest Dissertation and Theses.
As illustrated in the flowchart in Figure 1, our database search yielded 1,478 manuscripts. We reviewed the titles and abstracts of these manuscripts and identified 234 that could possibly meet inclusion criteria. Next, we reviewed the reference sections of identified eligible articles, as well as relevant reviews and books (e.g., Fuligni, 1998; Hernandez, 1999; Rumbaut, 1997). The reference trail yielded four additional potentially eligible manuscripts. In total, we retrieved and reviewed 238 manuscripts. Of these, we excluded 44 that did not report generational status, 23 that included only one generation of immigrants, 38 that did not measure grades or test scores, 19 that focused on adults, 4 that were conducted outside of the United States, 9 that were in a foreign language, 1 that reported on data collected before 1965, 6 that were theoretical or review articles, and 1 that could not be located by the library staff.
Flowchart of the manuscript search, inclusion, and exclusion process.
An additional 50 manuscripts were deemed to be eligible, but they did not report the necessary statistical values to estimate an effect size. When contact information could be located, the corresponding authors were contacted up to twice via phone or e-mail. For dissertations and theses, we also contacted corresponding committee chairs. For two manuscripts, we could not locate contact information for the author or committee chair. For 28 manuscripts, contact attempts did not yield any response. Additionally, 10 manuscripts were excluded because the authors could not locate the data necessary to estimate effect sizes. A remaining 10 manuscripts were included in our meta-analysis, with data provided directly from researchers. Thus, our search procedures yielded 53 eligible manuscripts with codeable effect size data. Independent sample t tests indicated that studies for which effect size information was obtained in manuscripts or through researchers did not significantly differ from those for which effect size data were unavailable on any of our coded demographic or methodological variables (all ps > .05).
Reliability of Search Procedure
To determine the reliability of our search procedure, both the first and second authors screened manuscripts based on our inclusion and exclusion criteria. The first author screened all 1,478 titles and abstracts that resulted from our database search. The second author screened a subset of 500 abstracts from this search. Each author independently identified manuscripts that met criteria for the meta-analysis. There was 100% agreement between the two authors.
Coding of Studies
The first and second authors independently coded all study characteristics and effect sizes. The following characteristics of academic achievement were entered as moderators: the indicator of academic achievement examined, coded as school grades or standardized test scores, and the academic subject examined, coded as mathematics or English. Notably, the available body of literature did not yield sufficient data to analyze individual academic subjects other than mathematics and English. We coded the academic subject for effects based on both grades and standardized test scores. The following characteristics of immigrant youths were also examined as moderators: the race-ethnicity of samples, coded as Asian, Black, Hispanic, White or Combination (any combination of the four racial-ethnic groups); the SES of samples; and the mean age of participants in years.
Our coding of ethnic-racial groups was admittedly broad, pooling together diverse immigrant groups. However, the available body of work precluded more homogenous groupings. Similarly, our coding of SES was limited, as data were typically provided for the entire study sample, not separately by each generation group. Last, the following methodological moderators were analyzed: the source of academic data, coded as school records or self-report; the year of data collection; and the type of sample studied, coded as a nationally representative sample or a community sample. When the year of data collection for a study was not reported in a manuscript or provided by researchers, it was coded as 2 years prior to the publication date.
Calculation of Effect Sizes
To calculate effect sizes, we generated a standardized mean difference from the means and standard deviations of different immigrant generations’ academic achievement. If means and standard deviations were not provided in reports or from researchers, the standardized mean difference was estimated using a variety of sources, including t tests, F tests, correlation coefficients, and logistic regression coefficients (for formulas, see Borenstein, Hedges, Higgins, & Rothstein, 2009; Lipsey & Wilson, 2001). Hedges’s (1981) correction, which accounts for the upward inflation of effect sizes in small samples, was applied to all effect sizes prior to analyses. To maintain consistency with the existing literature, the later generation was coded as the control group. Thus, effect sizes in the positive direction denote better academic performance among more recent immigrants.
To ensure the independence of data (Hedges & Olkin, 1985; R. Rosenthal & Rubin, 1986), manuscripts analyzing the same sample of youths were considered as a single study. Multiple estimates for the same sample were averaged to yield one standardized mean difference per study. Thus, each set of participants contributed only once to each analysis. Furthermore, studies typically compared the academic outcomes of first- versus second-generation immigrants and second- versus third-or-later–generation youths using the same sample of second-generation students. A shifting unit of analysis was adopted in such cases (Cooper, 1998); in other words, multiple effects reported for a unique set of individuals in one study were treated as independent to the extent that only one effect size from each group was pertinent to a specific analysis.
