Though the immediate effect of Reading Recovery (RR) is both strong and well established, the longer term or sustained effect has been less studied and the evidence regarding it has been less conclusive. Michigan Reading Recovery students (n = 328) were compared to control students (n = 264) while in first (2009–2010), third (2011), and fourth grades (2012), using propensity score matching to generate 3 levels of eligibility. Although the immediate effect measured at mid-year of first grade on the Observation Survey was large (1.17), the effect by the end of first grade on the same measure was .51, and by third grade, the effect was .16 on the state reading test. The overall effect completely diminished by fourth grade, but it was significant (.35) for the most eligible students in reading, and for moderately eligible (.34) and most eligible students (.35) in writing. The sustained effect overall was present but diminished by third grade, and was sustained into fourth grade for those students at greater risk. The findings suggest that RR instruction should be better tailored to the initial literacy profiles of individual students to maximize the longevity of the effect for all participants.