The demographic composition of students transferring out of KIPP schools and out of comparison middle schools is also broadly similar. At both KIPP schools and comparison schools, students who leave early are significantly more likely to be black males and significantly less likely to have limited English proficiency than students who remain. Those who leave early from KIPP are also less likely to be Hispanic, whereas at comparison middle schools the proportion of Hispanic students who leave is nearly identical to the Hispanic proportion of students who remain enrolled. In addition, students who leave KIPP are no more likely than those who stay to be FRPL-eligible (both groups have high eligibility rates) or to be in special education, whereas students who leave comparison middle schools are more likely than those who stay to be FRPL-eligible and to participate in special education programs. Thus, if anything, it appears that the attrition from comparison middle schools is more likely than attrition from KIPP to draw away students who are disadvantaged relative to the students left behind.
We find that, on average, KIPP middle schools admit students who are similar to those in other local schools, and patterns of student attrition are typically no different at KIPP than at nearby public middle schools. In both groups of schools, students who leave before completing middle school are substantially lower-achieving than those who remain. KIPP schools replace fewer of these students in the last two years of middle school, however, and, compared to district schools, KIPP schools tend to replace those who leave with higher-achieving students. Nonetheless, while this difference in replacement patterns is noteworthy, it cannot account for KIPP’s overall impact on student achievement. In particular, the literature on peer effects suggests that KIPP’s student replacement pattern could produce only a small fraction of KIPP’s actual impact on student achievement. A large part of KIPP’s cumulative effect occurs in students’ first year of enrollment, before attrition and replacement could have any effect.
Compare and Contrast essay High School vs.
How much of KIPP’s impact on student achievement might be explained by the fact that KIPP tends to attract higher-achieving new students in the upper grades? One way to estimate the possible size of peer effects at KIPP is to combine our findings with other research on how peers’ prior scores affect student achievement. Unfortunately, published estimates of the effect of peer ability on student achievement range widely, from close to zero to nearly half a standard deviation impact for each standard deviation of difference in peer achievement. Even if the largest estimates of peer effects are correct, however, the improvement in peers’ prior test scores would appear to benefit KIPP students’ achievement only by about 0.07 to 0.09 standard deviations after four years at KIPP. KIPP’s cumulative impacts in middle school are three times that size, so even the largest estimates of the size of peer effects suggest that they are unlikely to explain more than one-third of the cumulative KIPP impact.