Skip to main content
Log in

Migration background and educational tracking

Is there a double disadvantage for second-generation immigrants?

  • Original Paper
  • Published:
Journal of Population Economics Aims and scope Submit manuscript

Abstract

Research on immigrants’ educational disadvantages documents substantial immigrant–native achievement gaps in standardized student assessments. Exploiting data from the German PIRLS extension, we find that second-generation immigrants also receive worse grades and teacher recommendations for secondary school tracks than natives, which cannot be explained by differences in student achievement tests and general intelligence. Second-generation immigrants’ less favorable socioeconomic background largely accounts for this additional disadvantage, suggesting that immigrants are disproportionately affected by prevailing social inequalities at the transition to secondary school. We additionally show that differences in track attendance account for a substantial part of the immigrant–native wage gap in Germany.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Institutional subscriptions

Fig. 1
Fig. 2
Fig. 3

Similar content being viewed by others

Notes

  1. In Austria, students are also tracked at age 10; in the Czech Republic, Hungary, and Slovenia, at age 11; and in Belgium and the Netherlands, at age 13 (for a more comprehensive review, see Woessmann 2009).

  2. See Empfehlungen zur Arbeit in der Grundschule (Beschluss der Kultusministerkonferenz vom 02.07.1970 i.d.F. vom 06.05.1994), as cited in KMK (2010).

  3. p. 436 Pietsch and Stubbe (2007) find that 83.4 % of the parents follow the teacher’s recommendation, while 6.7 % attend a lower secondary school, and 9.9 % a higher secondary school than recommended by the teacher.

  4. The figure displays the distribution of reading performance for exemplary purposes. Similar overlaps exist when focusing on math performance or on a combined measure of test scores in the two domains.

  5. This illustration is helpful to clarify the idea of the double disadvantage; but in reality, the transition to secondary school tracks in Germany, in addition to having more than just two school tracks, is not a deterministic process as suggested by Fig. 3. In particular, it should be kept in mind that there are no objective, clear cutoff rules for receiving a recommendation for a particular type of school since teachers do not base their recommendation on objective tests but on subjective assessments of their students’ educational potential.

  6. See also the discussion in p. 32 Schnepf (2002).

  7. For details on the construction of the estimation sample, on the treatment of missing values, on the measures of cognitive skills, as well as descriptive statistics of students’ background characteristics, see Appendix A.1.

  8. Results are almost identical when estimating ordered logit, ordered probit, and linear probability models.

  9. More precisely, we calculate the average of discrete or partial changes over all observations, using the finite-difference method for categorical variables and the calculus method for continuous variables.

  10. Note that returns to socioeconomic background characteristics do not differ significantly between second-generation immigrants and natives.

  11. In a linear regression, the students’ grades in German and mathematics account for 70% of the variation in teacher recommendations. Note that the second-generation immigrant dummy does not enter significantly in this regression.

  12. In a related study, Kiss (2011) confirms the existence of grade disadvantages in mathematics for second-generation immigrants in primary school based on a matching approach but cannot find evidence for differences in grading within secondary school tracks between natives and second-generation immigrants at the age of 15.

  13. We find some evidence that part of the grade disadvantage is due to unobserved heterogeneity between schools. Estimating the same specification with class fixed effects did not change the results compared to the school fixed effects specification. These results are available from the authors upon request.

  14. In sociology and psychology, it is well established that at points of transition in the educational system, the impact of students’ socioeconomic background on educational outcomes tends to be amplified (for a brief overview see Maaz et al. 2008). This finding, however, has to date received much less attention in the economic literature on educational production.

  15. See Spence (1973) for the seminal paper in the signaling literature, and Altonji and Pierret (2001) for recent evidence on the existence of employer learning and statistical discrimination.

  16. In an earlier study, Schmidt (1997) shows that the gap in earnings between native Germans and ethnic German migrants is substantially reduced when controlling for participation in advanced secondary schooling (a proxy for attending the highest school track) and participation in postsecondary education.

  17. Note that these results correspond to the results for second-generation immigrants in Germany reported in Table 4 of (Algan et al. 2010) apart from the fact that we do not further distinguish between different countries of origin of second-generation immigrants. We thank Albrecht Glitz for providing us with the relevant programming code to replicate their results. See Appendix A.2 for details on the estimations presented and Table 8 for descriptive statistics on the estimation sample.

  18. Note that there are no systematic differences in the amount of missing information on teacher recommendations between natives and second-generation immigrants.

