Achievement effects as measured by the Stanford Achievement Test 9 were estimated for students in grades 1-11 that participated in America's Choice, Co-nect, Different Ways of Knowing, Success for All (SFA), and Urban Learning Centers comprehensive school reform (CSR) designs implemented in Los Angeles Unified School District between 1999 and 2002. Propensity score matching was used to construct comparison groups, and difference-of-differences was used to estimate achievement effects as a function of the duration of student participation and how early in their educational careers students participated in each reform. This study overcame methodological issues identified by Borman, Hewes, Overman, & Brown (2003) and yielded effects estimates that were generally less than prior research has suggested. Results suggested that comparison group selection methods accounted for smaller effect sizes rather than an artifact of where the reforms were implemented. No consistent evidence across the 5 CSR designs was found that earlier intervention caused larger gains. However, students that began SFA in grades 1 or 2 benefited between 0.10 sigma and 0.16 sigma in reading, mathematics, and language arts. Conditional on a statistically significant average achievement effect, there was strong evidence that longer participation caused larger effects. Policy makers thinking of implementing a CSR design must consider that prior effects research might have overestimated achievement effects due to less internally valid research methods. (Contains 7 tables and 1 figure.)