Hispanic Student Decline and English-Learner Mediation

What is this question really asking?

From 2022 to 2024, Hispanic students' math scores dropped 3.4 points nationally and 4.1 points in Texas — more than any other racial or ethnic group. Many Hispanic students are also classified as English learners. The question is: did the decline happen because of something about being Hispanic, or because of something about learning English during disrupted schooling?

Why classification matters. This is a mediation analysis — we are testing whether English-learner status is the pathway (the mediator) through which Hispanic identity connects to the score decline. If you turn off the mediator — by comparing only non-EL Hispanic students to non-EL White students — and the gap shrinks or disappears, language status is the mechanism.

How you would investigate it

1. Split Hispanic students into two groups. Group A: Hispanic students who are classified as English learners. Group B: Hispanic students who are not. If Group A's scores dropped much more than Group B's, language interruption during remote school is a strong candidate explanation.

2. Rule out poverty as the real driver. Hispanic students are more likely to qualify for free or reduced-price lunch (a standard proxy for low family income). Compare Hispanic and White students at the same lunch-status level. If the gap shrinks dramatically, income — not ethnicity or language — may be the underlying factor. This is called testing for a confound.

3. Check if the pattern is the same everywhere. In Texas, over half of students are Hispanic. In Vermont, fewer than 5% are. If Hispanic students declined similarly in both states, the mechanism is not about being concentrated in particular schools or districts.

4. Run a formal mediation model. A mediation analysis quantifies how much of the Hispanic decline "flows through" the EL pathway. The classic approach (Baron & Kenny, 1986) runs three regressions. Modern alternatives use bootstrapping (Preacher & Hayes, 2004). If EL status explains 70% of the decline, language is the main story. If it explains 10%, broader forces — school funding, segregation, teacher quality — are at work.

Data you would need

What	Source	Example
Scores by race × English-learner status	NAEP Data API (crosstab)	Hispanic EL: −5.2 pts; Hispanic non-EL: −1.8 pts
Scores by race × poverty status	NAEP Data API (crosstab)	Hispanic free-lunch vs. non-free-lunch
School % LEP enrollment	NAEP school survey	C046501
State EL enrollment trends	NCES	Texas EL: 18% → 20% of enrollment

NAEP API variables: SDRACE, LEP, SLUNCH3, C046501, C044006

Analytic method: Mediation model with SDRACE as predictor, LEP as mediator, and 2022–2024 score change as outcome. Oaxaca-Blinder decomposition at the state level decomposes the Hispanic decline into within-EL and within-non-EL components.

References: Baron, R. M., & Kenny, D. A. (1986). The moderator–mediator variable distinction. Journal of Personality and Social Psychology, 51(6), 1173–1182. | Preacher, K. J., & Hayes, A. F. (2004). SPSS and SAS procedures for estimating indirect effects. Behavior Research Methods, Instruments, & Computers, 36(4), 717–731.