What Effect Do Charter Schools Have on Test-Score Gains for Students Who Transfer Between Traditional Public Schools and Charter Schools?
. . . .In five out of seven locales, these nonprimary charter schools are producing achievement gains that are, on average, neither substantially better nor substantially worse than those of local TPSs. In Chicago (in reading) and in Texas (in both reading and math), charter middle schools appear to be falling short of traditional public middle schools. Results that include charter schools at every tested grade level (i.e., those that start in kindergarten as well as those that serve exclusively middle- and high-school grades) are, in most cases, similar to the results that are limited to nonprimary charter schools, providing no evidence that charter-school performance varies systematically by grade level.
The inclusion of kindergarten-entry charter schools in the analysis makes a substantial difference to our estimate of their achievement impacts in only one location. In Ohio, as in most of the other sites, the average performance of nonprimary charter schools is indistinguishable from that of nonprimary TPSs. But when the K-entry charter schools are included in the analysis, the estimated impact of Ohio’s charter schools is significantly and substantially negative. The dramatically lower estimated performance of Ohio’s K-entry charter schools appears to be attributable not to grade level per se but to virtual charter schools that use technology to deliver education to students in their homes. Virtual schools constitute a large part of the enrollment of K-entry charter schools in Ohio, and students have significantly and substantially lower achievement gains while attending virtual charter schools than they experience in TPSs. This result should be interpreted cautiously, because students who enroll in virtual charter schools may be quite unusual, and their prior achievement trajectories may not be good predictors of their future achievement trajectories (pp. xiii-xiv).
. . . .
Randomized experiments are often considered the gold standard in research because, by assigning subjects randomly to the treatment condition or control condition, they ensure that differences observed later are the result of treatment rather than the result of background differences between the subject groups. A few studies are beginning to examine oversubscribed charter schools that randomly admit students through lotteries. For instance, Hoxby and Rockoff (2004) found that four Chicago charter schools that admitted students by lottery were outperforming TPSs as measured by students’ subsequent achievement. Later, Hoxby and Murarka (2007) used a similar design to evaluate 47 charter schools in New York City and likewise found a small positive achievement effect for students attending charter schools. Abdulkadiroglu et al. (2009) found that a subset of charter middle and high schools in Boston that used admission lotteries also found positive impacts (sometimes large). And Mathematica Policy Research is engaged in a federally funded national study of oversubscribed charter middle schools that admit students by lottery; results are not yet available (see Mathematica Policy Research, undated).
These lottery-based studies have strong internal validity: Researchers can be confident that the participating charter schools caused the achievement advantages for the students who were admitted in their lotteries. But although the studies should produce internally valid and reliable results for the set of charter schools and students examined, they may have limited implications for charter schools that lack lengthy waiting lists and do not use lotteries to admit students. In other words, these studies have weak external validity. Charter schools with lengthy waiting lists might well be those that are better than average. Indeed, the only study that has begun to examine the issue found that charter schools using admission lotteries appeared to be more effective than charter schools that were not oversubscribed (Abdulkadiroglu et al., 2009) (p. 22).
Comment: In other words, charter chains like KIPP that have an unending national advertising campaign and that open one or two stores in urban areas where there are many crumbling schools, have an easier time keeping their desks full, even as 40-60% of students who enter KIPP in 5th grade do not make it to 8th grade at KIPP. Low performers or perpetual "miscreants" can be dumped, allowing in others with parents who are willing to subject themselves and their children to the KIPP treatment. A concentration of high performers become the survivors with the miraculous test scores, and the PR machine starts all over again.
. . . .
Table 3.3 presents estimated impacts for nonprimary charter schools in each of the sites. Of the 14 estimates across the seven locales, 11 suggest charter-school impacts that are indistinguishable from those of TPSs. Three of the estimates are significantly negative, for middle schools in Chicago in reading (at 0.01) and middle schools in Texas in both subjects (at 0.01 in reading and 0.05 in math). High schools are not included in this analysis in Chicago and Texas because test data are unavailable in successive high-school grades in those sites. (Chicago’s high schools are analyzed using a different method in Chapter Four.)Reference:
Interestingly, the results for nonprimary schools in Table 3.3 for many of the sites are quite consistent with the results in Table 3.1 that included elementary schools. All of the results that are statistically indistinguishable from 0 in Table 3.1 remain so in Table 3.3. In Milwaukee, the small positive result in math in Table 3.1 declines slightly and loses statistical significance in Table 3.3, but the point estimate does not change much. Similarly, Denver’s positive math effect reduces in magnitude and becomes statistically insignificant. The significantly
negative results in Chicago (reading) and Texas (both subjects) remain significantly negative in Table 3.3.
Ohio, however, presents a different story. The substantial negative results suggested in Table 3.1 disappear when focusing on the subset of nonprimary charter schools. For Ohio’s charter middle schools (our test data in Ohio end at eighth grade, so high schools are not included), estimated achievement impacts are indistinguishable from those of TPSs (pp. 37-38).
. . . .
Those results suggest that the performance of nonprimary charter schools is approximately on par with that of TPSs in most of the sites, though middle schools in Texas and Chicago appear to be falling behind (p. 39.)
. . . .
The analysis suggests that nonprimary charter schools are producing achievement gains that are approximately equivalent to those of TPSs in most locations, with moderately negative effects in math and reading in Texas middle schools and in reading in Chicago middle schools. While our results for Ohio’s virtual charter schools should be viewed with a level of caution because of the uniqueness of the students who attend these schools and because much of the analysis relies on charter schools with primary grades, they suggest that these schools should be examined more carefully because of the poor achievement results. We find support for the conclusion that, in most locations, charter schools do not do well in their first year of operation but subsequently improve (though sometimes this improvement is sufficient only to produce a result that is somewhat less negative than in the first year of operation). Finally, we find that charter schools in most locales have marginally greater variation in performance than TPSs, as measured by the achievement-impact estimate for each school. Ohio is a notable exception: Its charter schools have a much wider range of variation in performance than its TPSs (p. 51).
Zimmer, R., Gill, B., Booker, K., Lavertu, S., Sass, T., Witte, J. (2009). Charter schools in eight states: Effects on achievement, attainment, integration, and competition. Santa Monica, CA: Rand Corporation.