Do DoDEA schools in Germany close the gap?

Rather than wasting my time reading the American Conservative or watching Bonanza on ME TV, I took a look at the 2011 TerraNova math and reading scores of Blacks and Whites in every DoDEA school in Heidelberg and Bavaria. As can be seen below, the averaged math and reading gap, n-weighted by Black sample size, was 0.6 SD. For comparison, the 2011 civilian NAEP averaged gap was 0.9 SD.. In no country (Guam, Japan, Korea, Okinawa Germany, Italy, Uk, Belgium, Netherlands, Spain, Italy, and Turkey) was there not a DoDEA B/W gap. Apparently, being abroad or in a non-civilian school doesn’t magically obliviate the differential.

Similar results can be found with regards to the SAT for all DoDEA schools — unfortunately scores are not broken down by region. The overall DoDEA B/W (M+W) SAT score difference is approximately 0.75 SD, as compared to a national difference of one sigma:

As Brown (2006) noted, the military-civilian difference in the magnitude of the B/W gap can mostly be explained by differences in parental selectivity:

The central message of the data on military parents in Table 2 is that military parents have higher test scores than their civilian counterparts. This difference is .36 standard deviations for white parents, .74 for blacks, and .64 for Hispanics. In interpreting this result, note that the DMDC data refer to the military parent. Since the military uses AFQT scores as a screen, it is not surprising that the military parent has above-average test scores. The non-military parent would also be expected to have above-average scores to the extent that assortative mating is important. Bouchard and McGue’s (1981) survey suggests that a correlation between spouses (for IQ scores) of .33, though they note that values of .5 are often used. If this range can be applied to AFQT scores, it suggests that mothers’ scores would be higher in military families by about .12-.18 for whites, and .25-.37 for blacks and .21-.32 for Hispanics. If we could select one parent at random from military families and civilian families, we would have a test score gap in favor of military parents of roughly .25 standard deviations for whites, .52 for blacks, and .45 for Hispanics…

…Multiplying this inter-generational test-score correlation times the difference between AFQT scores of military and civilian parents gives us an estimate of the effect of these differences on test scores of their children. For white children, the .358 difference in scores among parents would translate into a .15 difference in scores among children – i.e., essentially all of the .17 σ difference in Table 9. For black children, parents’ scores account for .32 σ, compared to an overall difference of .48 σ. Among Hispanic children, differences in parents’ scores account for a .27 σ difference, compared to an overall difference of .60 σ. Thus, for black and Hispanic children, parents’ test scores account for some but not all of the observed test score advantage of their children. Brown, 2006. Relatively Equal Opportunity in the Armed Forces: Impacts on Children of Military Families


Ad hoc sociological explanations are not needed.

About these ads

8 thoughts on “Do DoDEA schools in Germany close the gap?

  1. It’s been almost impossible to enlist in the military over the last 20 years if you score at the 30th percentile or below on the highly g-loaded AFQT. Before the end of Cold War, when the Army was larger, they’d recruit farther down, but the bottom 10% have been off limits by law since the early 1950s, with an unfortunate exception for awhile during Vietnam that was one of MacNamara’s brainstorms.

    Granted, that’s normed off the NLSY and I would guess that NLSY participants don’t study up for the test but people who want to be allowed to enlist do, so that would move the percentiles around a little. But, still …

  2. To spell it out, here we have 26 schools in Germany run by the Pentagon for military children who grow up in relatively standardized environments and are typically the genetic offspring of at least one individual who had to score 92 or higher on what’s pretty close to an IQ test. And yet 26 out of 26 schools see a white-black gap. In some, the gap is small enough to be nugatory, but still … 26 out 26, and an average of 0.6 sd.

    And black recruits don’t come from the bottom of the societal barrel, by any means. According to the 1994 book “All We Can Be” written by a couple of sociologists of military life, black recruits have virtually the same parental income as white recruits. Lots of black recruits are second or third generation military.

    This kind of thing suggests we’re seeing regression toward the mean in action.

    • Steve, thanks for the elaboration. I’ve always felt disinclined to expand upon points that seem obvious.

      As for selectivity, Brown (2006) provides quite specific estimates:

      “While test scores of those entering the military are a well-publicized indicator of recruiting success, data on scores of those who re-enlist are not generally available. The Defense Manpower Data Center provided data on members of the Armed Forces on active duty in December 1997. I selected those with dependents born in 1984 (i.e., those 13 years of age). AFQT scores were available for nearly all enlisted personnel. These are shown in line 4 of Table 2. AFQT scores are not available for officers; I imputed these (based on race/ethnic group and years of schooling) using NLSY data. Mean AFQT scores including those imputed for officers are on the fifth line of the table.17

      The central message of the data on military parents in Table 2 is that military parents have higher test scores than their civilian counterparts. This difference is .36 standard deviations for white parents, .74 for blacks, and .64 for Hispanics. In interpreting this result, note that the DMDC data refer to the military parent. Since the military uses AFQT scores as a screen, it is not surprising that the military parent has above-average test scores. The non-military parent would also be expected to have above-average scores to the extent that assortative mating is important. Bouchard and McGue’s (1981) survey suggests that a correlation between spouses (for IQ scores) of .33, though they note that values of .5 are often used. If this range can be applied to AFQT scores, it suggests that mothers’ scores would be higher in military families by about .12-.18 for whites, and .25-.37 for blacks and .21-.32 for Hispanics. If we could select one parent at random from military families and civilian families, we would have a test score gap in favor of military parents of roughly .25 standard deviations for whites, .52 for blacks, and .45 for Hispanics.”

      And so on…(the paper is worth the read if you’re interested in the topic)

      Given Brown’s figures, the mean Black enlistee IQ is about 95. Given the coefficient of assortative mating of about 0.33, assuming the number of military-military relations are negligible, the Black mid parental IQ should be about 92. With regression, the offspring IQ should be a little under 90, assuming no other factors involved. This would explain why they perform 0.3 SD better than the kids of civilian Blacks.

  3. For a lot of blacks who enlist, “It is a job, not an adventure.” Many go for supply jobs rather than combat jobs. Quite reasonably, they are looking for a well-ordered work environment that teaches skills that they can continue to employ in middle age in say, corporate or government office jobs as they collect their military pensions. Black enlistees tend to come from the sensible lower middle-class.

comments do not require an email

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s