  19. First-generation immigrants, defined as individuals born outside of Germany who have either only foreign citizenship or who obtained German citizenship through naturalization, are excluded from the sample.

  20. In principle, one would also have to subtract the log of weeks per month, but this is a constant and will be captured in the constant term in the regression analysis.

References

  • Algan Y, Dustmann C, Glitz A, Manning A (2010) The economic situation of first and second-generation immigrants in France, Germany and the United Kingdom. Econ J 120(542):F4–F30

    Article  Google Scholar 

  • Altonji JG, Pierret CR (2001) Employer learning and statistical discrimination. Q J Econ 116(1):313–350

    Article  Google Scholar 

  • Ammermueller A (2007) Poor background or low returns? Why immigrant students in Germany perform so poorly in the programme for international student assessment. Educ Econ 15(2):215–230

    Article  Google Scholar 

  • Autorengruppe Bildungsberichterstattung (2008) Bildung in Deutschland 2008. Ein indikatorengestützter Bericht mit einer Analyse zu Übergängen im Anschluss an den Sekundarbereich I. Bertelsmann, Bielefeld

  • Autorengruppe Bildungsberichterstattung (2010) Bildung in Deutschland 2010. Ein indikatorengestützter Bericht mit einer Analyse zu Perspektiven des Bildungswesens im demografischen Wandel. Bertelsmann, Bielefeld

  • Bauer P, Riphahn RT (2006) Timing of school tracking as a determinant of intergenerational transmission of education. Econ Lett 91(1):90–97

    Article  Google Scholar 

  • Becker M, Lüdtke O, Trautwein U, Baumert J (2006) Leistungszuwachs in Mathematik. Z Padagog Psychol 20(4):233–242

    Article  Google Scholar 

  • Brunello G, Checchi D (2007) Does school tracking affect equality of opportunity? New international evidence. Econ Policy 22(52):781–861

    Article  Google Scholar 

  • Dustmann C (2004) Parental background, secondary school track choice, and wages. Oxford Econ Pap 56(2):209–230

    Article  Google Scholar 

  • Gang IN, Zimmermann KF (2000) Is child like parent? Educational attainment and ethnic origin. J Hum Resour 35:550–569

    Article  Google Scholar 

  • Heller K, Perleth C (2000) Kognitiver Fähigkeits-Test für 4. bis 12. Klassen, Revision. (KFT 4-12+R). Beltz Test GmbH, Göttingen

    Google Scholar 

  • Kiss D (2011) Are immigrants and girls graded worse? Results of a matching approach. Educ Econ (forthcoming)

  • KMK (2010) Übergang von der Grundschule in Schulen des Sekundarbereichs I und Förderung, Beobachtung und Orientierung in den Jahrgangsstufen 5 und 6 (sog. Orientierungsstufe). http://www.kmk.org/fileadmin/veroeffentlichungen_beschluesse/1970/1970_07_02_Empfehlungen_Grundschule.pdf. Accessed 03 Nov 2010

  • Maaz K, Trautwein U, Lüdtke O, Baumert J (2008) Educational transitions and differential learning environments: how explicit between-school tracking contributes to social inequality in educational outcomes. Child Dev Perspect 2(2):99–106

    Article  Google Scholar 

  • Meghir C, Palme M (2005) Educational reform, ability, and family background. Am Econ Rev 95(1):414–423

    Article  Google Scholar 

  • Pekkarinen T, Uusitalo R, Kerr S (2009a) School tracking and development of cognitive skills. IZA discussion papers 4058. Institute for the study of labor (IZA), Germany

  • Pekkarinen T, Uusitalo R, Kerr S (2009b) School tracking and intergenerational income mobility: evidence from the finnish comprehensive school reform. J Public Econ 93(7–8):965–973

    Article  Google Scholar 

  • Pietsch M, Stubbe TC (2007) Inequality in the transition from primary to secondary school: school choices and educational disparities in Germany. Eur Educ Res J 6(4):424–445

    Article  Google Scholar 

  • Riphahn RT (2003) Cohort effects in the educational attainment of second generation immigrants in Germany: an analysis of census data. J Popul Econ 16(4):711–737

    Article  Google Scholar 

  • Schmidt CM (1997) Immigrant performance in Germany: labor earnings of ethnic German migrants and foreign guest-workers. Q Rev Econ Financ 37:379–97

    Article  Google Scholar 

  • Schneeweis N (2011) Educational institutions and the integration of migrants. J Popul Econ 24(4):1281–1308

    Article  Google Scholar 

  • Schnepf SV (2002) A sorting hat that fails? The transition from primary to secondary school in Germany. Innocenti working papers 92, UNICEF

