Primary and secondary effects in the explanation of disadvantage in education: the children of immigrant families in France

Abstract

This paper explores the prospective transition of immigrant and native students in France from lower to upper secondary school. Because they are more likely to be tracked to less prestigious (vocational) tracks, immigrant and immigrant‐origin students are significantly disadvantaged at this key academic stage in comparison with the children of native families. Primary and secondary sources of educational disadvantage are explored to explain this phenomenon. Primary effects appear to account for the entire initial disadvantage, while secondary effects could have a positive impact for immigrant‐origin students. Nonetheless, immigrant families appear to be more conservative than native families and may need stronger evidence that their children will succeed in upper secondary school.

Keywords:

• educational disadvantage
• immigration
• France
• secondary school

Notes

1. A law passed in 1985 (Loi Informatique, fichiers et libertés) prevented public surveys from asking about country of birth in an attempt to fight discrimination and racism. In 1992 a survey called Mobilité Géographique et Insertion Sociale (Geographical Mobility and Social Insertion), conducted by the French demographer Michèle Tribalat and her colleagues Benoît Riandey and Patrick Simon, allowed the identification of migrants by nationality. This created an intense public and academic debate instigated by the demographer Hervé le Bras, who criticised Tribalat for rendering evidence about the ethnic stratification of French citizens in the context of increasing xenophobia and the growing popularity of the ultra‐nationalist party Front National.

2. In Panel‐95, 62.13% of students went to séconde général et technologique. Some 30.20% went on to BEP and only 5.47% went to CAP.

3. Once we controlled for the grades in 1995 (Model 2), the coefficients of the non‐mixed categories of immigrant children were be positive and remained statistically significant.

4. Note that in this model only the indicators of prior performance significantly affect the likelihood of proceeding towards one track or the other. It is important to emphasise that these two variables changed little.

5. The assumption behind these models is that the loss of cases in the dependent variable results in sample selection. This is a form of censoring in which the truncation of a dependent variable depends upon the value of another variable. In this case, the independent variables (X) and the selection variables (S) must be observed for all of the sample, regardless of whether our dependent variable is observed. In this context, estimates obtained using ordinary least squares techniques will be both biased and inconsistent, but Heckman two‐stage modelling can capture the selection and produce consistent estimates.

6. This questionnaire was only collected for the students in 3^ème. Logically, if the student repeated, the questionnaire was only filled when he reached 3^ème.