We are pleased that the authors of the Civil Rights Project (CRP) report on racial segregation in charter schools have chosen to respond to our reanalysis of the 2007-08 data. This dialogue is important as we attempt to move toward the most appropriate analytic strategies for this question. Indeed, the CRP response begins by highlighting the continued significance of school segregation and highlights the benefits of integration (of all sorts) for students. We completely agree with the CRP authors on this point. However, we take issue with three points made (or not made) in the CRP response.
1. Methodological differences
2. Characterization of the RAND study
3. Neglecting to consider our concluding points
Methodological Differences
Our primary criticism of the CRP methodology was the following, taken word for word from our article:
In every case, whether the authors examine the numbers at the national, state, or metropolitan level, they compare the racial composition of all charter schools to that of all traditional public schools.
We believe that these comparisons may well generate misleading conclusions because charter schools are not sprinkled across the nation randomly. Rather, charters are disproportionately opened in disadvantaged urban areas where families and students need additional educational options.
Appropriately, the CRP response begins by defending the methodology of its analysis. Before getting into the nuts and bolts of the analysis, it is important to note that the CRP authors do not offer any rebuttal to the criticisms we made concerning their use of national and state comparisons. This is in spite of the fact that the national and state-level comparisons were central components of the report. Every statistic in their executive summary and in the accompanying press release came from the national and regional comparisons. Perhaps both we and the CRP authors agree that we cannot infer anything from the comparisons made at such high levels of aggregation.
In our re-analysis published in Education Next, we focused most of our efforts reanalyzing the CRP analysis comparing charter and traditional public schools at the CBSA level –the Core-Based Statistical Area that the Census Bureau uses to define a metropolitan area. It is our view that this analytic strategy was the most appropriate strategy used in the CRP report, but as we show, even it generates misleading conclusions.
The CRP authors defend their analytic choices and criticize our reanalysis by making two claims. First, they argue that, because charter schools are not constrained by district boundaries, the metropolitan area represents the proper point of comparison. Second, they assert that the original report includes an analysis (Table 22) restricted only to cities which serves to confirm their findings. We disagree on both counts. In the paragraphs that follow, we will show that our central cities approach provides a far more appropriate unit of comparison, and we will demonstrate that the CRP Table 22 examining data for cities across the nation is off base.
Metropolitan regions are not a good point of comparison
Although charter schools can (and sometimes do) attract students across traditional boundary lines, we do not believe it is reasonable to expect that students move freely across entire CBSAs to attend charter schools for a number of reasons. First of all, CBSAs are very large. For example, in the case of Washington D.C., if the entire CBSA were an appropriate point of comparison, charter students would be crossing state lines (since the Washington D.C. CBSA also includes Virginia, Maryland, and West Virginia) and city boundaries in the 5,000 square mile region in an effort to travel to charter schools in the heart of inner city D.C. Of course, it doesn’t make sense to compare, for instance, the charter schools in Washington D.C. (where 93% of the charter schools in the metro region are located) to the traditional public schools in Front Royal, VA, which is 63 miles away!
This is the case in most CBSAs, which are extremely large. The Phoenix CBSA, for example, measures more than 150 miles across. Houston’s CBSA is over 140 miles across. If the geographic barriers weren’t enough, in many of the central cities of the CBSAs used in the CRP report, a student must either be a resident of the city (such as in Boston, see here) or, if they wish to travel from outside the city, they must pay tuition to the receiving charter school (see Washington D.C.’s rules here).
We thought it impractical that students would travel great distances to span these large CBSAs, so we looked for some evidence on the distance most charter students travel. A recent report by Teske, Fitzpatrick, and O’Brien (2009, link found here) found that most students do not travel more than 10 miles to attend a charter school. The authors reported that 92% of all students attending charter schools in the Denver and DC areas traveled 10 miles or less, 72% travelled less than five miles, and 25% travelled less than a mile.
Second, the CRP report is primarily concerned with schools that are minority hyper-segregated; are we truly concerned that minority students are transferring from integrated suburban schools to venture into central cities to attend racially segregated schools? Of course not. The vast majority of students transferring into these charter schools are leaving nearby traditional public schools. Simply put, the odds are stacked against students travelling great distances and crossing city boundaries to attend these mostly-minority charter schools.
To support their approach, the CRP authors cite a 1998 “study from Arizona that found that charter schools within one traditional public school district pulled students from 21 distinct districts.” This same citation was given as evidence to support their method in their original report and its accompanying press release.
Despite contacting the CRP, the study’s original authors, and the Goldwater Institute which published the original report, we have not been able to obtain a copy of this apocryphal report and are thus unable to verify the specific claims made. Nevertheless, it is particularly important for readers to know that Arizona has some of the smallest districts in the nation. Phoenix alone has 30 unique school districts (305 if charter districts are counted), and the state of Arizona has 612 districts. It hardly seems appropriate to use a district count from Arizona — a state unique both in terms of district size and number of charters — as evidence that students across the nation are willing to travel hundreds of miles to attend inner-city charter schools. It simply doesn’t make sense.
The CRP “city” analysis misses the mark
The CRP response to our Ed Next article points out that the authors of the CRP report did consider the city as a unit of comparison and present the results in Table 22 on page 63 of the original report. This table presents the results of a nationwide analysis, breaking down schools as located in one of three categories — city, suburb, and town/rural. This does seem reasonable enough on its face, and their results show that 52% of students in city charters are attending hyper-segregated schools, as compared to only 34% of city traditional public school students. However, there is again a key flaw in this analysis in terms of the comparability of the charter sample to the traditional public school sample.
