This article has been retrieved
times since May 30, 2000
This article has been retrieved
Education Policy Analysis Archives | ||
Volume 8 Number 25 |
May 30, 2000 |
ISSN 1068-2341 |
|
Editor: Gene V Glass, College of Education Arizona State University
Copyright 2000, the
EDUCATION POLICY ANALYSIS ARCHIVES. Articles appearing in EPAA are abstracted in the Current Index to Journals in Education by the ERIC Clearinghouse on Assessment and Evaluation and are permanently archived in Resources in Education. |
Mathematics Achievement by
Immigrant Children:
|
|
Abstract In this study, I examined academic achievement of immigrant children in the United States, Canada, England, Australia, and New Zealand. Analyzing data from the Third International Mathematics and Science Study (TIMSS), I gauged the performance gaps relating to the generation of immigration and the home language background. I found immigrant children's math and science achievement to be lower than the others only in England, the U.S., and Canada. Non-English language background was found in each country to relate to poor math and science learning and this disadvantage was stronger among native-born children—presumably children of indigenous groups—than among immigrant children. I also examined the school variation in math performance gaps, using hierarchical linear modeling (HLM) to each country's data. The patterns in which language- and generation-related math achievement gaps varied between schools are different in the five countries. |
|
The public school system as an institution plays a critical role educating immigrant children and facilitating their participation in the larger society. This system in the U.S., succeeded in integrating European immigrants, is now facing a serious challenge as newcomers of non-European heritage have become the primary source of immigration over the decades. This shift in origins of the immigrants is a most striking development in U.S. immigration history (Fix, Passel, Enchautegui, & Zimmermann, 1994). Asians and Hispanics are the fastest growing groups among foreign-born population in the U.S., rising from 1.5 percent each in the early 1990s to 25 percent and 43 percent, respectively, in 1990 (Bureau of Census, 1993). Asian and Hispanic children, respectively, represent 3.5 percent and 14 percent of the U.S. elementary and secondary student enrollment in 1992, more than doubled from the 1.2 and 6.4 percents in 1976 (NCES, 1995). Many developed nations share this challenge. The trend of globalization has brought rising waves of foreign labors, refugees, and immigrants into affluent countries. Today, the U.S., Canada, Australia, New Zealand, France, Germany, Britain, and other European countries are receiving newcomers from different regions of the world. The public schools in these countries confront the daunting task to educate children of immigrants. Given the gravity of the issue, ironically, educators know little about the schooling of immigrant children. Little research has systematically dealt with the issue. It is unclear as to how the new generations of immigrants do in the school system and what their great diversity has to do with their schooling. It is even more uncertain about how schools are acting to help immigrant children learn math and science, subjects that are critical for competing in today's technology-oriented labor market. No baseline comparison is available regarding education of this group in the U.S. and other nations. The lack of knowledge about immigrant children's education and general well being concerns educators and policymakers. The Federal Interagency Forum on Child and Family Statistics has published annual reports on children (Federal Interagency Forum on Child and Family Statistics, 1998). But the reports contain little information specifically about children of immigrant background. A recent study of immigrant children released by the National Research Council and the National Institute of Medicine points out that there is virtually no public dissemination of information on even the most basic indicators of the conditions of children in immigrant families (Hernandez & Carney, 1998). In a policy study report, the National Commission on Immigration Reform also calls for increased attention to and resources for immigrant children's schooling (see Schnaiberg, 1997). My study was intended to remedy this shortage of knowledge by comparing math and science performance of immigrant children in five English-speaking countries. |