For our meta-analytic moderation analyses, effects for separate groups of individuals in the same study were treated as independent. Although it is possible for effects from different samples within the same study to share some dependencies (Landman & Dawes, 1982), these dependencies are likely to be minimal. Standard practice in meta-analysis is to regard such subsamples as nonoverlapping (Lipsey & Wilson, 2001). When data were available from the same set of participants for multiple levels of a moderator (e.g., achievement indicator, academic subject, source of academic data), we selected one level of the moderator using random number lists. For continuous moderators (e.g., age), averaging effect sizes across levels of the moderator would have reduced variability in the moderator variable.
Interrater Reliability
We calculated interrater reliability with intraclass correlation coefficients (ICCs) for continuous variables, kappas for nominal variables, and weighted kappas for ordinal variables. Reliability was acceptable for all variables (all ICCs > .84; all κs and weighted κs > .75). Discrepancies were discussed, and the resulting consensus was used in further analyses.
Data Analysis Plan
We ran separate analyses comparing the academic achievement of first- versus second-generation immigrants and second- versus third-or-later–generation immigrants. We opted to perform separate comparisons rather than summarize a bivariate correlation between generational status and academic achievement for two reasons. First, the majority of studies provided achievement data in the form of group means and standard deviations, rather than as correlations. Second, pairwise comparisons do not assume that the association between generational status and academic achievement is linear, and some theories suggest that second-generation youths may outperform both first- and third-or-later–generation students (Kao & Tienda, 1995).
Tests of meta-analytic moderation were run using an analog of the analysis of variance F test using the SPSS Macros METAF (Lipsey & Wilson, 2001) under the mixed-effects model. The mixed-effects model assumes that the variance beyond sampling error can be attributed partly to systematic differences between studies and partly to random sources. The analog F test partitions the total heterogeneity in effect sizes into a portion explained by the predictor, QΒ, and the residual pooled within groups, QW. A significant QΒ indicates that the groups differed in their mean effect sizes, whereas a significant QW means that significant within group heterogeneity still exists even after accounting for variance due to the between group predictor. Significance tests for Q have low power when the number of studies is small, so we also calculated I2, which quantifies the amount of variation in effect sizes due to heterogeneity rather than sampling error (Higgins, Thompson, Deeks, & Altman, 2003). I2 values of 25%, 50%, and 75% are interpreted as low, moderate, and high heterogeneity, respectively.
Results
Descriptive Statistics
As summarized in Figure 1, there were 53 manuscripts that met our inclusion criteria and had codeable data available. These 53 manuscripts included data from 53 independent samples, with 10 manuscripts providing data on multiple samples and 6 samples being reported on in multiple manuscripts. Based on the recommendations of Lipsey and Wilson (2001), we use the term studies hereafter to refer to these nonoverlapping samples. These 53 studies sampled a total of 101,022 participants (12,125 first-generation, 33,343 second-generation, and 55,554 third-or-later generation). The mean sample size was 253 youths (range of 4–2,807) for first-generation immigrants, 629 youths (range of 3–11,862) for second-generation immigrants, and 2,137 youths (range of 4–21,819) for third-or-later–generation immigrants.
School grades were used as a measure of academic achievement in 40 studies, and standardized test scores were used in 25 studies. Achievement in mathematics was reported in 19 studies, achievement in English in 19 studies, and achievement across several academic subjects in 41 studies. One study did not report the academic subject for any of its effects.
Consistent with U.S. immigration patterns, most studies sampled youths of Hispanic (n = 31) or Asian origin (n = 17). There were seven investigations that included Black immigrants (e.g., from the Caribbean) and seven that focused on White immigrants (e.g., from Europe). Unfortunately, we could not extract data separately by racial-ethnic group for 12 subsamples, referred to as “Combo” in Table 1. There were 28 investigations reporting on low-SES students, 15 on middle-SES students, and 3 on high-SES students; 12 studies did not describe the participants’ SES. On average, participants were 12.67 years old (SD = 4.15). Providing the variability needed for our moderator analyses, the subsamples included in our meta-analysis spanned from kindergarten through the 12th grade.
Studies included in the meta-analysis
Note. G1, G2, G3 = first-, second-, third-or-later–generation immigrants, respectively; SES = socioeconomic status; N/A = data not available; R = records; SR = self-report; C = community sample; N = national sample; CILS = Children of Immigrants Longitudinal Study; ECLS = Early Childhood Longitudinal Study; ELS = Education Longitudinal Study; HSB = High School and Beyond; NELS = National Education Longitudinal Study.
