Former Atlanta superintendent of schools Beverly Hall is now the poster child for cheating on standardized tests. But Hall claims that Atlanta schools really did perform better under her leadership, and as evidence she points to gains on the national NAEP test, widely considered to be reliable and not easily gamed. Dana Goldstein comments:
Although NAEP security procedures are generally considered more stringent than those used in state and district-level testing, there are reasons to be skeptical of Atlanta’s gains on the national exam as well. Between 2002 and 2009, the demographics of Atlanta NAEP test-takers changed considerably; the number of white students taking the test doubled, and the number of Hispanic students also went up. In Atlanta, white and Hispanic children tend to score higher than black children, which led Professor Mark Musick, a former NAEP chairman, to estimate that as much as 40 percent of Atlanta’s gains could be due to changes in which students sat for the exam.
I don't quite get this. Why not just look at the NAEP results for black, white, and Hispanic kids separately and see how they did? I don't feel like doing that for every combination of kids and tests, but a quick look tells me that reading scores for black 8th graders increased from 233 to 249 during Hall's tenure, and math scores increased from 241 to 262. That's no defense of Hall, but it seems pretty straightforward to figure out how Atlanta's kids did and how that compares to other big cities. Why estimate?