In a forthcoming book, I devote about one third of the pages to "Principles of Data Interpretation, Or, How Not To Get Statistically Snookered by Ideologues, Politicians, Incompetents and Crooks." I note how "A Nation At Risk," with 9 NAEP trends, only one of which would support crisis rhetoric, used only that one. And so forth.

A similar thing happened in the December 22, 1999 edition of the Wall Street Journal, the viper's nest of the title, an appellation donated by Slate magazine. Slate meant it only in regard to the editorial page and it is true that most WSJ stories, as opposed to editorials, have been fair and balanced. The one at hand is not. A page one story by Daniel Golden lauded the domestic military schools (as opposed to DODDS schools abroad) for doing so well while civilian schools with similar levels of poverty and minority proportions failed so miserably. The answer, of course, is that populations are not similar and that Golden had picked one piece of data where the military schools looked especially good.

This piece of data is the most recent NAEP writing test where these schools finished second only to Connecticut. They did this in spite of the fact that 40% of the students are black or Hispanic and in spite of the fact that 50% of the kids are eligible for free and reduced lunch.

I don't have the writing report and will be giving data from the 1996 Math, 1996 Science and 1998 Reading Report Cards. They are maddeningly unstandardized so not all data is there for all three.

Two quick points on the poverty stats:

For only the reading report card is the military eligibility 50%. The Math Report Card, for instance gives it as 35% for 4^th graders and 29% for 8^th graders. The respective civilian proportions are 34% and 30% (all civilian data in this report are for public schools only). The science report card gives the data only for 8^th graders: 24% for military kids, 29% for civilians. The variability of the data suggests that something is wrong with the NAEP sampling plan or with the report.

The scores of military kids who are eligible for free and reduced lunch are higher than civilian school kids' scores. The science score for eligible military 4^th graders is 218 compared to 207 for civilians. For 8^th graders, it is 260 vs. 252.

However, you do not have to be living in abject poverty to be eligible and the military kids are not. Free lunch qualification extends to 175% of poverty. You won't find military kids living in the conditions found in Southeast Los Angeles or Spanish Harlem (in which conditions, I remind you, children spend 91% of their lives from birth to age 18). No doubt more of them come from two-parent families as well.

THE CLINCHER THAT GOLDEN FORGOT:

The most cogent, compelling statistic that Golden totally ignored (on purpose?) is that the parents of students in the military's schools are much better educated than those in public schools. Parental educational level is typically the most potent predictor of test scores. The Science and Math Report Cards give these figures

Percent at Each Educational Level

When we compare children of like-educated parents, the Department of Defense Mystique disappears:

NAEP SCORES, GRADE 8

So, dear readers, when we compare children from similar backgrounds, mirabile dictu, they score the same

The magic of the military is a mirage.

Some other things not considered by the article: school and class size. Golden notes that there are 34,000 students in 71 schools. That's fewer than 480 kids per school. He doesn't provide any statistics that permit calculation of class size although he says it's "about the same as the national average."

By looking only at writing scores, Golden fails to notice the variability in scores from subject to subject. In reading, military schools do almost in writing, but in science they rank only 18^th and in math 26^th. Looks like they're focusing on literacy and paying for it in math and science.

By calling attention to ranks, Golden calls attention away from scores. In reading, where the schools tie with Massachusetts for second place at the 8^th grade level, the military schools score at 269 while the national average is 261. Small increases in a score will produce large increases in ranks.

Golden also does allow that the military spends $7,279 per kid, 23% more than the national average. Their best performance, though, on reading, is only 3% better than the national average.(on reading, 269 vs. 261).

Military kids do move around a lot but the schools share a common curriculum which helps. In addition, according to Golden, many base teachers were themselves "Army brats" and know the special needs of the children of military.