Data were obtained from self-reports in 14 studies and from school records in 33 studies. For seven studies, the source of data was not specified for any effects. Six studies described nationally representative datasets such as the National Longitudinal Study of Adolescent Health (Add Health), the Early Childhood Longitudinal Study, and the National Education Longitudinal Study. Data collection across studies ranged from the years 1969 to 2007. Most investigations included community-based samples (n = 47).
Mean Generational Differences
Of the 53 studies, 48 compared first- versus second-generation youths, and 26 compared second- versus third-or-later–generation youths. Overall, academic achievement did not significantly differ between first- and second-generation immigrants under the random effects model (g = 0.01, p = .82). Although second-generation students tended to outperform their third-or-later–generation peers (g = 0.12, p = .03), the effect size was small in magnitude (J. Cohen, 1988). Importantly, there was significant heterogeneity in the generational comparisons of academic achievement (G1 vs. G2: Q = 129.83, p < .001, I2 = 49.93; G2 vs. G3: Q = 187.70, p <. 001, I2 = 85.08). This suggests that moderator variables should be considered.
Characteristics of Academic Achievement
Academic Indicator
As illustrated in Table 2, we found support for our hypothesis that generational differences in academic performance would be significantly moderated by the indicator of achievement. Although school grades did not significantly differ between first- and second-generation students, second-generation youths outperformed third-or-later–generation peers. First-generation immigrants obtained lower standardized test scores than their second-generation peers. However, differences on standardized scores between the second- and third-or-later generation were not significant.
Summary of moderator analyses for characteristics of academic achievement
Note. CI = confidence interval. Positive effect sizes indicate an immigrant advantage, and negative effect sizes indicate immigrant risk.
p < .05. **p < .01.
Academic Subject
Our second hypothesis was that first-generation students, compared to second-generation youths, would perform less well on assessments of English, given that many do not have complete mastery of the English language (Larson, 2004). Contrary to our hypotheses, academic subject did not emerge as a significant meta-analytic moderator (Table 2).
Characteristics of Immigrant Youths
Race-Ethnicity
Our third hypothesis was that the immigrant advantage would be strongest among Asian immigrants and weakest among Hispanic immigrants. As shown in Table 3, the race-ethnicity of samples significantly moderated effect sizes across our generational comparisons of achievement. In contrast with our expectations, academic performance did not significantly differ between first- and second-generation Asian, Black, Hispanic, and White students. Consistent with our hypothesis, among Asian youths, second-generation immigrants academically outperformed third-or-later–generation immigrants. This pattern of immigrant advantage did not emerge for Black, Hispanic, or White students. There was significant evidence of immigrant advantage among our samples representing multiple racial-ethnic backgrounds, when comparing both first- versus second-generation youths and second- versus third-or-later–generation students.
Summary of moderator analyses for characteristics of immigrant youths
Note. CI = confidence interval. Positive effect sizes indicate an immigrant advantage, and negative effect sizes indicate immigrant risk.
p < .05. **p < .01.
Socioeconomic Status
Our fourth hypothesis was that the protective characteristics of being an immigrant would be most pronounced within communities of low SES. As shown in Table 3, the difference in academic performance between first- and second-generation youths of low SES was not significant. However, we found evidence of immigrant advantage when comparing the achievement of second- versus third-or-later–generation students in low-SES communities. We also detected evidence of immigrant risk when contrasting the academic performance of first versus second-generation youths in middle- and high-SES samples.
Mean Age
We also explored whether generational differences in academic achievement varied by the mean age of the sample. As depicted in Table 3, there was no significant variability in generation differences related to the age of immigrant youths.
Methodological Characteristics
Data Source
Next, we explored whether generational differences in academic achievement differed based on whether data were gathered via self-reports or school records. Because self-report data were available only for grades, we restricted our analyses to studies reporting grades. There were also insufficient data to compare differences between second- and third-generation immigrants. As depicted in Table 4, first-generation students tended to outperform second-generation youths based on self-reports, but the two generations did not significantly differ based on school records. There was also significant heterogeneity among studies with self-reported data but not among those using school records.
Summary of moderator analyses for methodological characteristics
Note. CI = confidence interval. Positive effect sizes indicate an immigrant advantage, and negative effect sizes indicate immigrant risk.
p < .05. **p < .001.