Literature Review and Research QuestionsThe available research on immigrant children's school performance is inconclusive even regarding the basic conditions of performance. Some studies suggest that the children of immigrants do better in school than the rest of American children; their performance is above averages (Rumbaut, 1996; also see Viadero, 1998, Lapin, 1998). In social adaptation, physical and mental health, foreign-born immigrant children were also seen to fare at least equally well as other children in the U.S. (Hernandez & Charney, 1998). On the other hand, there is evidence that immigrant children, especially Hispanics and others with impoverished background, suffer poor academic achievement and lower educational attainment (e.g., McPartland, 1998; Vernez & Abrahamse, 1996). A foremost concern for research is to provide clear description of this population's schooling with solid baseline indicator of performance.Aggregated comparisons may mask crucial variation within the immigrant population. For example, while Hispanic adolescents of all generations have grade point averages and math test scores that are lower than those of white adolescents in U.S.-born families (NCES, 1998a), academic achievement of immigrant students appears to decline by generations (Hernandez & Charney, 1998). The social, economic, and cultural factors that either protect or disadvantage immigrant children are not well understood. Thus, baseline indicators should also summarize performance differences by important subcategories of the immigrant children, such as generation of immigration, sex, native language, and socioeconomic status. A small number of recent studies of immigrant children's academic achievement provide some insights for understanding the variation among immigrant children's academic achievement. For example, Hao and Bonstead-Bruns (1998) used the concept social capital to explain immigrant children's academic performance. This concept, though useful in understanding the behavioral and cultural attributes of immigrant groups affecting academic learning, is less relevant to study of the functioning of institutions, such as public schools. It is not clear from such research as to how schools could reduce the detriment caused by meager social capital for an immigrant child. Theories and research are needed to sort out institutional factors that account for the wide variation and the changing pattern of this population's academic performance. As a preliminary study intended to address some of these concerns, I examine the following issues in the analysis. Generation DifferenceThe generation of immigration distinguishes a number of demographic characteristics among children from immigrant families. Compared with children in U.S.-born families, first-generation immigrant children (the foreign-born) are more likely to experience high poverty; to have a large family with both parents; and their parents are more likely to have attained little education yet to participate in labor force (Hernandez & Charney, 1988). Second-generation children (those born in the U.S. to at least one foreign- born parent) tend to experience substantially less risk than do first-generation children, but are likely to lose psychological resilience that the first generation often demonstrates. Such cross-generation distinctions imply different risks and strength for immigrant children's schooling.The analysis first addresses the question about the performance gap relating to the generation of immigration in different countries. The second question is to what extent this gap differs across schools in each country. To answer this question, the analysis explores the variation of the generation gap across schools in each country. With the international test results available from the data, it should be particularly interesting to see how the school- level variation of the gap differs across countries. The resulting baseline indicator may reveal the extent to which the overall school setting relates to the variation of the gap—in contrast to the extent to which individual factors account for the variation. Future study may elucidate school roles in reducing the generation gap by examining specific school factors relating to the variation of the performance gap. Language BarrierLimited English proficiency handicaps immigrant children's learning on key subject areas such as mathematics and science. Language barriers are often more detrimental for children of low socioeconomic background. Living in socially and linguistically isolated communities, poor immigrant children can hardly improve their new language skills and the language barriers persist over the school years. On the other hand, bilingual proficiency, defined as the mastery of both the mother tongue and a new language, is found to be a strength for immigrant children's cognitive growth (e.g., Bumberger & Larson, 1998; Hao & Portes, 1998).I first estimate the size of the math and science performance gaps related to non-English language background in each country. I then examine the variation of the gaps between schools in each country. While these baseline indicators are descriptive, they imply the extent to which the overall school context is associated with the variation of the gap-- relative to the individual level variation. The analysis may provide a ground for further study of specific school functions in reducing language-caused performance gap for immigrant children. School Variation of Performance GapDoes