Polarization versus professionalism: military and civilian views on the domestic use of the military

Abstract

Scholars and policymakers warn that with rising affective polarization, politicians will find support from the public and permission from military professionals to use military force to selectively crack down on political opponents. We test these claims by conducting parallel survey experiments among the US public and mid-career military officers. We ask about two hypothetical scenarios of domestic partisan unrest, randomly assigning the partisan identity of protesters. Surprisingly, we find widespread public support for deploying the military and no significant partisanship effects. Meanwhile, military officers were very resistant to deploying the military, with nearly 75 percent opposed in any scenario. In short, there is little evidence that public polarization threatens to escalate domestic disputes, and strong evidence for military opposition.

1. Introduction

I have long viewed domestic division as our greatest national security vulnerability. Political polarization is a “force multiplier” that worsens other threats and cripples our ability to combat them.

— Amb. Susan Rice¹

We do not need to militarize our response to protests. We need to unite around a common purpose.

— Gen. (Ret.) James Mattis²

In recent years, scholars and national security officials have warned that political polarization poses a growing threat to US national security, particularly during periods of domestic unrest. During the summer of 2020, the Trump administration deployed federal agents to suppress progressive protesters in Portland and Washington. In response, many prominent national security officials broke ranks to express their alarm. Less than a year later, many of the same voices criticized the White House and the Defense Department for their reticence to deploy the national guard to defend the US Capitol from pro-Trump insurrectionists. Notably, these voices sought to convince military leaders and political supporters of President Trump to resist partisanship and oppose the president's actions.

The proper role of the military in the face of domestic unrest raises two long-standing concerns (Bennett, 2006; Head and Mann, 2016). First, some worry that incumbent leaders will succumb to the temptation to selectively deploy military forces against domestic opponents. When incumbents are free to deploy (or withhold) military forces into domestic unrest—cheered on by supporters and unchecked by military professionals—it gives the government a veto on security and free expression (Kohn, 2002). Second, others worry that the militarization of civilian policing itself is a threat to democratic governance. Employing “emergency” powers in “normal” provision of public safety, they argue, undermines trust in democratic institutions and misdirects key national security resources from foreign to domestic threats (Pion-Berlin and Arceneaux, 2000; Kuehn and Levy, 2021; Jenne and Martínez, 2022).

Despite the high normative stakes of these questions, there is little systematic evidence on support for military involvement into scenes of domestic unrest. When do American voters and military professionals support domestic deployments of the military? In this paper, we ask two main questions. First, we investigate whether out-group partisanship generates public support for misdirecting the military against domestic opponents. Do American voters favor deploying the military into scenes of domestic unrest, and to what extent do partisan considerations shape this support? Second, we examine whether professional norms constitute a bulwark against military participation in domestic unrest (Brooks, 2020; Brooks and Grewal, 2022). Do military professionals object to deploying the military for domestic purposes in general? To what extent are members of the US armed forces immune from partisan influence?

We present the first systematic survey evidence comparing public and military attitudes toward domestic uses of the armed forces. We conduct two parallel sets of survey experiments investigating when American citizens and military officers support deploying the military into scenes of domestic unrest. The first survey includes a nationally representative sample of over 800 US voters, recruited online by Qualtrics. The second survey includes an elite sample of 137 mid-career military officers studying at the US Naval War College. The officers at USNWC are broadly representative of leadership-track officers, representing all six branches of service and a wide variety of specialties and backgrounds. Deploying a combination of direct attitude questions and survey experiments, we investigate when and why each group favors deploying the military into scenes of domestic unrest.

The centerpiece of these surveys is a pair of vignette experiments posing situations similar to the 2020 Portland protests and 2021 Capitol insurrection, with the partisan identity of the protesters randomly assigned. The survey then asks about several potential public responses, including deploying military personnel and militarizing police. The primary purpose is to measure overall support for deploying the military and the degree to which partisan valence affects this support.

Contrary to our expectations, we find relatively little partisan bias among the public and strong professional norms among the military. First, public attitudes toward military deployment in domestic unrest are surprisingly durable and nonpartisan. Although the majority of the public is willing to support militarized responses to domestic protests, they are just as likely to favor an armed response against copartisans as outpartisans. Our findings run counter to recent scholarship showing that voters’ attitudes toward many other political norms, even those surrounding national security, are fragile when posed against partisan calculations (Tomz and Weeks, 2020; Krebs et al., 2021). Despite widespread concerns about political violence and partisan extremism, the public shows a surprising consensus in favor of law and order.

Second, military personnel hold very strong professional norms against military participation in partisan politics. In our survey, the vast majority of officers opposed the military being deployed into domestic unrest regardless of the circumstances. These differences appear to be driven by officers’ personal and professional commitments to military nonpartisanship. Military officers were supportive of armed responses to protest, but only under police auspices. Meanwhile, these views were most strongly held among longer-tenured officers, and were not shared by veterans or military family members. Our findings confirm claims that military officers take organizational norms seriously, even when those norms are out-of-step with public attitudes.

2. Polarization and professionalism in military interventions

With rising partisan polarization in the US, national security scholars and policymakers alike have warned about the potential consequences during periods of domestic unrest. Two normative concerns stand at the forefront.

The first concern surrounds partisan bias: leaders might selectively deploy military force against domestic political opponents. When policymakers view political opponents with suspicion and enmity, they may prioritize domestic political rivalries over foreign threats. More concerning, partisan supporters in the public may permit or even encourage politicians to target opponents. While it is certainly possible for national security agencies to take these actions on their own, it has historically been more common for civilian leaders and the public to “pull” them into roles beyond their traditional scope, particularly in democracies with a robust tradition of civilian control of the military (Harig and Ruffa, 2022, 88).