Sample Type
We conducted exploratory analyses to examine the potential moderating role of sample type (i.e., community or national sample). As shown in Table 4, sample type did not significantly explain the variation in effect sizes for our comparison of first- versus second-generation students. For our comparison of second- versus third-or-later–generation students, sample type did significantly explain the variation in effect sizes. Students from community samples, but not national samples, demonstrated a significant immigrant advantage.
Year of Data Collection
Last, we explored whether generational differences in academic achievement varied by the year of data collection. As depicted in Table 4, there were no significant generational differences in achievement based on the year of data collection.
Publication Bias
To evaluate publication bias, we first computed bivariate correlations between sample sizes and absolute effect size values for each of our significant effects (Lipsey & Wilson, 2001). Significant correlations are indicative of a publication bias and denote that larger effect sizes are more likely to be published than smaller ones. Negative correlations indicate that studies with smaller sample sizes are more likely to report stronger effects, whereas positive correlations indicate that studies with larger sample sizes tend to report stronger effects. As summarized in Table 5, studies with larger community samples tended to report larger academic achievement differences between second- and third-or-later–generation immigrants. The correlations between samples sizes and absolute effect sizes were not otherwise indicative of publication bias. To address the file drawer problem, we also calculated Orwin’s (1983) fail-safe N for each of our significant effects (Lipsey & Wilson, 2001). The fail-safe N evaluates the number of estimates with a value of zero needed to reduce the mean effect size to .05, a negligible level. Using R. Rosenthal’s (1979) criterion, a fail-safe N smaller than 5k + 10 is indicative of publication bias. As depicted in Table 5, the fail-safe N was smaller than the criterion for the majority of our analyses. Our findings may, therefore, be influenced by the file drawer problem.
Correlations between the absolute value of the effect and sample size, and fail safe Ns for statistically significant effect sizes
Note. Significant rs and fail-safe Ns smaller than 5k + 10 are indicative of publication bias.
p < .05. **p < .01. ***p < .001.
Discussion
The current meta-analysis not only represents the first quantitative summary of generational differences in academic achievement but also reconciles discrepant findings by examining moderators of these generational differences. Consistent with our hypotheses, we found evidence that first- and second-generation students obtained slightly better academic outcomes than third-or-later–generation youths. The mean differences between generations were notably heterogeneous. To understand the heterogeneity in generational differences, the current meta-analysis considered a number of moderators, including characteristics of academic achievement, characteristics of immigrant youths, and methodological characteristics of studies. Our analyses revealed that second-generation students are at a unique advantage, outperforming their first-generation peers on test scores and obtaining higher grades than their third-or-later–generation counterparts. Consistent with theories of segmented assimilation, immigrant characteristics were related to achievement differences, with stronger evidence of the immigrant advantage for Asian youths, those within low-SES communities, and select community samples. Our meta-analytic results further highlight the importance of study methodology, suggesting that the immigrant advantage may be overstated in studies that use self-reported academic data rather than obtaining measures from school records. Together, our findings suggest that although an immigrant advantage exists, this advantage is smaller than previously thought and is stronger on some measures of achievement and among some subgroups of immigrants.
Immigrant Risk or Immigrant Advantage?
Consistent with Kao and Tienda’s (1995) theory of immigrant optimism, the current meta-analysis revealed that second-generation students outperformed first-generation youths on standardized test scores (but had similar school grades), and they outperformed third-or-later–generation immigrants on school grades (but had similar test scores). The second-generation advantage may derive from these students’ protective cultural values combined with availability of resources. Protective cultural values include the dual frame of reference that allows immigrants to be positive about their situation by comparing it favorably with those still living in their native country (Kasinitz et al., 2008). Immigrant youths derive motivation to achieve from the sacrifice made by their parents to afford them opportunities to succeed (Caplan et al., 1991). Perhaps the immigrant advantage is attributable to high parental educational expectations (Fuligni, 1997) or the antiauthority and antieducation attitudes found among many mainstream American youths (Rumbaut, 1997). In support of these views, recent immigrants exhibit the fewest truancy and disciplinary problems, spend the greatest amount of time on homework, place the most importance on grades, and spend the least amount of time socializing with friends outside of school (Rosenbaum & Rochford, 2007). In one study, measures of achievement values, parental expectations, and peer support for academics accounted for a large percentage (70%) of generational differences in achievement (Fuligni, 1997).
The other factors influencing the outcomes of immigrant youths concern the availability of resources and include parental education, resources of schools and neighborhoods, and English proficiency. On average, more recent immigrants may benefit from protective values (Rosenbaum & Rochford, 2007) and later generations may benefit from greater access to resources (Kao & Rutherford, 2007). Our findings support the contention that second-generation youths may be at a unique advantage given the maintenance of protective cultural values, paired with adequate resources.
