I am thrilled to see the number of posts that have been generated by my original question regarding whether or not Romney is likely to have a military problem (see here, here, and here on The Monkey Cage and here at The New Republic). I think this is one of the best features of the Monkey Cage: that we can raise a topic for discussion and due to the incredible expertise of our readers, within the next few days we can essentially have a forum on the issue. With that in mind, here is one more comment from Major Jim Golby, an Instructor in the Department of Social Science at the United States Military Academy at West Point and a Stanford Ph.D.
A number of people have noted that military officers have different opinions than enlisted soldiers; I agree. Officers also are older, whiter, richer, and maler than are enlisted soldiers. Consequently, I think this distinction misses the main point on the issue of military voting. In his initial post, Josh asked, “Does Mitt Romney have a Military Problem?” The answer to that question is no; there is no such thing as the military vote, be it officer or enlisted.
My argument is that, once we know a person’s party ID and standard demographic characteristics – race, gender, age, income, etc. – we gain no new information about how someone will vote by knowing that she is enlisted or an officer or a veteran. Jeremy Teigen’s paper demonstrates this point for veterans in the 2004 election; mine does the same for officer opinions. I actually use the TISS data that Paul Gronke mentioned in his earlier comments, but I reach different conclusions than the authors presented in the original book. I control for Party ID, which the original authors omitted. I find that controlling for Party ID wipes out the ‘civil-military gap’ finding, though I do find that military officers who identify as Democrats usually are moderate Democrats. I present some evidence suggesting that liberals enter the military at low rates and leave at relatively high-rates after their initial commitment; I am not able to identify why liberal leave at higher rates.
The evidence in Jason Dempsey’s book, Our Army, also underscores this point for soldiers. Dempsey compares what he calls the ‘NAES Virtual Army’ – a group of civilian respondents weighted to resemble the army or the officer corps on relevant demographic characteristics – to the general Army population and the officer corps. Once he controls for demographic factors, the evidence shows few differences between civilians, enlisted soldiers, and the officer corps. For example, see Dempsey’s table below on political ideology:
There is evidence, however, that there are fewer Democrats in the Army. The ranks may not be as tilted toward the GOP as some may believe, but there is evidence that it might be titled away from the Democratic Party. Although the percentage of Republican identifiers in the military is similar to that of the general population, the percentage of Democrats is substantially lower in the Army (C&S in the table) than among the NAES general population (without demographic weights). According to Dempsey’s estimation, only 11 percent of the Army identifies with the Democratic Party compared to 33 percent of the civilian population (in the 2004 NAES).
This issue is not an officer-enlisted issue. Democratic ID among junior enlisted soldiers and NCOs is 9 percent and 12 percent, respectively, compared to 19 percent among lieutenants and 11 percent among colonels. We need to treat all of Dempsey’s data on Party ID with some caution, however. Because of the restrictions on his survey, he did not ask Party ID. Instead, he estimated soldiers Party ID using an algorithm. Nevertheless, the more accurate understanding that Dempsey’s data gives may help change some misperceptions about the representativeness of our nation’s military.
In short, there is not much evidence that any of this matters for voter behavior. Paul Gronke is correct that there is not as much data about service members as we would like. Nevertheless, based on the evidence we have, veterans and members of the military are likely to vote for the GOP at slightly higher rates than civilians, but not more than civilians who share the same demographic characteristics that they do. Overall, there is very little evidence that there is a military voting bloc.