Ukrainian refugees struggling to integrate into Czech school social networks

The study examines how Ukrainian refugee students, who rapidly entered Czech schools after the 2022 Russian invasion, integrate into classroom peer networks. The research questions are: (1) To what extent are Ukrainian refugee students integrated into classroom social networks, considering both friendship and exclusion ties? (2) How does classroom ethnic composition (proportion of Ukrainian students) relate to their level of social integration? The context includes an unprecedented influx of Ukrainian children into an educational system in an historically ethnically homogeneous society, with many newcomers lacking Czech language proficiency and experiencing psychosocial stress. Given the critical importance of peer relationships for psychosocial adjustment, mental health, and academic outcomes, mapping refugees’ peer integration is both timely and policy-relevant. The study also responds to optimistic institutional reports suggesting seamless integration, which the authors sought to verify using network methods capturing students’ own perspectives.

Prior research shows that ethnic minority and first-generation students often experience exclusion and form homophilous (same-ethnicity) friendships. Language barriers and war-related psychosocial stress further hinder refugees’ peer relationship formation. Evidence on how classroom or school ethnic composition shapes cross-ethnic friendships is mixed: some studies suggest diversity increases inter-ethnic contact and acceptance, while others find that higher same-ethnic presence increases intra-ethnic clustering and can relate to worse social adjustment. Few studies have focused specifically on refugees, and even fewer have examined classroom composition effects with network-analytic modeling that accounts for network structure. This study adds early evidence specific to Ukrainian refugees and explicitly models mechanisms of tie formation.

Design and setting: Cross-sectional sociometric network study in 12 mixed (Czech–Ukrainian) lower secondary classrooms across six schools in Brno, Czech Republic. Data were collected October–December 2022, shortly after refugees’ arrival but allowing some time for relationships to form. Sample: N=266 students in grades 5–9 (ages 11–15); 56 Ukrainian (21.05%) and 210 Czech (78.95%). Slight majority of boys. Two classrooms per participating school were randomly selected from schools with high refugee enrollment (30–60 Ukrainian students). Data collection: Pen-and-paper bilingual (Czech/Ukrainian) sociometric questionnaire administered in class by trained researchers. Two unlimited nomination items captured directed ties: (1) friendship (“Write the names of the classmates you are friends with.”), (2) exclusion (“Write the names of the classmates with whom you would not like to share a desk.”). Gender and ethnicity were obtained from school records. No missing data were reported. Network construction: For each classroom, multiplex directed networks were constructed with nodes as students and two tie layers: friendship and exclusion. Multiplex modeling acknowledges possible cross-layer dependence (e.g., exclusion reducing likelihood of friendship). Analytic strategy: Multiplex exponential random graph models (ERGM) were estimated for each classroom in XPNet. A common model specification assessed: (a) sender/receiver propensities by ethnicity (Ukrainian vs Czech) and gender; (b) ethnic and gender homophily; (c) structural effects on friendship (reciprocity, path closure/transitivity, cyclic closure, multiple 2-paths, alternating in/out stars); (d) limited structural effects on exclusion (reciprocity); (e) cross-layer dependence between friendship and exclusion. Convergence and goodness-of-fit were checked via simulation (1000 networks per model); fits were good for friendship and modest for exclusion due to fewer structural terms. One classroom (Network 12) required a reduced specification due to near-perfect collinearity among structural terms; in several classrooms, no exclusion ties among Ukrainian students prevented estimating ethnic homophily on exclusion, so aggregate estimates for that term are tentative. Meta-analysis and meta-regression: Classroom-specific ERGM estimates were aggregated using random-effects meta-analysis (metafor in R). A mixed-effects meta-regression examined whether the percentage of Ukrainian students in a classroom moderated three friendship effects: ethnic homophily, Ukrainian sender, and Ukrainian receiver effects.