school has something to do with the performance gaps? It is conceivable that the average performance and the performance gap between immigrant children and the other children may vary across schools. Schools with different demographic composition, resources, and curricular and instructional programs theoretically could achieve different levels of excellence and equity. Relevant to policymaking, gauging such school-level variation is crucial for further assessing institutional role in achieving educational equity. Understanding the school-level variation in performance gaps and school features relating to such variation can help school improve equity. In this preliminary analysis I only examine the school-level variance in math achievement gap relating to the generation of immigration and language backgrounds. |
Data SourceTIMSS is the most comprehensive and rigorous international education comparison ever (NCES, 1998b). I extracted TIMSS Population 1 (students of grades 3-4 or ages 8-9) data of five English-speaking countries including the U.S., Canada, England, Australia, and New Zealand, with unweighted samples size of, respectively, 10,670, 14,639, 5,584, 10,433, and 4,670. Conducted in 1995, TIMSS researchers tested the mathematics and science knowledge of more than half a million students in 41 countries at three grade levels—primary, middle, and end of secondary school. TIMSS ensured that the participating students in each country were representative of its population. It generated information on the background and math and science achievement tests for children of the participating countries. While tests on math and science were administered to students, survey data were collected from teachers, schools, as well as students. The resulting information encompasses student demographic background and math learning experience; teachers' background and instruction; and school facilities, program provisions, and demographic attributes. Information for identifying foreign-born children is available, including the nation of birth for both the parents and the child.The TIMSS nationally representative sample designs generated data for the population of each target age group (or grade level) in a country. The sample for a given age group in a country was selected in a two-level stratified design. In this design, a school sample representative to the national population of schools was drawn first, and within each selected school, typically one classroom at the target grade level was selected for the test and survey. While certain minority groups were oversampled, sample weights were provided to compensate the bias resulting from the oversampling. Unit nonresponse bias was corrected by sample weights as well. The tests were designed through collaboration among experts from the participating countries. Recognizing vast differences in social and educational context, the tests were meant to measure students' general math and science knowledge and skills at the given age/grade. The results were widely accepted as valuable for cross-national comparison, given the caution of contextual differences among the participating nations (Forgione, 1998). Four items were used to identify students' immigrant background. They presented information about the child's birthplace (foreign- or native-born in one of the five countries), the number of years living in the current country, and the foreign-born status of the child's mother and father. I defined a child as a first-generation immigrant if the child was foreign- born regardless of the birthplace of the parents, and a second-generation immigrant if the child was born in the current country to one or both foreign-born parent; and the rest were considered as non-immigrants. With a data item about student home languages, I categorized students as a non-native language speaker if he or she reported that a language other than the TIMSS test language (English) was “often” or “always” spoken at home. Analytical MethodsThe analysis included two components. To generate baseline indicators of the overall performance patterns, I ran a series of descriptive analysis. To estimate school-level variance of performance gaps, I conducted two-level hierarchical linear modeling.Descriptive AnalysisDescriptive analysis entailed comparing means of the test scores for the groups of interest. As specified earlier, baseline indicators of math and science performance gaps between immigrant and non-immigrant children will be estimated in a comparison of means with significance tests (all at the p<. 