You are on page 1of 31

Are immigrant students better off in certain educational systems or certain schools?

Jaap Dronkers, Rolf van der Velden & Allison Dunne


Maastricht University

---------------------------------------------------------------------------------------------------------ECER 2011 Urban Education, Freie Universitt Berlin

JnkJaap ers

Background of the paper


Effect of parental background on achievement varies systematically in different educational systems. Higher average scores in comprehensive systems. Only two levels: educational systems (i.e. countries) & students (e.g. Hanushek & Wmann (2005). Dunne (2010) and Dronkers (2010) introduced a 3-level model: 1. educational systems; 2. schools and 3. students. Conclusion: Characteristics of the schools (composition; diversity) has differential effects on achievement and serve as intermediaries between educational systems and students.

The problem with Dunne and Dronkers


No track as separate level of analysis: confusion of school characteristics (like school composition) and curriculum taught in schools. Identification of schools by administrative unit: invalid measurement of real school context. Schools can have different tracks that constitute a totally different environment to students. Omitting the track as a separate level can serious seriously bias the results.

Contribution of this analysis


Inclusion of track as separate level of analysis. Use of track in stead of administrative unit of school to estimate composition characteristics. Separate analyses of native and immigrants children: different effects of schools and educational systems. In this paper we focus on immigrants.

Research question
What are the effects of educational systems, school-composition and track characteristics on educational achievement of 15-years old students with immigrant background? What are the effects of socio-economic and ethnic characteristics?

Hypotheses
1. Track-level explains most of the effect of socioeconomic (ESCS) school-composition, especially in stratified systems. 2. No difference in ESCS school-composition effects in different systems, after control for track. 3. Strongest effect of individual ESCS in comprehensive system, after control for track and ESCS schoolcomposition. 4. Students with immigrant background have the highest achievement in strongly stratified systems or in comprehensive educational systems.

Data
OECD PISA wave 2006: tests of 15-year-olds in linguistics, math and science. Questionnaires for students and school principals. 57 countries participated, but only 15 collected data on country of origin of immigrants: AU, AT, BE, CH, DK, FI, DE, GR, LT, LI, LU, NZ, NO, PT, Scotland. Earlier research showed the importance of including country of origin (Levels, Dronkers & Kraaykamp, 2008). This analysis: 15 countries of destination, 72.300 natives + 8.600 immigrants from 35 countries of origin.

Cross-classified analysis with four levels


1. Country of destination (educational system characteristics) 2. Country of origin 3. Tracks-within-schools (track + school characteristics) 4. Students (students + family characteristics)

Identification of tracks (1)


PISA identifies separate schools. But the school-level reflects administrative unit of educational institution. The track-within-schoollevel reflects daily reality of teaching and learning environment. 2 cross-national indicators of tracks: A. Lower versus Higher; B. General versus Vocational. However, this is done by national coordinators and is not internationally comparable nor sufficiently detailed.

Identification of tracks (2)


E.g. Finland places all students in lower general, while Scotland places all students in higher general. In some cases we were able to repair this: e.g. in Germany we were able to distinguish between general education with and without access to higher secondary education (i.e. higher and lower general tracks) and between lower and higher vocational tracks. But this could not be done in all cases, which will lead to some noise in identification of tracks (and thus underestimation of the effects).

Dependent variable
Linguistic performance Repeated the analyses with performance in math and science: -> same results This measures more scholastic ability

Individual level characteristics


Parental ESCS: composite index of occupational status, educational level and material/cultural resources at home. Grade. standardized around modal grade. Female. Regions of origin: 1. Eastern-Europe; 2. non-Islamic Asia; 3. Islamic countries; 4. Western OECD countries; 5. SubSahara Africa. Second generation migrant. One parent migrant, other parent native. Home language same as in destination country.

