The Heritage Foundation's Center for Data Analysis recently published a study on troop demographics.[1] That study's conclusions appear to perfectly contradict those of a similar study by the National Priorities Project (NPP).[2] Both studies analyzed the same Pentagon data and employed the same kind of zip code analysis of recruits' homes of record. Many reporters and policymakers have asked how to interpret these two findings. This paper discusses how and why they differ. In short, NPP's data is accurate, but its primary conclusion is misleading. 

The Conflict

On November 1, NPP announced the release of its online database with a press release headlined "Military recruiters enlist lower and middle income youth." NPP's analysis is based on Department of Defense data on Army, Navy, Air Force, and Army Reserve enlistees in FY 2004. According to the press release, "Lower and middle-income communities experience higher military enlistment rates than higher income areas." It also included a charged quote from the NPP's executive director: "As the Iraq War continues and the number of soldiers killed and wounded mounts, this data makes clear that low- and middle-income kids are paying the highest price."

The NPP analysis received extensive coverage, including a front-page headline and large chart in the Washington Post.[3] The Post accepted NPP's conclusions unchallenged:

[T]he military is leaning heavily for recruits on economically depressed, rural areas where youths' need for jobs may outweigh the risks of going to war.

All of the Army's top 20 counties for recruiting had lower-than-national median incomes, 12 had higher poverty rates, and 16 were non-metropolitan.

On November 7, the Heritage Foundation's Center for Data Analysis released its 21-page "Who Bears the Burden?" report on military recruitment. This report uses Department of Defense data on all active-duty enlistees in FY 1999 and the January to September period of 2003. The report's key finding is that "There are slightly higher proportions of recruits from the middle class and slightly lower proportions from low-income brackets," and its lead author (also the author of this paper) suggested that "Congress needs to remain steadfast in opposing coerced conscription and expose the myths of racial and class exploita­tion in military recruiting."

How could two analyses of similar data differ so widely in their conclusions? A look at the NPP study's underlying methodology answers this question.

Flawed Reasoning

Almost all of NPP's findings are based on an analysis of what it calls the "top 20 counties by recruitment rate," which forms the basis the Post's chart. NPP reasons that because these counties are rural and rural areas are poorer on average, recruits are therefore poorer than average, which suggests exploitation. This reasoning is flawed. Serious researchers and statistically literate journalists would not use a non-random sample of twenty counties to represent the state of the military.

Common sense suggests that a handful of counties with small populations will always have the highest recruitment rates. It only takes one enlistee in a county with only twenty equivalent youths to achieve the top ranking. Indeed, four of NPP's top twenty counties had just four enlistees, and the average per county was 14.[4] In all, NPP's top twenty counties accounted for just 275 recruits, less than two-tenths of one percent of all the recruits in 2004. By contrast, the zip code area 28314 in Fayetteville, North Carolina, was the home of 111 recruits in 2003, more than any other ZIP code in the nation.

NPP should contrast its data with similar data from either (1) the "bottom twenty" counties or (2) some other time period. If one considers the bottom 20 counties, the same "poor, rural" concentration will result. Counties with zero recruits will tend to have small, rural populations with lower-than-national-median incomes.

This is easy enough to confirm: there were over eight-hundred 5-digit areas with zero recruits in both 1999 and 2003, resulting in a tie for the "bottom" recruitment rate. Unsurprisingly, the average household income of these zip codes is lower than the national median.

When ranked in terms of total population, the absolute bottom zip code is 02215, an urban neighborhood that abuts the "Harvard Bridge" in Boston and had zero recruits. The average household income for the "bottom 20 zip codes" is $22,724. That is, the 20 largest U.S. neighborhoods with no enlistees have median incomes that are half the national median.

In contrast, the zip code with the highest recruitment rate is 78254, just outside of San Antonio, Texas. That neighborhood has a median household income of $76,000. This finding flies in the face of NPP's analysis.

A Comprehensive Approach

To fully understand the nature of military enlistment, one must take a comprehensive approach and employ statistics that include the entire recruit population. What makes the comprehensive data so impressive is that for every two recruits who came from the poorest 20 percent of neighborhoods in 2003, three came from the richest neighborhoods.

Several aspects of the NPP approach are commendable, particularly its searchable database of counties- and high schools-level recruitment. But the sad fact is that most of the NPP analysis is based on "top 20 county" figures that are not representative. Policymakers should be especially wary of the conclusions that the rural nature of the military is any different this year than previous years.

And finally, there is no logical dissonance that military recruits generally come heavily from both rural and wealthier neighborhoods. While rural areas tend to be poorer, not all are. More to the point, urban and suburban areas still provide four out of every five enlistees. Arguing that the military is disproportionately rural and therefore made up of individuals from low-income families is no better logic than arguing that the military is disproportionately black and therefore made up of individuals from low-income families. It is true that blacks make up a disproportionate share of military recruits, but the blacks that enlist tend to be better educated and from wealthier neighborhoods than comparable civilians. Only with a comprehensive analysis of the recruitment data are conclusions of this sort possible. On this, NPP falls short.

Tim Kane, Ph.D., is Bradley Research Fellow in Labor Policy in the Center for Data Analysis at The Heritage Foundation.