District analysis: Dual language immersion yields similar growth to non‑DLI classrooms; home‑boundary availability strongly predicts enrollment

Provo City School District staff presented an 11‑year longitudinal analysis of dual language immersion (DLI) programs to the school board on May 13 and told members the data do not show a clear academic-growth advantage for DLI students compared with peers in non‑DLI classrooms after controlling for observable covariates.

Director of Assessments Nate Mitchell said the analysis used hierarchical linear modeling to control for socioeconomic status, multilingual learner (ML) status, special education status and other covariates. Using Acadience (DIBELS-type) subtest scaled scores from grades 3–6 as the longitudinal outcome, Mitchell said the district’s models showed no statistically significant growth advantage for DLI students overall. "Comparing DLI programs to non‑DLI programs, the growth rates are exactly the same," he said.

Mitchell highlighted one consistent predictor of DLI participation: whether the program is available at a student’s home-boundary school. He reported an R‑squared of 0.76 for the predictor "program offered at home boundary," meaning the bulk of enrollment variance was explained by local availability. Board members asked whether moving DLI to magnet sites could reduce participation; Mitchell said the data suggest that magneting programs away from neighborhood schools would likely reduce local uptake.

On class-size and program composition, district data presented by Mitchell showed a district-average elementary class size of 22.8 students. Average class size at DLI schools (including both DLI and non‑DLI classrooms) was 23.0; DLI classrooms within DLI schools averaged 24.2 students while non‑DLI classrooms within the same buildings averaged 21.9.

On persistence and program completion, Mitchell said that most students who attrit from DLI leave in early grades (first through third), while students who persist past third grade are likely to remain through secondary grades. Across the longitudinal cohort he examined, 39 percent of students who were ever enrolled in DLI were also enrolled in the district’s higher‑level "Bridge" courses in the target language at some point; Bridge courses are 3,000‑level concurrent‑enrollment courses (the district said those who complete the typical Bridge sequence approach a university minor in the language).

Mitchell also compared two Spanish immersion implementation models in Provo: a "state model" strand (Canyon Crest) and a district model (Timpanogos). He reported no statistically significant difference in growth on the Acadience-derived metric between the two schools after adjustment, and said ML students in the district model showed stronger outcomes than ML students in the state-model school, though the speaker cautioned the ML subpopulation sizes differ between sites.

Board members asked for additional analyses and data the team agreed to provide as the board continues the discussion: breakdowns showing how many enrolled students live in Provo versus outside the district, AP exam pass rates for students attempting the language AP (a gatekeeper for Bridge classes), longitudinal boundary data to examine whether boundary changes affected enrollment, and counts of IEPs/behavioral referrals in DLI versus non‑DLI classrooms to clarify workload differences between teachers.

The board did not take any immediate policy action; the presentation was framed as analytic input for an ongoing district discussion about program use, resource allocation and boundary planning.