05 level if not otherwise specified). All the remaining indicators will be generated by breaking down the test data by two categorical variables, immigrant status and non-English language background, with significance tests.I ran the procedure with data for each country. The five plausible values for estimating performance on mathematics were used. The estimates from the five runs were then averaged as the final estimates in the comparisons (see TIMSS User's Guide for rationale for this special approach, International Study Center, 1988). Student-level sample weight (TOTWGT) was used to correct bias from unequal sampling of some student groups and unit nonresponse. I used jackknife procedures to correct the design effects caused by the stratified clustering sample design (rather than simple random design). See Chapters 5 and 7 of the User's Guide (International Study Center, 1998) for rationale of using sample weights and special procedures for correcting design effects. HLM ProcedureTo assess school-level variance of the performance gap relating to immigrant status, I used hierarchical linear modeling (HLM) technique (Bryk & Raudenbush, 1992). HLM was appropriate for this part of analysis because in the TIMSS design students as level-1 units were nested in schools (level 2) and HLM enabled me to separate the variance by two levels and to formally estimate the portion of variance taking place at school level.In an unconditional (one-way ANOVA) with random effect model, I estimated variance separately at the student and school levels. This model answered the question as to whether schools differed from each other in average math performance. It provided basic estimates for making decision if it was necessary to further model the variance at the two levels. The unconditional models were:
As the school level variance was sufficiently large (10 percent or more of the total variance, measured with the intraclass correlation coefficient) for each country, I specified random coefficient models to estimate school-level variance of the math achievement mean and achievement gaps associated to home language and the first- and second generation immigrant backgrounds (all student-level predictor variables were centered around the school mean). At level 1, the equation had the overall achievement mean, the average achievement differences relating to the non-English language and immigrant status, and the random error, Yij = ß 0j + ß 1j (LANGUAGE) + ß 2j (FIRST_G) + ß 3j (SECOND_G) + r ij . At level-2, the equations included no school variables but only the school average math score (the intercept) and the estimates of the variance around the average measures of the three gaps (the slope): ß qj = µ q0 + u qj where q=1, 2, 3. In case the gaps did not vary statistically significantly at the school level, the random effect u qj was removed from the equation and the effect was estimated only as fixed. I used the software package HLM (version 4.03) for the analysis, running the Plausible Value procedure available from the package (Bryk, Raudenbush, & Congdon, 1996). This procedure included the five plausible values as the outcome variable and automatically averaged the resulting estimates after the runs. Normalized student level weight and school level weight were used in the procedures for generating the estimates to the student population in each country. |
FindingsStudents of Immigrant and Non-English BackgroundsEach of the five countries' elementary student populations contained a substantial portion of students with immigrant and non- English backgrounds (see Figure 1). Australia and New Zealand had the highest rates of immigrant students of both the first and second generations, followed by Canada and the U.S. Strikingly, the second generation immigrant children comprised almost one third of Australia's population of third and fourth graders. The U.S. had a relatively high proportion of children of non-English background (16.7 percent), though this group was fairly large in Canada and New Zealand as well.
 Performance Gaps Associated with Immigrant StatusThe math achievement gaps to the disadvantage of immigrant students took place only in England, the U.S., and Canada, not in Australia and New Zealand. This pattern is particularly evident in the gap between non-immigrant and the first generation immigrant children (Figure 2). In England, the gap in math score was 41 point, in the U.S., 60, and in Canada, 44, all statistically significant; whereas in Australia and New Zealand, the gap was not observed. In the U.S. and Canada, the non-immigrant children scored higher than the second-generation immigrant children; but in England, this difference was not statistically significant.
 The patterns of performance gaps associated with immigrant status were similar in science (Figure 3). In short, immigrant students lagged behind in math and science learning in England, the U.S., and Canada, but they did not in Australia and New Zealand.
 |
