Long-Term Impacts of Reading Recovery through 3rd and 4th Grade: A Regression Discontinuity Study

Grade 3 N=9,879; Grade 4 N=6,345 (73-83% student attrition)

Lowest grade-1 readers who received Reading Recovery, followed to grades 3-4; US.

Multisite fuzzy regression-discontinuity on the Observation Survey eligibility cutoff. CONTESTED: the What Works Clearinghouse (2023) ruled the study INELIGIBLE as an RDD (it analyzed a forcing variable not used to assign students), reviewed it as a QED, and rated it 'Does Not Meet WWC Standards' for baseline equivalence. Independent critic Jon Baron (Evidence-Based Policy) calls the negative finding 'not a reliable finding' (73-83% attrition; ~0.85 SD baseline gap). Downgraded B->C.

The same cohorts that showed large positive grade-1 gains scored ~0.19-0.43 SD BELOW comparison children on grade 3-4 state tests - a fadeout-to-reversal. But the study is design-contested (WWC ruled it ineligible as an RDD; 73-83% attrition), so it supports 'no credible evidence of durable benefit' rather than a confident 'RR harms.'

RD would control selection if valid, but the design's validity is disputed by WWC and independent critics.

No independent long-term RCT of Reading Recovery exists. This is the only US long-term estimate; its negative sign is real in-sample but the causal read is contested.