Alternatively, the better academic performance of immigrant students could also reflect the cognitive advantages of bilingualism. A recent meta-analysis showed that bilingualism is consistently associated improved attentional control, working memory, metalinguistic awareness, and abstract representation skills (Adesope, Lavin, Thompson, & Ungerleider, 2010). The dominant explanation for these cognitive advantages suggests that the regular use of two languages requires bilinguals to select the language required, inhibit intrusions from the other language, and maintain cognitive flexibility (Bialystok, 1999; Bialystok et al., 2005; Yoshida, 2008). The acquisition and maintenance of two languages that vary in structure and form facilitate an understanding of the symbolic nature of language and how language works (Bialystok, Majumder & Martin, 2003; Campbell & Sais, 1995; Galambos & Hakuta, 1988).
Indicators of Academic Achievement as Moderators of Study Findings
How academic achievement is defined and measured appears to be of crucial importance. Certainly, the predictors of grades and test scores overlap substantially, as both are designed to tap into academic knowledge and skills. Yet school grades may be affected more by attitudes and engagement. Teachers incorporate behaviors, such as effort and participation, into final grades (McMillan, Myran, & Workman, 2002). More recently immigrated youths place more value on education and expend more effort in school (Fuligni, 1997; Suárez-Orozco & Suárez-Orozco, 1995). Thus, generational differences in school grades may be a reflection of generational differences in motivation or engagement.
We also found a pattern whereby first-generation youths obtained lower standardized test scores than second-generation students. It may be that recent immigrants are disadvantaged on standardized tests because their limited English proficiency makes it difficult to understand test items, which can be linguistically complex (Duran, 1989; Garcia, 1991; Larson, 2004). Limited English proficiency can also lead to slower reading and test-taking speed (Duran, 1989; Garcia, 1991; Larson, 2004). A recent analysis of nationally representative data shows that students who enter kindergarten without English proficiency “catch up” with their same-SES peers in middle school, but many still do not meet national standards (Kieffer, 2011). Although English language learners (ELLs) may be able to compensate for this weakness in the classroom environment (e.g., by spending more time on homework or seeking help from teachers and peers), it may be more difficult to offset this limitation in time-restricted testing situations. These results may also mean that there are “true” disparities in knowledge. Immigrant youths are disproportionately more likely to be found in homes, schools, and neighborhoods with fewer resources (García-Coll & Magnuson, 1997; Suárez-Orozco et al., 2010). These factors are less likely to affect achievement at the local level (i.e., in school grades) than at the national or state level (i.e., on standardized tests).
Contrary to our hypotheses, the academic subject measured (i.e., English vs. mathematics) did not emerge as a meta-analytic moderator. We had expected that the immigrant advantage would be weaker for English than for mathematics, given that many first- and second-generation children live in homes where English is not the primary language (Larson, 2004) and coming from a non–English speaking home may put children at a slight disadvantage in English learning in school (Han, 2006). Perhaps no differences emerged because English language skills are equally important for achievement in subjects such as mathematics (e.g., to effectively understand word problems).
An alternative hypothesis is that academic subject interacts with achievement indicator, a hypothesis that could not be tested in the current study. There is some evidence, for instance, that the immigrant disadvantage in English is more apparent on standardized tests than on school grades (Kao & Tienda, 1995; D. A. Rosenthal & Feldman, 1991). Immigrant youths’ increased academic effort is more likely to help them compensate for emergent English skills in courses than on standardized tests. Given our sample size, it was not feasible to tease out the independent contributions of academic subject (i.e., English vs. mathematics) and academic achievement indicator (i.e., grades vs. test scores). Although a consensus is emerging that immigrant children are disadvantaged on standardized tests of reading (e.g., Crosnoe & Fuligni, 2012), further investigation of the predictors of achievement in various academic subjects will provide a fuller understanding of the strengths and needs of immigrant youths.
Characteristics of Immigrants as Moderators of Study Findings
As expected, we found that the immigrant advantage was stronger among Asian youths than other racial-ethnic groups. These findings are consistent with prior reports (e.g., Fuligni, 1997; Han, 2006; Kao & Tienda, 1995). For instance, Kao and Tienda (1995) found that the immigrant advantage for achievement was clear among Asian immigrants, weak among Black students, and nonexistent among Hispanic youths. The results in the current study could reflect an immigrant advantage among Asian students, or poorer achievement among later generations of Asian immigrants as compared to other ethnic groups. Prior work suggests that the former explanation is more likely, given the documented advantage of Asian immigrants, as a group, relative to newly immigrated Black and Latino families. For instance, Asian immigrants arrive with higher SES, more preimmigration resources, and greater parental educational attainment and may be less likely to face negative academic stereotypes (Crosnoe & Fuligni, 2012; Fuligni, 1997; Pong & Hao, 2007; Zhou & Kim, 2006).