Immigrant and Language BackgroundNon-English home language is clearly a disadvantage to students' math and science learning regardless of immigrant status. In each group (non-immigrant and first and second generations of immigrant children), those whose home language was not English averaged substantially lower score than the rest of the students in math (Tables 1). Further, the language disadvantage was more acute among native-born children than among immigrant children. Consistent in each country, the second-generation immigrant children with non-English home languages did better in math than the non-immigrants with non-English home languages. I speculate that the latter was likely to be the indigenous groups or the groups that experienced persistent social and linguistic isolation, e.g., the American Indians and Hispanics in the U.S. Unfortunately, TIMSS contains no data to allow me confirm this assumption. With exception of the U.S. and Canada, this pattern holds between the first generation immigrant children and non-immigrants as well, though to a lesser extent.Average Math Achievement Scores by Immigrant Status and Home Language: TIMSS Population 1 (grades 3 or 4 and age 9 or 10) in the Five Countries
Two-level AnalysisSchool level variance was substantial and statistically significant in all the five countries (Table 2). As indicated by the intraclass correlation coefficients, school-level variance proportional to the total variance around the given country's average math achievement ranged from 9 percent (Canada) to 26 percent (New Zealand). This finding suggests that to a considerable extent, students' math scores in each of the five countries tended to cluster around their school average scores. The reliability of the achievement measure was around 0.80, with exception of Canada, where the estimate was only 0.49. These baseline statistics justified further two-level modeling to examine the performance gaps relating to the language and immigrant status.Table 2
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| Parameters | England | U.S.A. | Canada | Australia | New Zealand |
| Average school mean g00 | 483.51 | 505.18 | 501.88 | 522.53 | 470.85 |
| Reliability of the dependent variable |
0.88 | 0.82 | 0.49 | 0.75 | 0.86 |
| Intraclass correlation | 0.19 | 0.18 | 0.09 | 0.16 | 0.26 |
| School-level variance uoj | 1,700.63 | 1,755.81 | 1,349.94 | 1,864.95 | 2,102.65 |
| Parameter | England | U.S.A. | Canada | Australia | New Zealand |
Student-Level Effects (Level-1 models) |
|||||
| Intercept (overall mean achievement) µ00 | 482.47*** | 504.95*** | 501.86*** | 522.46*** | 470.85*** |
| Non-English language difference µ10 |
-38.26*** | -39.15*** | -14.09*** | -20.68*** | -30.06*** |
| First generation immigrant difference µ20 |
-21,87*** | -26.95*** | -44.75*** | 13.64** | 2.42 |
| Second generation immigrant difference µ30 |
8.42* | -4.25*** | -12.21*** | -7.61* | -5.46 |
School-level variance (Level-2 models) |
|||||
| School mean achievement, u0j | 1710.05*** | 1783.27*** | 1352.19*** | 1868.44*** | 2087.59*** |
| Non-English language difference, u1j |
674.44* | ||||
| First generation immigrant difference, u2j |
|||||
| Second generation immigrant difference, u2j |
1076.18*** | 421.99** | |||
| Student level variance (Level-1 random effect), rij |
7121.75 | 7387.03 | 13760.00 | 10017.03 | 5988.68 |
* p<.05; ** p<.01, *** p<.001
The symbol "—" indicates that the random variance was too small to model and thus the associated variable was specified only as a fixed effect in the model.
Note: All student-level predictor variables were centered on school means.
Booth, A., Crouter, A. C., & Landale, N. (Eds.). (1997). Immigration and the Family: Research and Policy on U.S. Immigrants. Mahwah, NJ: Lawrence Erlbaum Associates, Inc.
Bryk, A.S., & Raudenbush, S.W. (1992). Hierarchical Linear Models: Applications and Data Analysis Methods. Newbury Park, CA: Sage.
Bryk, A. S., Raudenbush, S.W., & Congdon, R. T. (1996). HLM: Hierarchical linear and Nonlinear Modeling with the HLM/2L and HLM/3L Programs. Chicago, IL: Scientific Software International.
Duran, B. J.; Dugan, T. & Weffer, R. E. (1998). Language minority students in high school: the role of language in learning biology concepts. Science Education, 82, 3, 311-41.
Federal Interagency Forum on Child and Family Statistics. (1998). America's Children: Key National Indicators of Well-being. Website: http://childstats.gov/ac1998/ac98.htm.
Fix, M., Passel, J. S., Enchautegui, M. E. & Zimmermann, W. (1994). Immigration and Immigrants: Setting the Record Straight. Washington DC: Urban Institute.
Forgione, P.D. (1998). What We've Learned From TIMSS About Science Education in the United States. Washington, DC: National Center for Education Statistics. Website http://nces.ed.gov/Pressrelease/science/index.html.
Hao, L. & Bonstead-Bruns, M. (1998). Parent-child differences in educational expectations and the academic achievement of immigrant and native students. Sociology of Education, 71, 3, 175-198.
Hao, L. & Portes, A. (1998). E Pluribus Unum: Bilingualism and loss of language in the second generation. Sociology of Education, 71, 4, 269-294.
Hernandez , D. J. & Charney, E. (Eds.). (1998). The Health and Well-being of Children in Immigrant Families. Washington, D.C.: National Academy Press
Huang, G. G. & Weng, S. (1998). Minority post-secondary education attendance, high school desegregation and student characteristics. Race, Ethnicity and Education, 1, 2, 241-265.