Tracks-within-schools characteristics (1)


Ethnic diversity: 0 = all students are from the same country of origin; 1 = students are equally from all origin countries. ESCS diversity: 0 = all parents are in the same ESCS category; 1 = students are equally from all ESCS categories. Percent students from immigrant regions: % Eastern Europe; % non-Islamic Asia; % Islamic countries; % Western OECD countries; % Sub-Sahara Africa. Average ESCS. Average parental ESCS per trackwithin-school

Tracks-within-schools characteristics (2)


Selective admittance of students to the school: selective; some selection (ref.); no selection. Teacher shortage (standardised per country). Student-staff ratio (standardised per country). Urbanisation: city; small town (ref.); rural. School size. Private public: public dependent (ref.); private dependent; private independent.

Country level characteristics


Stratification of educational system:
Highly stratified: AT, CH, DE, LI Moderately stratified: BE, GR, LU, PT Not stratified: AU, DK, FI, LT, NO, NZ, Scotland

Age of selection: 10-16

Strongest effects of individual and school-variables in different systems

Different importance of tracks in different educational systems


Differences of fit of models support first hypothesis: Track explains a substantial part of school-composition effect. School-composition explains more variance than track in all educational systems. No effect of higher track in comprehensive systems (which is logical). Positive effects of higher track in moderately and highly stratified systems: -1, +32 and +16. Negative effect of vocational track in all educational systems: -54.

Different importance of ESCS schoolcomposition in different educational systems


Model without tracks shows that ESCS schoolcomposition has stronger effects in the moderately and highly stratified systems than in the comprehensive systems. This is because track indicates both curriculum and neighborhood effects. However, after inclusion of track the effect of ESCS school-composition is strongest in comprehensive and highly stratified systems: 1 st.dev. increase -> +24, +18 and +32.

Different importance of individual ESCS in different systems


Direct effects of parental ECSC on reading scores strongest in comprehensive: 1 s.d. increase -> + 25, +9 and +9. But bivariate correlations between parental ECSC and reading scores are equal in these systems. Influence of parental ECSC on selection of immigrant students into different tracks and schools is equal in all systems (deviant for native students).

Ethnic composition effects

Ethnic share in schoolcomposition


Only positive effects of % Eastern-Europe and % non-Islamic Asia pupils No significant effects of the % Islamic, Western-OECD, and Sub-Saharan Africa pupils.

Negative effect of ethnic school-diversity


Ethnic school-diversity has a negative effect on achievement scores in all educational systems, but only for immigrant students. This result confirms partly the earlier outcomes by Dronkers (2010). No significant effect of ESCS diversity.

Effects of origin country (1)

Effects of origin country (2)


Same home language (+24) nullifies immigrants-native difference (-26). Second generation (+9) decreases 1/3 of immigrants-native difference (-26). Non-Islamic Asian origin (+19) decreases 2/3 of immigrants-native difference (-26). Islamic country origin (-35) doubles immigrants-native difference (-26).

Scores pupils with high and low parental ECSC in different schools and systems

No uniformly best or bad system (1)


Best scores of pupils with highest ESCS parents: Comprehensive if school ESCS=high & high general: 625 Comprehensive if school ESCS=low & high general: 466 Comprehensive if school ESCS=high & vocational: 572 Comprehensive if school ESCS=low & vocational: 404

No uniformly best or bad system (2)


Best scores of pupils with lowest ESCS parents: Stratified if school ESCS=high & high general: 551 Moderately if school ESCS=low & high general 365 Stratified if school ESCS=high & vocational: 481 Moderately if school ESCS=low & vocational: 278

Analysis with 4 instead of 2 levels

Results with 4 instead of 2 levels


A two-level model: overestimates the effects of parental ESCS and grade; underestimates effects of educational systems; overestimates amount of variance at individual level; fits less with data.

Conclusions
Results highlight the importance of including track and school factors and countries of origin. Inclusion of the track-level is necessary to avoid overestimation of the school-composition effect, especially in stratified systems. Effects of origins of immigrant children and ethnic school diversity should not be ignored. Educational systems are not uniformly good or bad, but they have different consequences for different groups.

More information
Full paper: Are immigrant students better off in certain types of educational systems or schools than in others? On the effects of educational systems, school-composition, track-level, parental background and country of origin on the achievement of 15-year- old immigrant students. Available at: http://www.eui.eu/Personal/Dronkers/English/ECE R.pdf All correspondence to: j.dronkers@maastrichtuniversity.nl

You might also like