Effectiveness of Four Instructional Programs Designed to Serve English Learners

English-learner students in a large district followed from elementary through middle school

English-learner students in English Immersion, Transitional Bilingual, Developmental Bilingual and Dual Immersion programmes

The identification strategy is the interesting part: the authors compare students whose PARENTS EXPRESSED THE SAME PROGRAMME PREFERENCE but who ended up in different programmes, which absorbs a large share of the usual family-selection confound without being a lottery. It remains observational, and the residual threat is whatever determined which preferring family got which programme. Outcomes are English Language Arts and mathematics test scores in ENGLISH, tracked as growth trajectories rather than levels.

English Language Arts scores of English learners in ALL bilingual programme types grow at least as fast as, and sometimes faster than, those in English Immersion. The same generally holds in mathematics, with the exception of Developmental Bilingual programmes, where average scores grow more slowly than in English Immersion. Latino English learners do better longitudinally in bilingual programmes than their Chinese counterparts. No differences in programme effectiveness by initial English proficiency.

Medium. Matching on expressed parental preference removes the largest and most obvious selection channel, which is why this sits above the rest of the observational literature; but programme assignment among equally-preferring families is still not random.

Converges with the Portland lottery on the key practical point - bilingual instruction does not cost English achievement - using an entirely different identification strategy on a different population. That convergence is the strongest thing this cluster has.