International Study Center. (1998). User's Guide for the Third International Mathematics and Science Study (TIMSS) and U.S. Augmented Data Files. Boston, MA: Boston College.
Lapin, L. (1994). Study: Immigrants' Kids Learn Better. The Sacramento Bee, February 23, 1994, p.1.
McPartland, J. (1998). Project #7126: The Adaptation of Immigrant Children in the American Educational System Center for Research on the Education of Disadvantaged Students (CDS). Johns Hopkins University website http://www.csos.jhu.edu/default1.htm.
NCES (1998a). The Condition of Education 1998. Washington, DC: the author.
NCES (1998b). Pursuing Excellence: A Study of U.S. Twelfth-Grade Mathematics and Science Achievement in International Context. Washington, DC: the author.
NCES (1995). The Condition of Education 1995 / Supplemental Table 40-2 Enrollment in public elementary and secondary schools, by race/ethnicity: 1976, 1984, 1986, 1988, 1990, and 1992 on website: http://nces.ed.gov/pubs/ce/c9540d02.html.
Oakes, J. (1990). Opportunities, Achievement. and Choice: Women and Minority Students in Science and Mathematics. Review of Research in Education, 16, 2, 153-166.
Rumbaut, R. G. (1996). The New Californians: Assessing the Educational Progress of Children of Immigrants. CPS Brief, 8, 3, 1-14.
Rumberger, R. W. & Larson, K. A. (1998). Toward explaining differences in educational achievement among Mexican American language-minority students. Sociology of Education, 71, 1, 69-93.
Shavarini, M. K. (1996). California's Immigrant Children, book review. Harvard Educational Review, 66,3, 668-674.
Schnaiberg, L. (1997). Panel urges greater focus on immigrant children's needs. Education Week 17, 29, p. 18
U.S. Bureau of Census. (1993). We the American…Foreign Born. Washington DC: the author.
U.S. Congress. (1980). Science and Engineering Equal Opportunities Act, Section 32(B), Part B of P.L. 96-516. Washington, DC: the author.
Valdles, G. (1998). The World outside and inside schools: language and immigrant children. Educational Researcher, 27, 6, 4-18.
Van, J. (1994). Immigrant Children Get Better Grades, Study Finds. The Chicago Tribune, February 23, 1994, p. 6.
Vernez, G. & Abrahamse, A. (1996). How Immigrants Fare in U.S. Education. Santa Monica, CA: RAND Center for Research on Immigration Policy.
Viadero, D. (1997). Immigrant children succeed despite barriers, report says. Education Week, 17, 29, p. 14, Appendix
Wainer, H. & Steinberg, L. S. (1992). Sex Differences in
Performance on the Mathematics Section of the Scholastic
Aptitude Test: A Bidirectional Validity Study. Harvard
Educational Review, 62, 3, 323-336.
Senior Research Analyst
Synectics for Management Decisions, Inc.
North Moore St.
Arlington, VA 22209
Phone: 703.807.2324 703.528.2857
Email: garyh@smdi.com
Gary Huang is a research sociologist working in education and public policy related areas. His research interests include minority children's socialization and schooling, institutional effects on learning, rural education, and cross-cultural research and communication. He works mainly with large-scale survey data analyses but has a broad interest in other approaches to research. Supported by a grant from U.S. Department of Education, he is currently working on a cross-dataset synthetic analysis looking at the district-level resource and student-level academic achievement relationships.
Copyright 2000 by the Education Policy Analysis ArchivesThe World Wide Web address for the Education Policy Analysis Archives is epaa.asu.edu General questions about appropriateness of topics or particular articles may be addressed to the Editor, Gene V Glass, glass@asu.edu or reach him at College of Education, Arizona State University, Tempe, AZ 85287-0211. (602-965-9644). The Commentary Editor is Casey D. Cobb: casey.cobb@unh.edu . EPAA Editorial Board
|