Barack Obama is playing to a variety of audiences while he travels abroad this week, with stops in Iraq and Afghanistan as well as Europe. One of them is an interesting voting group that could pack some surprises: Active-duty U.S. military.
Conventional wisdom holds that U.S. service members – including the 500,000 currently serving overseas – are a disproportionately Republican and conservative group. But that assumption is challenged by a unique survey of the U.S. Army done in 2004 by Maj. Jason Dempsey, then of West Point, and Prof. Robert Shapiro of Columbia University, via Columbia’s Institute for Social and Economic Research and Policy.
Their data show that the officer corps indeed is disproportionately conservative and Republican – but that enlisted service members, who make up the bulk of the population, are not. They’re essentially no more conservative, and no more apt to be Republicans, than the U.S. population as a whole. Fewer are Democrats; more, independents.
On ideology, while 63 percent of Army officers identified themselves as conservatives, only half as many enlisted members, 32 percent, said the same. The combined total, 38 percent, is very close to the aggregate for the U.S. public overall in ABC/Post polls this year, 34 percent. Twenty-one percent of all Army service members were liberals – again roughly matching the U.S. public overall.
The survey had to estimate partisan identification in a roundabout way. It asked respondents to place the Democratic and Republican parties on an ideological spectrum, then to place themselves on the same spectrum, then asked if they identified with one of the parties (but not which one). The answers were used to project party allegiance.
The result: Fifty-one percent of Army officers were identified as Republicans, but that fell to 23 percent of enlisted personnel. The net was 29 percent – again very close to the public overall, 27 percent in ABC/Post surveys this year.
There was a big difference in estimated Democratic allegiance: Only 11 or 12 percent of officers or enlisted service members were identified as Democrats. Instead 37 percent of officers, and a whopping 66 percent of the enlisted ranks, were independents, for a net total of 60 percent of U.S. Army personnel.
Independents, as it happens, are the quintessential swing voters in presidential elections.
Ideology Est. party ID Lib Mod Cons Dem Rep Ind Officers 14% 23 63 12% 51 37 Enlisted 23 45 32 11 23 66 All 21 41 38 11 29 60
Gen. pop. 23 40 34 37 27 31
(Army data from Jason Dempsey and Robert Shapiro, survey of U.S. Army personnel, Institute for Social and Economic Research and Policy, Columbia University. Gen. pop. data is 2008 ABC/Post aggregate.)
VOTE? – Another question is how many active-duty military actually vote. The Pentagon, which runs the Federal Voting Assistance Program aimed at encouraging turnout, commissioned a survey in 2005 in which 73 percent of uniformed military respondents reported voting in the 2004 election, compared with 57 percent in a 2000 study. (See the full report here.) If so, that’s better than the 60 percent turnout among all eligible voters in 2004.
But the reliability of that survey is in question. Dempsey and Shapiro’s Army survey produced a much-lower 43 percent turnout figure for 2000. (Full disclosure: Shapiro consults with ABC News on exit poll analysis.) Dempsey is skeptical of the FVAP figure for 2004; as he puts it, take a bunch of 18- to 24-year-olds, move them around every few years, and it’s hard to see three-quarters of them voting.
Moreover, the Government Accountability Office has raised questions about the FVAP survey, saying its "estimates and conclusions should be interpreted with caution" because of its response rate, which was low by GAO standards. Scott Wiedmann, deputy director of the FVAP, told me this week that his group agrees with the GAO criticism, avoids projecting the survey results beyond the individuals who participated (though the FVAP report doesn’t read that way to me), and is reworking the methodology to produce better data after this fall’s election. Wiedmann referred my detailed methodological questions to others at FVAP; see their reply here. (Aficionados will note the creative calculation of response rate, as well as the vote question that offers two yesses vs. one no.)
There’s also the question of overstatement of voting in polls, not just because of presumed social desirability bias, but also – and for my money, more so – because of civic-engagement bias – the fact that people who participate in polls are more civically engaged, and therefore are also more likely indeed to have voted. In any case, the FVAP reply to my questions goes so far as to predict a lower estimate for 2008, simply because they’ve tweaked their question wording to make it easier for respondents to say they didn’t vote.
The bottom line is that there’s plenty of room to debate both the size and direction of the military vote; Dempsey and Shapiro's data suggest you could drive a Humvee through the holes in the conventional wisdom on the subject. Dempsey will be reporting his full survey results in a forthcoming book on the social and political attitudes of the U.S. Army – meaning there may be more surprises yet to come.