The Mysterious Decline in Black Recruits (Crap Journalism Edition)

Today saw a raft of stories about the perilous decline in black military recruits (go here for a list of such stories).

The basic take? This Wash Post piece is representative. Headlined "Steady Drop in Black Army Recruits: Data Said to Reflect Views on Iraq War," it notes that blacks now make up less than 14 percent of recruits, down from 23.5 percent in 2000.

Let's leave aside for the moment whether this is a good thing or a bad thing. After all, blacks made up about 12 percent of the US population in the 2000 Census and there's a perennial argument that their historic overrepresentation in the all-volunteer military meant such a system was worse for them than a draft. It may simply be expected that over time, any given ethnic group's representation in the military will approach its proportion in the population at large. Who knows?

But there's a couple of elements of crap journalism in the general spin of this story. The Post also notes:

Hispanics have increased from 10.4 percent of new recruits in 2000 to 13 percent in 2004; whites went from 61 percent in 2000 to 65 percent in 2004; and Asians or Pacific Islanders made up less than 1 percent of new soldiers in 2000 but nearly 5 percent in 2004.

So the story might have been headlined something like: "Asians and Pacific Islanders Bursting with Patriotism, Increase Number of Recruits 5X." Or something touting the increases, in percentage terms, of white and Hispanic recruits.

More important, neither the Post story, nor any of the others I skimmed, actually gave raw numbers of recruits. Instead, all we get are relative percentages. Which is fine, but also woefully incomplete.

How many total recruits were there in 2004 vs. 2000? I mean, if the number of total recruits doubled in those four years, it would mean that, in absolute numbers, there were more black recruits in 2004 than 2000. Without those figures, there's no way of knowing. Have recruitments been flat? Or what? It shouldn't have taken a reporter very long to find out and report that along with the percentage shifts.

Comments

[D Anghelone]

You'd also have to look at age distributions within groups to see from where are the potential recruits.

[MB]

That damn liberal MSM!

Oh, wait, Stars and Stripes "spun" it the same way:

http://www.estripes.com/article.asp?section=104&article=26657&archive=true

[Mr. Nice Guy]

Another thing the Washington comPost failed to point out is that many people consider the military as a jobs program during peace time. If there is actual fighting.. well, forget about it.

[biologist]

here's a fine argument in favor of mandatory statistics education in high school, which should lead to more critical thinking by the populace, the education of whom is vital to self-governance

a one-sample chi-square test would answer the above criticisms/ questions very satisfactorily

[jet]

If my memory is correct (a big if), chi-square tests are necessary when looking at data generated from random samples. The data in question here, from military recruits, appears to be "population" data. If so, then any differences across groups (expressed in percentages) are "statistically significant" differences.

[SR]

This reminds me of the screaming headlines a year or two ago about how blacks were "disappearing" from major league baseball. Yet when you actually read the stories, you learned that the number of *whites* playing major league baseball was declining as well. The numbers for Hispanics and Asians increased significantly, however.

[js]

"here's a fine argument in favor of mandatory
statistics education in high school, which should lead to more critical thinking by the populace ..."

Haha, you're joking right? Afterall, who remembers any of the course they had to take in high school for reasons they could not fathom and that they had no interest in?

Perhaps, here is a fine argument, for concerned adults to read any of the numerous books showing various tricks that can be done with statistics.

[themantheththelegend]

biologist and js,

Or maybe it's an argument for major news organizations to keep a statistician on staff to peer review articles thatemploy statistics. The frequency with which this sort of thing happens should be pretty embarrasing to papers like the Post which I'm sure fancy themselves as the informers of the masses.

[smacky]

Perhaps, here is a fine argument, for concerned adults to read any of the numerous books showing various tricks that can be done with statistics.

There is a really great book that literally illustrates how statistics can be manipulated graphically. It's called, "The Visual Display of Quantitative Information" by Edward R. Tufte. It's a great book, especially if you like to look at a lot of pictures. There is a very pertinent chapter in it called "Graphical Integrity", chapter 2, that shows how easily information can be misinterpreted or misrepresented. And there are even real examples of false statistical interpretations that were actually printed for reading by the general public! To say that publishers would be embarrassed about this kind of stuff would be a stretch in some cases -- I'm sure some of them intentionally spin the statistics in order to favor their hidden agenda (which I am suspicious of, in the case of these Army recruitment stats).

[PapayaSF]

Not to to diss Tufte, but a better recommendation for general readers would be the slightly dated but classic How to Lie With Statistics by Darrell Huff. Short, funny, and with amusing illustrations, the lessons of the book have stayed with me for 30 years.

[thoreau]

The important thing about a statistics class is not that you remember the formula for the chi-square test or how to calculate a standard deviation or regression coefficients. The important thing is that you learn what sorts of questions to ask about data. You don't have to know the equations for multiple regressions to know that there are ways to control for variables. You don't have to remember how to calculate a correlation coefficient to ask whether a number changed in absolute terms or only relative terms. You don't have to remember the difference between a chi-square test and t-test to ask whether comparisons were done in a manner that really only changes one variable.

And so forth.

[saw-whet]

PapayaSF, I just bought How to Lie With Statistics, Crimes Against Logic and Bad Thoughts. Anyone else have suggestions along those lines?

[Don]

Another thing the Washington comPost failed to point out is that many people consider the military as a jobs program during peace time. If there is actual fighting.. well, forget about it.

Blacks tend to be overrepresented in the non-combat arms portions of the military, and underrepresented in Special Forces and other "tip of the spear" units. In other words, they make use of the military as a job program, and that explains their historically high enlistment rate. Now that it is less of a job program and more of a combat role, it makes sense that their enlistment drops.