When interpreting such findings, however, we must bear in mind that within major sending regions, there are differences in the educational performance of immigrant children. Chinese immigrants, for example, have been found to obtain higher grades than Filipino students (Rumbaut, 1995). Among the 14 Asian samples analyzed in the current meta-analysis, 6 were of Chinese descent, 2 were Hmong, 1 was Vietnamese, 1 was Taiwanese, and 4 had a mix of ethnic backgrounds. These numbers unfortunately do not afford us adequate statistical power to analyze differences within Asian immigrants. There is mounting evidence, however, to indicate the academic advantages of East Asian groups (e.g., Chinese and Koreans; Chang & Le, 2005; Dandy & Nettlebeck, 2002), some of which may be attributable to the high education level and financial resources of immigrant parents from these backgrounds (Crosnoe & Fuligni, 2012). Southeast Asian groups (e.g., Vietnamese, Laotians) tend to experience more structural disadvantage, and the data supporting their academic advantage are less well established (Kim, 2002). As data accumulate, the field will be in a better position to draw conclusions regarding the immigrant adjustment of specific ethnic groups within broad racial categories.
Our results indicated that generational differences also differed by other immigrant characteristics, including SES. As hypothesized, we found an immigrant advantage favoring second- over third-or-later–generation students in low-SES samples. We found the reverse effect in middle- and high-SES samples, whereby second-generation youths are performing less well than their third-or-later–generation counterparts. The educational aspirations of the community in which a student lives are linked to his/her achievement (Kroneberg, 2008). Accordingly, integrating into certain segments of contemporary youth society may entail some academic risks if these youths do not value achievement (Kroneberg, 2008). In low-resource settings, greater achievement strivings may be necessary for success. Conversely, students living in adequately resourced communities and attending adequately resourced schools may be more likely to endorse proeducation attitudes. In that case, differences in proeducation values between recent immigrants and later generations are likely to be less pronounced. In such communities, later generations may have greater English proficiency, more knowledge of the educational system, and other advantages that may facilitate their performance.
Methodological Characteristics as Moderators of Study Findings
Our results suggested that the immigrant advantage was stronger among samples consisting of more than one racial-ethnic group. We found medium effects favoring first- over second-generation students (g = .32) and second- over third-or-later–generation youths (g = .42) among mixed-race samples. In contrast, we found that achievement was similar across generations when the study sample was entirely Asian, Black, Hispanic, or White. This finding was unexpected, leading us to further examine the six studies that did not disaggregate generational differences by race. Our review showed that the majority of studies used samples that comprised primarily one ethnic or racial group but that also contained a significant portion of other racial groups. For instance, Mateu-Gelabert (1998) had a sample that was 75% Hispanic, 24% Black, and 1% Asian. Of the six studies, two had majority Hispanic samples, two had majority Black samples, one had a majority Asian sample, and one had no majority racial-ethnic group sample. These studies were typically designed not to directly answer questions about generational differences in achievement but to shed light on a related educational issue for immigrant youth. For instance, Plunkett, Behnke, Sands, and Choi (2009) examined the impact of family involvement in education. What accounts for the larger effect sizes for generational differences in achievement among this group of studies is not clear.
Our current meta-analysis yielded evidence that conflicting findings in the literature can be explained by methodological differences between studies, including measurement strategies and sample selection. Our results suggested that the immigrant advantage might be overestimated among studies that relied on self-reports of school grades rather than obtaining such measures from the participants’ school records. Perhaps some immigrant youths are more prone to inflate estimates of their own grades because they place more importance on school achievement or because a lack of familiarity with the education system makes them less aware of their own performance.
Similarly, studies using community samples reported stronger effect sizes than those with nationally representative samples. Researchers interested in studying the adjustment of immigrant youths may be more likely to select community samples wherein immigrant youths are doing better than would be expected. Researchers may also be selecting community samples with a critical mass of immigrants with shared linguistic or cultural characteristics that function as community resources to foster more positive outcomes (Portes & Zhou, 1993). Contrasting this with the weaker evidence of the immigrant paradox in national samples suggests that these communities are not entirely representative of all immigrant youths, and underscores the heterogeneity of adjustment outcomes among this population.
Limitations
A number of potential limitations of this project should be considered. As with all meta-analyses, our conclusions are valid only to the extent that the included studies represent the population of research. That is, our results may be biased if the included studies do not represent a random sampling of the population, as with the file drawer problem. Tests of publication bias using Orwin’s (1983) fail-safe N and R. Rosenthal’s (1979) criterion indicated that our findings may be affected by the file drawer problem. Although standard practice, the use of Orwin’s fail-safe N and R. Rosenthal’s criterion may yield conservative findings (Card, 2012). First, Orwin’s (1983) fail-safe N may be inaccurate, as it does not model heterogeneity.
Second, it assumes that the excluded studies in a meta-analysis have the same average sample size as the included studies. If the excluded studies tend to have smaller samples than the included studies, then the fail-safe N is underestimated. Third, there are no solid guidelines for determining how large Orwin’s fail-safe N should be. Although R. Rosenthal’s (1979) criterion provides a frequently used guideline, it is likely an overestimation. Consistent with these limitations, correlations of sample sizes and absolute effect sizes in our meta-analyses were, for the most part, not indicative of publication bias. Additionally, our inclusion of unpublished manuscripts and our use of random effects analyses allows for greater generalizability of findings compared to a sole reliance on published data and fixed effects models (Card, 2012).
As we proceed with research in this field, it is important to place the observed effect sizes in context. When we did find significant differences between generations, effect sizes for g tended to hover around 0.20. Such differences are considered small by most statistical standards (J. Cohen, 1988). Based on typical means and standard deviations of GPA (e.g., Kao & Rutherford, 2007; Padilla & Gonzalez, 2001), we calculated that a g of 0.20 approximately translates to a 0.16 difference in GPA on a 4-point scale. On standardized achievement tests, including the SAT, a g of 0.20 is equivalent to a 20-point difference for an examination with a mean of 500 and a standard deviation of 100. These differences are small in comparison with the effects of SES and race.
In a meta-analysis, Sirin (2005) found the achievement differences between low- and high-SES students to be large (g = .61). The Black-White test score gap is also substantial, ranging from 0.75 to 1.00 (Vanneman, Hamilton, Anderson, & Rahman, 2009). In comparison, the differences between generations of immigrants are much smaller. However, they can be quite meaningful. In a survey study of college admissions officers, one third of postsecondary institutions reported that a score increase of 20 points on the SAT mathematics subtest or 10 points on the SAT critical reading subtest would significantly improve students’ likelihood of admission (Briggs, 2009). Another way to contextualize the magnitude of our effects is to consider intervention programs aimed at improving student achievement. These programs often yield effect sizes in the 0.3 to 0.4 range (e.g., G. L. Cohen, Garcia, Apfel, & Master, 2006). Thus, the effect sizes documented in this meta-analysis suggest an immigrant advantage that is on par with the benefits of a targeted intervention.
Future Directions
The current study examined group differences between first-, second-, and third-generation immigrant youths. These intergenerational comparisons comprise a dominant methodology in the literature on immigrant youths, making it possible for us to conduct a quantitative summary of this work. Nevertheless, we cannot rule out the possibility that immigration patterns, changes in schooling, or other societal trends may explain the documented generational differences. The processes of change, adjustment, and acculturation can be gleaned only through longitudinal work that allows researchers to tease out the immigrant experience from normative developmental changes. By conducting this meta-analysis, we aimed to provide a broad-based summary of the existing literature, simultaneously delineating the gaps in our knowledge and spurring further investigation in these areas.
The present meta-analysis considered several moderators that might account for discrepancies in generational differences for academic achievement. Our results underscored that the detection of immigrant advantage varies by immigrant population and study characteristics. However, additional features of immigrants and studies may be considered in future research. For instance, our quantitative synthesis did not consider gender differences as the majority of research is composed of gender-balanced samples; this lack of variability limits the statistical power of a meta-analysis. Yet there is some evidence that the female advantage on academic outcomes in primary and secondary school (Fuligni, 1997; Kao & Tienda, 1995) might be more pronounced among immigrants (Brandon, 1991).
Another important moderator variable that could not be examined in this study was the English proficiency of the sample. Lack of mastery of academic English impedes academic performance (U.S. Department of Education, 2014). What further complicates the issue is that ELLs are not limited to first-generation students. In fact, the majority of ELLs are born in the United States (Capps, Fix, Murray, Passel, & Herwantoro, 2005). American-born students comprise 76% of English language learners (ELLs) in elementary school and 56% of ELLs in middle and high school (Capps et al., 2005). Most investigations of immigrant generation do not examine ELL status, and vice versa. The siloing of these two literatures makes it impossible to determine how much of the generational differences in achievement is due to low English proficiency or, conversely, to the cognitive advantages conferred by bilingualism.