  • Schnepf SV (2007) Immigrants’ educational disadvantage: an examination across ten countries and three surveys. J Popul Econ 20(3):527–545

    Article  Google Scholar 

  • Spence AM (1973) Job market signaling. Q J Econ 87(3):355–374

    Article  Google Scholar 

  • Thorndike R, Hagen E (1971) Cognitive abilities test. Houghton-Mifflin, Boston

    Google Scholar 

  • Woessmann L (2009) International evidence on school tracking: a review. CESifo DICE Rep 7(1):26–34

    Google Scholar 

Download references

Acknowledgements

We thank the three anonymous referees for insightful comments. We would also like to thank the participants of the ESPE conference in Essen, the EALE/SOLE conference in London, the annual congress of the IIPF in Uppsala, the annual meeting of the German Economic Association in Kiel, the EEA conference in Oslo, and seminars in Barcelona, Munich, and Tübingen. We especially like to thank Hanna Dumont, Oliver Falck, Tiago Freire, Albrecht Glitz, Rainer Lehmann, Stephan Thomsen, Ludger Woessmann, and Lei Zhang.

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Guido Schwerdt.

Additional information

Responsible editor: Klaus F. Zimmermann

Appendices

Appendix A: Description of data sets

1.1 A.1 The German extension of the Progress in International Reading Literacy Study (PIRLS-E) 2001

PIRLS is an internationally comparable assessment of reading literacy of primary school students. As in most countries, in Germany, 10-year old students were tested, all of which attended the fourth grade of primary school. The PIRLS-E 2001 database is unique in that it contains a wide range of objective and subjective measures of cognitive skills.

We use three types of cognitive measures in our analyses of the double disadvantage. First, we use a measure of students’ mathematics performance provided by the German extension of the PIRLS-E. Second, we use the test scores on two subscales of a standardized test of cognitive abilities, the Kognitive Fähigkeitstest for grade 4 by Heller and Perleth (2000): Verbal Analogies and Figure Analogies. This is the German adaptation of the “Cognitive Abilities Test” by Thorndike and Hagen (1971). A total response time of 7 or 8 min was devoted to these subtests. Both subscales measure an individual’s capacity for logical thinking and reasoning. Generally, a high share of total variance in the scores of the KFT subscales is accounted for by a factor termed “general intelligence,” with the highest factor loadings on the figure analogies subscale. Heller and Perleth (2000) point out that, on average, students with migration background show stronger differences in performance on the different subscales than native students, which is why we use the scores on the two subscales separately in all our analyses. Also, note that the authors warn against the interpretation of KFT results as indicating an invariant indicator of intelligence. An individual’s KFT test score is to be interpreted not as a measure of innate, invariant cognitive ability, but it is to be conceived also as an outcome of formal education, indicating an individual’s cognitive strengths and weaknesses, as well as potential need for remedial education.

Third, we analyze subjective measures of student achievement, namely grades in German and mathematics as well as teacher recommendations for the type of secondary school a child should attend at the end of grade 4. Both grades and recommendations are provided by the teachers. PIRLS-E also contains detailed information on students’ individual characteristics and parental background. Given the relatively large number of missing values for all measures of social background, we impute household income, parental education, and number of books at home, as well as kindergarten attendance and language spoken at home, using the method of multiple imputation by chained equations (MICE). This imputation approach gives valid inferences under the assumption that data are missing at random. We set the number of imputations M = 25 to keep the sampling error due to imputation relatively low.

Table 7 contains descriptive statistics on students’ background characteristics, and reveals that second-generation immigrants, on average, come from less privileged social backgrounds and have attended kindergarten for a shorter period of time. For our analyses of the second disadvantage for second-generation immigrants, we use data for West German states only since for historical reasons, there are very few second-generation immigrants in East Germany. Given that primary school has six grades in Berlin and Bremen, students’ families do not have to make a decision about which academic track to choose at the end of grade 4. We therefore drop observations from these two states.

Table 7 Descriptive statistics on students’ background characteristics

Additionally, Hamburg and Saarland have been excluded because there is no differentiation between lower and intermediate secondary school in grades 5 and 6. There remains a sample of N = 5,071 observations from seven West German states in the sample: Baden-Württemberg, Bavaria, Hesse, Lower Saxony, North-Rhine Westphalia, Rhineland-Palatinate, and Schleswig-Holstein. We further had to delete 1,165 observations (23.0%) because either the information on the teacher recommendation (N = 415) or on migration background (N = 827) was missing or both.Footnote 18 Moreover, we excluded from the sample all first-generation immigrants, i.e., all students who were not born in Germany (N = 519). Our estimation sample consists of 580 second- generation immigrants and 2,856 native students. In all regression models that contain mathematics performance, we also dropped all students from Lower Saxony since they did not participate in the mathematics test.

