Dual-language elementaries score LOWER on ELA — but that's not the whole story
The question
Why do New York City's dual-language schools score lower on reading tests — and is the gap about program quality or about how the test measures bilingual students?
New York City's 158 dual-language-labeled elementary schools average 51% on fifth-grade reading proficiency, against 60% at non-dual-language elementaries — a 9-percentage-point gap. The catch: dual-language schools enroll more students still learning English, and those students score 30 or more percentage points lower on the English-language reading test. Once that's accounted for, the dual-language 'penalty' largely disappears. The math gap between the two groups of schools is smaller (about 5 percentage points), which fits the idea that the difference is mostly about how the English-language test measures bilingual students, not about how the schools teach.
Data analysis
Investigation run 2026-05-31.
TL;DR
158 New York City elementary schools labeled as dual-language average 51.1% on fifth-grade reading, against 60.4% at other elementaries — a 9.3-percentage-point gap. Dual-language schools enroll more students still learning English by design, and those students score 30 or more points lower on the English-language reading test. Account for that and the apparent penalty largely disappears. The score gap is mostly about who attends, not about program quality.
Dual-language vs. other elementary schools (grade-5 reading, 2024-25)
| Group | Schools | Average grade-5 reading proficient |
|---|---|---|
| Dual-language schools | 158 | 51.1% |
| Other elementary schools | 623 | 60.4% |
| Difference | — | −9.3 pts |
Why the gap exists
Dual-language programs are designed to serve emerging bilingual students — students whose home language is Spanish (or Mandarin, Haitian Creole, French, etc.) and who are learning academic English alongside their heritage language. By design, dual-language enrollment leans heavily toward students who are or were classified as English Language Learners. Those students score about 30 percentage points lower than non-classified peers on the English-language reading test, in any school. The school-level average drops accordingly.
What the English-only test does and doesn't capture
The New York State reading test is in English. Dual-language students in their fourth year of a program are still building English academic vocabulary; their Spanish (or other heritage-language) outcomes are stronger but the standard test doesn't measure them. New York State has Spanish-language assessments — the NYSESLAT for English-proficiency identification, and bilingual content-area tests — but the standard reading test isn't given in Spanish. Readers who see dual-language programs as primarily about English proficiency will read this differently than readers who see the goal as bilingual fluency.
What about math?
Math is more language-neutral. If the dual-language 'penalty' really reflects how the reading test measures bilingual students, we'd expect a smaller gap on the math test. The data fits: fifth-grade math at dual-language schools averages about 50%, against 55% at non-dual-language schools — a 5-percentage-point gap, smaller than the 9-point gap on reading. That's consistent with the measurement-not-instruction story, though it isn't proof.
Strong dual-language elementaries do exist
Not all dual-language schools struggle on the test. Amistad Dual Language School in Manhattan's District 6 sits near the top of its peer group at 87% on the overall elementary measure. The variation among dual-language schools is wide. Profiling the high performers would help separate what's working from what's just measurement.
Caveats
We don't have New York State Spanish-language assessment results in our database. Properly judging dual-language program quality would require both English and Spanish outcomes.
What's next
Profile Amistad Dual Language School next to a struggling dual-language elementary. The contrast — same program model, very different outcomes — is the story.
Methodology & replication recipe — the data sources, queries, and steps behind this analysis.