CREDO released a new two volume study today titled Charter School Growth and Expansion. The title actually masks the ambitious range of research questions tackled and important findings presented. Though the study presents data on patterns of growth and expansion in the charter sector, with particular attention given to networks of schools, the report is solidly about public charter school quality. And the results suggest that what we thought about when quality determinations should be made within the life cycle of charter schools may have been wrong.
Before getting to what will likely be the most talked about aspect of the new CREDO study—the ranking of individual networks, specifically the (perhaps surprising) negative results for some well-known networks—I want to address the paradigm shift this study brings about in terms of thinking about charter school quality.
There has been a widely held belief that it could take several years for new charter schools to get their sea legs, and as a result we should not expect high levels of performance for the first couple of years of a charter school’s existence. In other words, if given time, charter school performance would increase, even for the low-performing or average school. Moreover, there is a body of research that confirmed this theory. These studies examined the question by comparing the performance of “mature” charter schools with “newer” charter schools, or even the same school at different points in time, and found that charter schools in later years performed better than new schools. However, what these studies overlooked, and what the CREDO study points out, is that the models did not control for schools closing due to low-performance (or the results of low-performance, such as low-enrollment) or more importantly, examine whether early performance predicted later performance.
The CREDO study shifts the research question from looking at average performance for charter schools based on the number of years open, to examining whether charter school performance in the first one or two years of operation predicts performance up to five years after opening. The short answer is that early performance of charter schools almost entirely predicts future performance. In other words, if a charter school starts out low-performing, it has a very slim chance of making improvements. This is sobering but important information about what we can expect from charter school performance.
Fortunately, there are bright sides to the initial quality story. There are public charter schools that hit the ground running and are top performers in their states in the first years of operation. Moreover, about two thirds of the schools that replicate are in the top 40 percent of school performance. So, in many cases, the right schools, in terms of performance outcomes, are being selected for replication.
And perhaps this shouldn’t be surprising, despite earlier research and what we may have wanted to believe about the unsteadiness of new schools. CREDO lays out a theory of action that suggests that school culture, norms, and behavior are set early, and not only are these school structures and processes indicative of school performance, they are really difficult to change. But if the norms and behaviors are good, the school will continue to have success.
Now to the flashier portion of the study—the performance of networks of charter schools. CREDO redefines how we understand networks of charter schools. They define CMOs as networks of three or more charter schools where the CMO holds the charter. Unlike previous definitions of CMOs and EMOs, CMOs in this study include both non-profit and for-profit management organizations. CREDO defines EMOs as organizations that secure contractual agreements from governing boards to operate schools, and they too can be non-profit or for-profit. By this definition, CMOs could contract with EMOs to operate schools, and several do. CREDO also creates a new category of network, the “super network,” to classify the unique structure of KIPP and a couple of other management organizations, which operate as a federation of independent networks under the umbrella of a larger national manager.
CREDO compares the results for CMOs and non-CMOs are to traditional public schools. Overall, the findings are mixed. In math, both CMOs and non-CMOs perform worse than traditional public schools. In reading, the two sets of charter schools perform better. In math, non-CMOs perform worse than CMOs, but in reading they perform better. The findings are much more varied by state, and the breakouts can be found in Table 1 of Volume II. And while CMO performance is close to non-CMO performance, over time CMOs demonstrate larger effect sizes (learning gains) compared with non-CMOs, in both math and reading.
While the overall results are mixed and small in size, the effects for subgroups of students are more promising for CMOs and non-CMOs. In nearly every category and subject area, CMOs and non-CMOs outperform traditional public schools for the following breakouts: Black, Hispanic, poverty, English language learners (ELL), and special education.
Among the four super networks mentioned in the report, KIPP and Uncommon Schools outperform non-CMOs and traditional public schools, while Responsive and White Hat experience negative and mixed findings, respectively. When networks are organized by whether they received investments from the Charter School Growth Fund (CSGF), the networks associated with CSGF are doing phenomenally well.
The tables below present networks with the top ten effect sizes for math and reading (from Appendix A in Volume II). It should be noted that the effect sizes do not represent overall performance levels, they are four year growth effects to show how the charter schools impact students who attend. What emerges from the two lists of top networks is that most are small CMOs in just a handful of states (AR, AZ, CA, DC, FL, NY, OH, PA, TX), many are on both lists, and a good portion of them serve at-risk student populations.
Top Networks: Growth in Reading
* In list of top networks for both math and reading
Top Networks: Growth in Math
* In list of top networks for both math and reading
CREDO acknowledges that there is variation in performance among schools within the overall network effect. So networks that show large positive effects may have individual schools performing lower than the average, and conversely low-performing networks may have individual schools performing higher than the average. However, the overall results by network are quite interesting and will take some time to digest.
Charter school quality is an increasingly important topic, not only for closing low-performing schools but for identifying high impact school models and educational practices that can be replicated to serve more students. The new CREDO study provides a lot of information that will help the sector think about measuring performance and making decisions based on the data.