Another major gap is our understanding of the experiences of undocumented immigrants. One can imagine that students without immigration documentation perceive huge barriers to attending college and gaining meaningful employment, and that it may be difficult for such youths to maintain enthusiasm toward learning and persistence in school in the face of hopelessness about the future and the lack of connection between school success and life success. Many barriers exist in recruiting and retaining undocumented immigrants as participants, yet knowledge about their experiences can have important implications for policy. As our knowledge about the immigrant experience builds, we may benefit from a consideration of person-environment fit as a determinant of the academic success of immigrant youths. Suárez-Orozco et al. (2010), for instance, demonstrated that some immigrant youths who are initially successful in school show a drop in their achievement after transitioning to highly demanding school environments before they acquire the language skills necessary to succeed in that environment. In other words, a child can struggle in one school setting but be successful in another. Through continuing empirical investigation, we can work toward developing a framework that identifies the needs of diverse immigrant students and the family, school, community, and political factors that promote their academic adjustment.
We were unable, in the current meta-analysis, to examine the simultaneous contribution of multiple moderator variables. We have reported the moderating effects of each of seven moderator variables independently. Immigrant generation in the United States is often correlated with race-ethnicity, cohort differences, and SES. Similarly, most studies in this area do not disaggregate analyses or findings by important outcome variables, such as academic subject or academic indicator. It is our hope that the current project will guide future investigators in identifying potential confounding variables when examining immigrant achievement, as well as delineate gaps of knowledge.
Conclusions
This study’s findings help identify the unique academic strengths and needs of the immigrant population. We are certainly not the first to note the significant heterogeneity of the growing number of children we have so far grouped as immigrants. However, the meta-analytic methodology employed enabled a unique synthesis of the current literature and allowed us to shed light on some of the discrepancies in the literature. The results of the current study provided support for immigrant optimism theory (Kao & Tienda, 1995), as well as theories that emphasized heterogeneity in integration trajectories, including the segmented assimilation theory (Portes & Zhou, 1993) and Alba and Nee’s (2003) new assimilation theory. Consistent with these theories, we found no generational differences overall in nationally representative samples but evidence of the immigrant advantage among select communities.
An important contribution of the current project is the identification of factors that account for this heterogeneity. Specifically, we found that characteristics of immigrants (including race-ethnicity) and characteristics of the integration context (as captured by SES) both affect immigrant achievement. Our finding that the immigrant advantage is stronger on grades than on test scores suggests that proeducation values may be one mechanism behind immigrant advantage. At the same time, the immigrant youths’ protective values are not always able to compensate for a lack of resources in their schools and neighborhoods. The picture emerging from this area of research is that the risk factors that take a toll on the engagement and achievement of nonimmigrant youths, such as poorly resourced, violent, and segregated schools, also take a toll on the engagement and achievement of immigrant youths (Suárez-Orozco et al., 2010). Some immigrant children have an especially high exposure to these risk factors. Yet our findings also point to resilience, overall, among these children. We hope that the current findings will spur other programs of research that will enable us to understand the heterogeneity of immigrant youth, so that we can attend to the factors that may impede their success, and capitalize on their unique strengths.
Footnotes
Notes
Authors
MYLIEN T. DUONG, PhD, is a Ruth L. Kirschstein postdoctoral fellow at the University of Washington, 2001 Eighth Avenue, Suite 400, Seattle, WA 98121; e-mail:
DARYANEH BADALY, MA, is a doctoral student in the Clinical Science program of the Department of Psychology at the University of Southern California, mailing address: 2101 Commonwealth Boulevard, Suite C, Ann Arbor, MI 48105, USA; e-mail:
FREDA F. LIU, PhD, is an acting assistant professor of Psychiatry and Behavioral Sciences at the University of Washington School of Medicine, mailing address: Mailstop OA5154, 4800 Sand Point Way NE, Seattle, WA 98105, USA; e-mail:
DAVID SCHWARTZ, PhD, is an associate professor of Psychology at the University of Southern California, mailing address: SGM 501, 3620 South McClintock Avenue, Los Angeles, CA 90089, USA; e-mail:
CAROLYN A. MCCARTY, PhD, is a research associate professor of Pediatrics at the University of Washington, 2001 Eighth Avenue, Suite 600, Seattle WA 98121, USA; e-mail: