POLITICAL GEOGRAPHY, MILITARY INSTALLATIONS, AND PARTISAN PREFERENCES

While some influential scholarship asserts that geographic “context should not count” in political analysis (King Reference King1996, 159; see also Jacoby Reference Jacoby2014), the notion that location shapes political preferences is well established across political geography. Lyons (Reference Lyons2017) shows that the pressures of location can affect partisanship more strongly than other critical factors, such as family partisanship. Furthermore, research consistently finds that geographic patterns and broader social-diffusion effects influence individual-level partisanship (Lyons Reference Lyons2011, Reference Lyons2017; Motyl et al. Reference Motyl, Iyer, Oishi, Trawalter and Nosek2014; Teigen, Shaw, and McKee Reference Teigen, Shaw and McKee2017).

Regarding proximity to military installations specifically, states with higher enlistment rates and richer “military culture” are less casualty averse (Gill and DeFronzo Reference Gill and DeFronzo2014, 727). Similarly, individuals who live in areas with proportionately large military veteran populations, have familial relationships with members of the armed services, or view their attachment to military veterans as a central part of their identity also hold distinct attitudes on key foreign policy issues (Gartner Reference Gartner2008, Reference Gartner2009; Krueger and Pedraza Reference Krueger and Pedraza2015; Lau, Brown, and Sears Reference Lau, Brown and Sears1978; Theys, Krueger, and Pedraza Reference Theys, Krueger and Pedraza2020). Thus, it seems reasonable to predict that living near a military base shapes one’s partisanship (Ambrosius Reference Ambrosius2016). As Chinni and Gimpel (Reference Chinni and Gimpel2010, 167) explain, “[A] 2009 Gallup survey found that current and former service members are more likely than civilian workers to be Republicans, and that partisan tie clearly has an effect in these communities—even among those not actually in the military. There’s a military culture in the Military Bastions that extends far beyond life on the military base itself.”

Indeed, Chinni and Gimpel (Reference Chinni and Gimpel2010) offer what likely is the most famous and cited argument regarding the influence of military installations on the politics of surrounding communities. In making their arguments, Chinni and Gimpel present a largely descriptive analysis of US counties, placing each county in one of 12 categories based on their chosen demographic characteristics, including average age, migration, religion, population growth or decline, income, and education. For 10 of these categories, locations are denoted as “monied burbs” or “Mormon outposts” (for example) based on the compositional characterization of who lives there. However, for two of their categories—“military bastions” and “campus and careers”—there is an implicit causal claim stemming from their reliance on non-compositional factors. For “military bastions,” Chinni and Gimpel (Reference Chinni and Gimpel2010) recognize that the defining characteristic of a county is more than a descriptive summary of county demographics; it is a recognition that they are making an implied contextual argument beyond the compositional one: that is, the presence of an armed forces installation.

Chinni and Gimpel (Reference Chinni and Gimpel2010) make the implicit causal case that there is something different about these counties apart from others similar to them in other respects. That is, there are other counties in Kentucky that share the demographics of Hopkinsville in Christian County that are not military bastions because they lack proximity to a military base. Thus, the current discourse predicts that people living near military bases will trend more Republican and, indeed, there is an underlying causal argument to this discourse.

Nevertheless, we argue for caution before accepting this assumption at face value. Conventional wisdom assumes that the military is overwhelmingly Republican. While that may have been true in the past, recent research shows that this tendency has not held true in the past decade. As Lupton and Margulies (Reference Lupton and Margulies2020) explain, “[P]olls of active-duty personnel show they are conservative, but not overwhelmingly so.” Changing demographics of the military at large—including an increase in the number of female, Black, and Latino personnel—suggest that the armed services may continue to become less conservative in the future (Shah Reference Shah2024). Thus, it should not be surprising that statistics from the 2020 presidential election indicate that military bases swung Democratic at a rate that outpaced the national average (The Economist 2021). Furthermore, when considering other key demographic factors, counties with military bases were no more or less likely to support Trump during the 2016 presidential election (Abascal and Centeno Reference Abascal and Centeno2017).

Conventional wisdom further assumes that military bases are similar in their geography and subsequent political influence. When most people think of a military base, they likely envision a large Southern base such as Fort Moore on a rural stretch of the Alabama and Georgia state line south of Columbus. Yet, this heuristic fails to appreciate the incredible geographic diversity of America’s domestic military installations. MacDill Air Force Base shares crowded space with two of Florida’s largest cities, Tampa and St. Petersburg. Similarly, the US Navy and US Marines operate multiple bases on the Southern California coast. More broadly, scholars increasingly push back against the notion that certain spaces are red while others are blue (Ambrosius Reference Ambrosius2016). Not all military personnel are Republican and not all military bases are rural. Accordingly, our counter-hypothesis indicates that there is no relationship between proximity to military bases and partisanship.

RESEARCH DESIGN

To assess this issue, we employ pooled cross sections of nationally representative samples from the 2006–2022 Cooperative Election Study (CES) surveys. The CES includes respondents’ zip codes and relevant survey questions—namely, partisanship and presidential vote choice with appropriate control variables. For our key independent variable of interest, we measure how much distance exists between each respondent’s zip code and the nearest military installation that houses active-duty forces. The stateside array of military bases includes individual branch installations as well as multi-branch joint bases. The reasons behind their locations are a complex and historical array of national security needs, political deal making, and other factors. From our perspective, it is not important to focus on why bases are located where they are as much as measuring their location—and the proximity of individual survey respondents to these locations.

To identify major military installations across the United States, we need to determine which installations to include and where those bases are located. We begin by working with Adler’s (Reference Adlern.d.) classification of “major” military installations listed in the Congressional District Data Codebook. It includes any installations that are named “Fort,” “Base,” “Camp,” “Air Force Base,” “Naval Base,” or “Submarine Base.” This classification, however, omits key major military installations, such as those designated as “Joint Base,” “Naval Station,” and service academies. We then cross-reference this list with the Department of Defense (DOD) list of military installations to identify which bases to include in our data. We exclude installations that have scant or no active-duty personnel units stationed at those locations. We believe that focusing on such “major” military installations best tests the spirit of Chinni and Gimpel’s (Reference Chinni and Gimpel2010) argument. We identify 113 bases that meet these criteria. We then specify the location of each base, using information provided by the DOD (Lupton and Teigen Reference Lupton and Teigen2025).

The way we geolocate respondents and their proximity to military installations is more straightforward. CES data provides each respondent’s zip code, which has two desirable attributes in terms of analytic utility: zip codes (1) are geographically small, providing as high geographic precision as possible with nationally representative survey research; and (2) tend to be residentially homogeneous.

We employ three similar but distinct ways to measure and operationalize a respondent’s geographic location vis-à-vis military installations. First, we use an aggregated model that includes all respondents and the Euclidean distance to their nearest base. Second, we zoom in on each military installation and run a model for only those respondents within 60 miles of each base. Third, we calculate total proximity to bases such that respondents who reside near multiple bases would have high scores, those near only one or two would have moderate scores, and those not near any bases would have low scores.

Our two dependent variables are Republican Party identity and voting for Republican presidential candidates. Party identity is coded as a dummy rendered from the seven-point party identity scale with “strong,” “not very strong,” and “lean” Republican identifiers receiving a 1, and we code independents, Democratic leaners, and Democratic identifiers as zero (Keith et al. Reference Keith, Magleby, Nelson, Orr, Westlye and Wolfinger1986). We track vote choice as a dummy variable, coding 1 for voters who chose John McCain, Mitt Romney, or Donald Trump and zero for those who chose Barack Obama, Hillary Clinton, or Joe Biden. We omit nonvoters or those who voted for minor-party candidates.

Our individual-level control variables include several well-understood correlates of political preferences—namely, family income ($70,000+ per year); race and ethnicity (white non-Hispanic versus other identities); education (bachelor’s degree or more); gender (male dummy); religion (Evangelical Christian association); age (bracketed into dummies); marital status (married living with spouse); military service experience (veteran dummy); year; and state. Year of survey and state of residence variables each omit one arbitrary category (i.e., 2006 and Alabama).

We include three additional compositional variables measured at the zip code level to control for space apart from military base proximity: (1) the percentage of the respondent’s zip code that comprises African Americans; (2) the percentage of the respondent’s zip code that comprises Hispanics; and (3) population-density quartiles. This inclusion controls for an important factor about many military bases. Although some stereotypical bases (e.g., Fort Campbell) exist in a milieu that often is thought of when imagining a rural base in a red state, other urban and diverse areas also have bases. San Antonio, San Diego, Los Angeles, and Washington, DC, are examples of locations with high racial and ethnic diversity, population density, and military installations. Finally, we include year- and state-fixed effects in each model.