Thursday, September 13, 2007

Looking for the cloud

Stories on testing of school children hardly ever give a full and fair picture. In this case, the reporter James Vaznis finds something wrong with the increase in MCAS scores in 2007.

The something is racial and ethnic disparities, and that is indeed a persistent problem. What Vaznis doesn't tell us until after the jump is that the long-standing disparity also improved. So I guess the bad news that every test score story must by journalistic convention include is that we don't live in heaven on earth.

Vaznis doesn't provide a socioeconomic breakdown of the MCAS data, but then the state doesn't provide it. There are privacy concerns of course. However, Vaznis could have looked at income averages and correlated them with average test scores. That might not be good enough for a peer-reviewed journal, but it would be good enough for a newspaper. In fact, I did it almost ten years ago for a local weekly.

There's a strong correlation between wealth and test scores, and it explains part of the ethnic and racial disparities. Of course, that explanation is only an explanation, not an excuse to do nothing. If we really want to be a classless society, we've got to educate our poor people well, or we doom social mobility. No real 'if' - the middle class nation we had forty years ago has been under persistent and successful attack from the right with little defense from the left.

Another bit of context missing from this story, as from most Boston Globe stories on test scores, is the fact that Massachusetts and Connecticut students are duking out for number one in the nation on nationally normed tests. The only fair conclusion is that our NCLB-mandated testing is much harder than, say, the similar testing done in Florida and Texas and put in place by two of the worst Bushist governors, Duhbya and Jeb.

One last point (and then pencils down!): The summaries shown here indicate that the 10th grade math test is too easy to tell the difference between really good math students and merely mediocre ones. This is the real standard that NCLB has set, and it encourages dumbing down the test. All the other distributions look reasonably normal, so far as I can tell from such coarse data.

No comments: