Teaching at the Low Level. Evidence on Direct Effects, Spillovers, and Parental Engagement in Remedial Education in El Salvador

When classroom instruction is locked to the curriculum and home support is limited, students with accumulated learning deficits face near-zero marginal returns to instruction. We study two remedial literacy interventions in rural El Salvador using an RCT that randomizes 606 households (with at least one low-performing student) from 26 schools to: (i) five months of twice-weekly, 90-minute small-group pull-out sessions during school hours, or (ii) the same sessions plus a low-cost parental engagement component (WhatsApp messages and three focus meetings), versus control. We estimate small treatment–control differences in literacy endline scores, but these are masked by sizable within-school spillovers: the average control student (22.7% treated peers) gains 0.10–0.14 SD, while the average treated student gains 0.20–0.25 SD relative to low-exposure controls. The parent add-on yields no additional gains and does not increase measured parental engagement. Evidence suggests scope for further gains through higher realized treatment exposure, improved targeting, and more intensive parent-facing designs that translate information into sustained at-home practice.