1.2 A.2 The German microcensus

The German Microcensus is the largest scale annually conducted household survey in Germany with a sample of 1 % of the German population. The statistical office provides public use files with information on 70 % random samples of the Microcensus data which contain up to half a million observations. We use Microcensus data for the years 2005 and 2006.

These data allow identification of second-generation immigrants based on citizenship and year of arrival in Germany. The reference native group consists of nonnaturalized German citizens born in Germany. We identify second-generation immigrants as individuals born in Germany who either hold only foreign citizenship or German citizenship that they obtained through naturalization.Footnote 19 This identification of second-generation immigrants as well as other sample restrictions correspond to the sample construction used in Algan et al. (2010). The data provides information on employment status, normal working hours per week, and net monthly earnings. We construct an approximate log hourly wage measure by subtracting the log of normal hours worked from the log of net monthly earnings.Footnote 20 Most importantly, the data also contains information on the type of secondary school certificate. We use secondary school certificates to proxy for track attendance as had been done previously in the literature (see Dustmann 2004). See Table 8 for descriptive statistics by gender and migration background.

Table 8 Characteristics of the employed population by migration background

Appendix B: Robustness check: missing information on migration background

One potential concern with our empirical analysis is that we have to exclude a relatively large number of observations due to missing information on migration background. Thus, our results might identify effects for a specific subset of second-generation immigrants. In order to assess the potential relevance of the group of students omitted due to missing information on migration background, we conduct three robustness checks, the results of which are shown in Table 9.

Table 9 Robustness check: missing information on migration background

The table shows average marginal effects after multinomial logit for our main specifications of interest (without controls for socioeconomic background). These results should be compared to the average marginal effects for second-generation immigrants reported in columns 1–3 of Tables 3 and 4.

First, we include all students with missing information on migration background as a separate category in our multinomial logit model (columns 1–3). Second, we code all students with missing information on migration background as second-generation immigrants (columns 4–6). Third, we code all students with missing information on migration background as natives (columns 7–9).

All three robustness checks fully support our main results. The results from estimations including students with missing migration background as a separate category show that students with missing information on migration background receive on average less favorable teacher recommendations for secondary school tracks than natives (column 1). However, when additionally controlling for differences in achievement (column 2) and controlling for differences in achievement and general intelligence (column 3), we no longer observe a statistically significant difference in teacher recommendations for this group.

This reflects the common finding that, in international student assessments, nonresponse in the background questionnaires is more common among students with low ability. Thus, students with missing information on migration background also receive on average less favorable teacher recommendations for secondary school tracks than natives, but this difference in teacher recommendations is almost entirely explained by differences in student achievement and general intelligence.

Although marginal effects for students with missing information on migration background in columns 2 and 3 are not significant, the signs of the estimates may indicate that students in this group are slightly less likely to receive a recommendation for high school and slightly more likely to receive a recommendation for secondary general school than natives. This is consistent with the hypothesis that this group of students, similar to the group of students with observed information on migration background, consists mainly of native students, but also includes a small group of second-generation immigrants.

Columns 4–6 (Tables 7, 8, and 9) show how our main results change if we code all students with missing information on migration background as second-generation immigrants (natives). Given the results presented in columns 1–3, we expect that the estimated average marginal effects for second-generation immigrants will become smaller in absolute terms. Overall, this is indeed what we see. For male students (panels (a) and (b) of Table 9), the estimated differences in teacher recommendations between natives and second-generation immigrants after controlling for cognitive skills are smaller in absolute values in comparison to the estimates presented in columns 1–3 in Tables 3 and 4. However, all estimates remain statistically significantly different from zero except for the estimated marginal effect for the outcome high school in panel (b) column 3. For female second-generation immigrants, the point estimates are smaller and partly insignificant, as shown in panels (c) and (d) of Table 9. This generally confirms our main findings reported in columns 1–3 in Tables 3 and 4.

Rights and permissions

Reprints and permissions

About this article

Cite this article

Lüdemann, E., Schwerdt, G. Migration background and educational tracking. J Popul Econ 26, 455–481 (2013). https://doi.org/10.1007/s00148-012-0414-z

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s00148-012-0414-z

Keywords

JEL Classification

Navigation