Descriptive patterns: - Both friendship and exclusion layers showed typical school-network features: density, centralization, reciprocity, and transitivity; exclusion layers were less cohesive (lower values) than friendship layers. Friendship and exclusion ties negatively correlated, indicating they rarely co-occur. - Descriptive homophily: friendship ties showed positive gender and ethnic homophily; exclusion ties tended to be mixed (negative homophily descriptively). ERGM meta-analysis (aggregate log-odds estimates; p-values): - Friendship structure: strong reciprocity (1.63, p<0.01) and path closure (1.39, p<0.01); alternating in-out star positive (0.77, p<0.01). - Friendship covariates: • Ethnic homophily: 1.83 (SE 0.24), p<0.01 – strongest covariate effect; students favored same-ethnicity friendships. • Ukrainians: lower tendency to send friendship ties (sender: -0.61, SE 0.16, p<0.01) and to receive friendship ties (receiver: -0.43, SE 0.16, p=0.01) than Czech classmates. • Gender homophily: 1.32 (SE 0.19), p<0.01. • Boys: lower sending (-0.61, p<0.01) and receiving (-0.54, p<0.01) of friendship ties. - Exclusion structure: reciprocity positive (0.68, p<0.01); intercept -1.41 (p<0.01), indicating sparser exclusion ties than friendships. - Exclusion covariates: • Ethnic homophily: 0.97 (SE 0.35), p=0.01 (aggregate estimate tentative due to estimation limits in some classrooms). • Ukrainians: no significant difference in sending (-0.42, p=0.19) or receiving (-0.26, p=0.17) exclusion ties vs Czechs. • Boys: higher likelihood to be targets of exclusion (receiver: 0.48, p=0.02); other gender effects not significant. - Cross-layer dependence: strong negative association between friendship and exclusion ties (-2.43, p<0.01), indicating these relationships rarely occur between the same dyads. Meta-regression (moderation by classroom % Ukrainian): - Ethnic homophily (friendship) increased with higher Ukrainian share: +0.07 in log-odds per 1% increase in Ukrainian students (p=0.04). - Ukrainian sender effect (friendship) decreased with higher Ukrainian share: -0.04 in log-odds per 1% (p=0.07, marginal). - No moderation for Ukrainian receiver effect (friendship) (p=0.86). Interpretation: - Networks formed heavily along ethnic lines. Ukrainian students were less connected via friendships (both as senders and receivers) but were not more explicitly excluded; they appeared neglected rather than rejected. Increasing the proportion of Ukrainian students was associated with stronger homophily and reduced Ukrainian initiation of friendships.

The findings address both research aims. First, Ukrainian refugees were not well integrated in classroom social networks: they had significantly fewer friendship ties sent and received, and friendships were strongly segregated by ethnicity. However, there was no evidence of higher exclusion directed to or from Ukrainian students, suggesting neglect rather than overt rejection. Second, classroom composition mattered: a higher percentage of Ukrainian students was linked to stronger ethnic homophily and a reduced tendency of Ukrainian students to initiate friendships. These results contrast with optimistic reports from school leadership and inspectors, likely reflecting perception biases when relying on external observations rather than students’ relational data. The results align with prior literature showing homophily and challenges for refugees’ peer integration, and they contribute by demonstrating composition effects using network models that account for structural dependencies.

This study provides early, network-based evidence on Ukrainian refugees’ peer integration in Czech classrooms. It shows that peer networks are strongly structured along ethnic lines and that Ukrainian students are less embedded in friendships without being overtly excluded. Importantly, higher classroom shares of Ukrainian students are associated with stronger homophily and lower friendship initiation by Ukrainians, indicating potential downsides of higher within-class concentration. Practically, schools and policymakers should distribute refugee students more evenly across classes (aiming to avoid exceeding ~20% per class) and implement interventions that foster cross-ethnic ties (e.g., inter-ethnic tutoring, cooperative learning, shared sports, extracurriculars) alongside intensive language support to reduce barriers. Future research should use longitudinal network designs, focus on student-reported ties, and examine how teacher practices and other exogenous factors can improve cross-ethnic relationship formation.

- Limited sample: 12 classrooms from six schools in one city; non-random selection; small sample constrained statistical approaches (e.g., precluded multilevel meta-analysis). - Measurement scope: only friendship and exclusion ties were collected; other relational types (e.g., bullying, romantic ties) were not captured. - Modeling constraints: one classroom exhibited near-perfect collinearity among structural terms, requiring a reduced model; in several classrooms, no exclusion ties among Ukrainian students prevented estimating ethnic homophily on exclusion, making aggregate estimates for that term tentative. - Cross-sectional design: cannot capture network evolution over time or causal dynamics. These limitations may affect generalizability and precision of some estimates.