According to the Census categorizations, cities are defined as a “territory inside an urbanized area and inside a principal city” of an MSA and can be either large, midsize, or small. Small cities can have populations of less than 100,000. Thus, the CRP analysis on Table 22 includes traditional public schools in small cities such as Appleton, WI, Ithaca, NY, and Round Rock, TX, which do not have charter schools and have very few minority students. Schools in these areas will never have minority populations of 90% or greater since very few minorities live in these white cities. In fact, many of the traditional public schools in the Table 22 CRP sample are located in states like South Dakota, North Dakota, or Nebraska–states that don’t even have charter school laws! It seems unreasonable to use these city schools as points of comparison, since they are completely different than the cities in which charter schools predominately locate.
In the end, a national sample of “cities” includes many traditional public schools that do not serve as a reasonable counterfactual to charter schools. Thus after some consideration of the objections by the CRP authors, we remain comfortable with the point made in our re-analysis — it is more appropriate to compare charter schools to traditional public schools within central cities of major metropolitan regions.[1]
Characterization of the RAND Study
The CRP authors then proceed to place their report in context by reprinting the text from their original report describing the results of the 2009 RAND study on this topic. We agree with the CRP authors that the RAND study provides a useful point of comparison, as the methodology of the report is sound and the reputation of RAND is stellar. However, the CRP authors simply misread and misstate the results, claiming that the RAND study “determined that in five of the seven locales, the movement of black students to charter schools meant these students attended more segregated schools.”
Here is what the RAND authors actually said, on page 19 of the report,
Overall, across the two analyses, it does not appear that charter schools are systematically skimming high-achieving students or dramatically affecting the racial mix of schools for transferring students. Students transferring to charter schools had prior achievement levels that were generally similar to or lower than those of their TPS [traditional public school] peers. And transfers had surprisingly little effect on racial distributions across the sites: Typically, students transferring to charter schools moved to schools with racial distributions similar to those of the TPSs from which they came. There is some evidence, however, that African American students transferring to charters are more likely to end up in schools with higher percentages of students of their own race, a finding that is consistent with prior results in North Carolina (Bifulco and Ladd, 2007).
The CRP authors appear to be drawing their inferences from the final sentence above and from the following figures from the RAND report (pulled from Table 2.3):
By the CRP logic, the five locales in which charters lead to more segregation for black students must have been Denver, Philadelphia, San Diego, Ohio, and Texas. However, to draw such a strong conclusion, despite the explicit RAND conclusion that charter students moved into schools with “racial distributions similar to the TPSs from which they came,” the CRP authors ignored the magnitudes of the changes in black enrollment.
In Philadelphia, for example, black students moved from segregated traditional public schools (84.2% black) to segregated charter schools (87.0% black). In Ohio, the pattern is similar (74.1% to 78.9% black). In each case, a reasonable conclusion is that the average charter student left a heavily black traditional public school for a heavily black charter school.
In Denver, black students transferred from traditional public schools that are 42.2% black to charters that are 51.0% black; both the TPS and the charters had about 15% white students. San Diego provides another interesting example. Here, the black students left TPS schools that were 25% black and entered charter schools that were 33% black. Of course, it seems a stretch to refer to charters that are about one-third black or one-half black as segregated black systems.
Clearly, the takeaway message from the RAND study — as the authors explicitly state — is that students who transfer into charters in these cities attend schools with racial compositions that are similar to the TPS attended in the prior year.
Neglecting our Concluding Points
Finally, we think it is important to reiterate two points — highlighted in our conclusion — that the CRP response does not address.
1) First, neither the traditional public schools nor charter schools are doing a particularly good job at drawing racially diverse student bodies. Those genuinely concerned with the racial segregation in schools should focus their attention on traditional public schools, where the vast majority (97%) of U.S. students are enrolled.
2) Families that send their children to charter schools are making a choice that best fits what they seek in an educational experience. To compare this choice to the forced segregation that occurred a half century ago is a trivialization of the true oppression that occurred. And to refer to these schools as “apartheid schools,” which implies that families are legally and physically required to attend segregated schools, is nothing more than alarmist rhetoric. Such charges would be more appropriate if they were leveled at traditional public schools where students in residential boundaries are forced to attend segregated schools.
The Civil Rights Project has a history of seeking justice and we commend that. But the organization is simply on the wrong side of this issue. First of all, the empirical data do not support the CRP claims. Moreover, although the group’s leaders have called for regulations to encourage more integrated charters, vocal critics of charters will certainly use the conclusions drawn from the CRP differently. And, if critics of charters were successful in limiting the growth of charter schools, the educational options available to poor and minority students would be further restricted. We and the CRP authors would certainly agree that this outcome would not enhance the civil rights of our nation’s disadvantaged students.
[1] This does not mean that we have landed upon the analysis we think is best; we simply believe that it is better than the CRP strategy. Indeed, as we stated in our original article, we believe our method results in a lower-bound estimate of the actual segregation in the traditional public schools that charter students would have otherwise attended. We are currently trying to improve our analysis by using zip code data, which may allow us to gather information on a more comparable set of traditional public schools.