Integrating Refugee Children into Schools. Guest Post by Pia Schiling

This is the 15th in this year’s series of posts by PhD students on the job market.

In 2020, 36 million children are going to school in countries that they were not born in, with a majority being educated in a language different to what is spoken at home. While many countries establish elaborate policies on how to integrate adult immigrants into their labor force, we know very little about integrating children into a school system where a different language is spoken than in their country of origin. Newly immigrated students in industrialized countries often join a system that is already defined by an achievement gap between immigrants and natives. Therefore, providing them with the best possible educational integration approach is crucial to lowering this gap from the beginning.

In my job market paper, we study the effect of separate, preparatory language learning classes on the academic success of newly immigrated refugee students in Germany. In particular, we compare standardized test scores and tracking into the academic high school as our outcomes for primary school-aged refugee students who join a preparatory class for language learning upon arrival to students directly integrated into regular classrooms exposed to the standard curriculum receiving additional language classes.

The School System and Refugee Children

The two most common approaches applied for newly arrived students are the parallel and the integrative model (as well as mixed approaches) (Massumi 2015). The parallel model separates newly immigrated children from regular classes into specific preparatory classes focusing on German language learning and preparing them slowly for a transition into regular classes. The idea is to provide a protected space for (refugee) children who are unable to understand the teaching language with a German as a second language qualified teacher and not overwhelm them in the regular class. The integrative model includes refugee children in regular classes from the start and provides them with additive language training. Even though they interact early on with their native peers, the model bears the danger that they are exposed to overly excessive demands regarding language requirements (Bruggemann 2016).

Refugee children arriving in Germany are initially allocated to a federal state based on a quota system. We use administrative data from one German federal state, where families are centrally allocated to accommodations throughout the city and school-aged children to schools by an administrative body, the school information center. While the offer of preparatory classes increased over the years, given the sudden demand for preparatory classes in 2015/16 due to the large refugee influx to Germany, not all students ended up being taught in preparatory classes.

The literature has shown the benefit of fast language acquisition for long-term social and economic integration of adult immigrants and direct spillovers to their children (for example Dustmann 2003, Foged 2022, Foged 2022a, Kanas 2022, Lochmann 2019).

For immigrated children, a school integration strategy that focuses on fast language acquisition through preceding classes could have similar effects, helping to overcome the large and persistent achievement gaps between native and immigrant students in many industrialized countries, which could be due to language barriers (for example Algan 2010, Giannelli 2016, OECD 2018, Ispholding 2016).

Identification

Our outcomes of interest are standardized test scores in grade five in German, Math, English, and Natural Science, as well as the observed school track since, in Germany, students are changing schools from elementary to secondary school after fourth grade.

Our main identification assumption relies on the random allocation of refugee children to neighborhoods and schools. While from the perspective of refugee families, the school they are allocated to and whether it offers a preparatory class is random and unrelated to student characteristics, we make sure that refugee accommodations are not allocated in worse neighborhoods, and that students are not sorted into preparatory classes by observable characteristics or to better or worse schools. To account for students in regular classrooms even though they attend a school that offered a preparatory class in the year of their arrival, we exclude those children in a robustness check, as we cannot observe students' exact arrival time and whether the preparatory classes were full, or if the children were not attending them based on their ability.

To further address the threat to our identification that there is sorting based on unobservable characteristics on the side of the school information center, we make use of older children (grade three and up) in earlier immigration years being more likely to attend a preparatory class, while for later schoolyears (and immigration years) more preparatory classes were initiated and also younger children attended. We apply an instrumental variable approach using the rollout by interacting the birthdate with the immigration year as an instrument for participating in a preparatory class. We hold the first visited grade and the region of birth constant and can assume that students who came to Germany from the same areas and started school in the same grade in different years are not systematically different from each other. Hence, the year and birthdate should only affect students' test results through being more or less likely to attend a parallel class (To account for students' test results being affected by their age in relation to other children in the cohort, we also test this instrument using the cohort or only birthyear interacted with immigration year and get very similar results). As a further test, we use the rollout of preparatory classes and compare students visiting the same school at different times when it offered a preparatory class or not.

Main Results

·       Averaging standardized test scores in German, Math, English, and Science, our results show that refugees who have attended a preparatory class in elementary school achieve significantly worse outcomes in their standardized test scores in fifth grade. Students who attend a preparatory class upon arrival perform 0.19 standard deviation points lower on average across the four tests than students who directly start in the regular classroom.

o   Using our IV strategy, our results stay similar, with both the coefficient and standard error increasing (-0.35 (0.12)), with an F-Statistic of 48.1 and a p-value of 0.342 on our overidentification test.

o   Using school fixed effects and the rollout of preparatory classes also leaves our result unchanged (-0.17 (0.08))

·       Figure 1 shows a significantly negative association between attending a preparatory class and the standardized test scores for all subjects, with students achieving the lowest (but not significantly different) test scores in Math and German. The lower test result in German could come from students not being surrounded by German peers for up to one year upon arrival, with fewer possibilities in their daily school interactions to pick up the language naturally. Fenoll (2018) has shown that the Math result of students are not dependent on their language skills, therefore cutting down curricular hours in Math for German lessons might harm the students in classes that are not very dependent on speaking the language.

·       Looking at our second outcome, refugee children that participated in a preparatory class in primary school are less likely to attend an academic school as the secondary school track. While only 20% of all refugee students attend an academic high school track (compared to 50% of German students), we only find a small negative effect that is sensitive to robustness checks and needs to be interpreted with caution.

·       Being placed in a school located in a neighborhood with an above-median foreign share harms refugee students in the regular classroom, while it does not make a difference for children in the preparatory class as we find a slightly positive coefficient on our interaction term.

Possible Mechanism

·       While we cannot observe social ties in our data, we see that children who visit a parallel preparatory class are in a classroom with more children from their initial class after changing to secondary school compared to children who were integrated directly into a regular class. This suggests that students might interact more with other immigrant students and speak less German with native speakers.

Policy Implications and Limitations

With the data available, we are unable to test for the intended psychological advantages of offering newly immigrated children a safe environment where they can learn the language first. Relying on findings by Alan et al. (2021), we can assume that in a classroom with pro-social behavior of native students, refugees could benefit from a safe space and natural language exposition. Refugee children’s language acquisition profits from being exposed to a larger presence of students who are native in the host language (Boucher et al., 2021).

Regarding their academic achievement, newly arrived refugee students attending a preparatory class perform significantly worse on standardized tests. Our results show that - different to adults - newly immigrated elementary school children do not seem to universally benefit from an integration program that first focuses solely on language learning. Given their young age and ability to learn quickly, they might profit from picking up the language from their peers in a natural environment better than attending language classes.

Pia Schilling is a PhD Candidate at the Free University of Bolzano. Her personal webpage can be found here.