Bob Somerby is complaining today about numerical illiteracy among our nation's elite reporting class. Item 1: the New York Times describes a 10-point improvement among fourth graders on the NAEP test as "small." In fact, it's roughly a full grade level. If you think that improving by a full grade level in a single decade is small, you're either crazy or innumerate.
Item 2: M. Night Shyamalan talks about the fact that American test scores are pretty high in "districts in which the poverty rate was less than 10 percent." However, the only income data we have for most test takers is related to the National School Lunch Program. Shyamalan is using eligibility for free or reduced meals as a marker of poverty. But it's not. And since here at MoJo we're dedicated to lighting a candle instead of cursing the darkness, here are the exact eligibility requirements for free and reduced-price lunches in 2013, courtesy of the Agriculture Department:
Obviously, folks eligible for reduced-price meals aren't exactly swimming in cash. Still, a family of three making $36,000 isn't anyone's idea of poverty, and it's misleading to say so. Eligibility for free meals would be a fairly decent proxy for poverty—they account for about a third of all NSLP meals—but unfortunately that data isn't collected separately. You either qualify for NSLP or you don't, and something like two-thirds of all schoolchildren qualify. It's a pretty broad brush, and there are damn few school districts in which fewer than 10 percent of kids qualify.
FWIW, this is why I've never bothered breaking down test scores by income. The only data available is eligibility for NSLP, and between the loose requirements and the virtual nonexistence of verification1, it simply doesn't mean very much. It can give you a very broad feel for how rich or poor a particular school or district is, but that's it.
1Which I'm all in favor of, by the way. This is a program that probably doesn't benefit from tighter scrutiny. Nonetheless, it makes it nearly useless as a proxy for poverty among test takers.