Using the military selectively to target opponents poses a threat to democratic governance. At the extreme end, the military may play a decisive role in transition from democracy to autocracy and vice versa (Nepstad, 2013; Singh, 2014). Even in less extreme situations, however, domestic military deployments introduce state force into partisan speech—particularly when the military is selectively deployed based on partisan incentives. Even when deploying the military is necessary to enforce the rule of law, it can still damage democratic norms in the long run. As Former Defense Secretary and General James Mattis wrote in 2020, domestic military deployment “erodes the moral ground that ensures a trusted bond between men and women in uniform and the society they are sworn to protect, and of which they themselves are a part. Keeping public order rests with civilian state and local leaders who best understand their communities and are answerable to them.”³

When political leaders use national security institutions for partisan purposes, moreover, it undermines public trust in those institutions among both the public and elites and threatens national security (Lythgoe, 2022). Opposition supporters may view institutions as tools of their partisan opponents, while regime supporters are encouraged to view opponents as security threats. During times of high polarization, this may accelerate growing distrust of political leaders, professional experts, and public institutions (Levi and Stoker, 2000; Iyengar et al., 2012; Abramowitz, 2018; Levendusky, 2018; Krebs et al., 2021).

The second concern surrounds militarization: deploying the military for domestic policing is itself corrosive to liberal democratic norms, even if done without partisan bias. Indeed, the very process of legitimizing the deployment of military force to address domestic threats has material and social impacts on civil-military relations and domestic politics (Laclau and Mouffe, 2014; Kuehn and Levy, 2021). While the material impact of militarization (e.g., risk of coup or military autocracy) receives priority in much of the existing literature, militarization's discursive power undermines democratic governance by addressing a “narrative that legitimizes and justifies the use of force in areas of security that are not the domain of the military” (Martínez and Bueno, 2023). In other words, militarization represents a longer-term process through which elites attempt to persuade the public that domestic deployment of the military is a “normal” tool of everyday politics. This is particularly relevant for more expansive uses of the military in domestic policing (Pion-Berlin and Arceneaux, 2000).

Militarization often emerges through securitization, or the process through which policymakers transform a policy issue into a national security concern by reframing the issue through the lens of existential threats (Waever, 1995; Buzan et al., 1998; Huysmans, 1998; Williams, 2003). Understood through the lens of securitization, increasing bipartisan support for the domestic deployment of the military can best be understood as a broader shift toward a permanent “state of emergency” in domestic politics (Schmitt, 2005, 2007, 2008). “[T]he military's internal use is justifiable on practical grounds, highlighting the lack of civilian state capacity, societal demands for security and basic services, as well as the armed forces’ organisational strength and logistics capacity,” explain Jenne and Martínez (2022, 59). As this process plays out, “societies begin to normalise the extraordinary, the state ‘forgets’ to build alternative capacities, and the military integrates new operational experiences into its organisational structure and role conception,” note Jenne and Martínez (2022, 59).

Considered in concert, the normative concerns of selectivity and militarization raise the stakes of public and military views on military action during domestic unrest, particularly in a highly polarized environment.

2.1 Partisanship and public attitudes

In highly polarized environments, public support for domestic use of the military may be highly biased by domestic political considerations. Of particular concern is affective polarization: the degree to which individuals distrust and dislike members of an opposing political party. In societies with high affective polarization, individuals typically regard members of opposing groups—both elites and ordinary members—with suspicion, contempt, or disgust (Linville and Jones, 1980; Tajfel and Turner, 1982; Kite and Whitley, 2016; Campbell, 2018; Kingzette, 2021).

In the past decade, scholars have documented rising levels of affective polarization along partisan lines, sometimes referred to as “negative partisanship” (Iyengar et al., 2012; Levendusky, 2018; Abramowitz, 2018; Druckman et al., 2021). For example, Boxell et al. (2022) find that American voters in 2020 on average rated in-party members 56 points higher than out-party members on a “feeling thermometer” ranging from 0 to 100—more than double the gap in 1978. This trend has coincided with growing partisan polarization on other dimensions: increasing ideological distance, separation of informational ecosystems, and increasing overlap of social and identity cleavages. Both among voters and party elites, this pattern has been asymmetric, with Republicans adopting more anti-Democratic attitudes than vice versa (Butler, 2009; Coppock et al., 2023).

Affective polarization shapes how voters and policymakers respond to domestic rivalries and security threats. Prior work has shown that American voters’ fear and contempt of outpartisans increasingly leads them to favor disparate responses based on partisanship. Voters judge outpartisans more harshly for the same behavior and trust their intentions less than copartisans, even outside of political contexts (Iyengar and Westwood, 2015). This should apply even in the context of physical security: recent studies have shown that voters and politicians both view security threats through partisan lenses, blaming outpartisans while absolving copartisans (Myrick, 2021). Even among national security professionals in ostensibly nonpartisan environments, distrust of domestic political opponents makes it more difficult to maintain common goals and adapt to new challenges under uncertainty (Schultz, 2017; Saunders, 2022). In a highly polarized environment, then, voters should view outpartisans as a greater threat to law and order than copartisans who behave similarly, and therefore favor harsher responses.

What should we expect to observe if partisanship shapes public attitudes toward domestic use of the military? If this is the case, American voters should be more likely to support deploying the military when such an intervention targets outpartisans rather than copartisans. That is, Democrats should be more likely to favor deploying the military against Republican protesters than Democratic protesters, and vice versa for Republicans. When domestic policing takes place in a polarized political climate, individuals often see it as inherently partisan, taken on behalf of either the in-group or out-group. Individuals may see action against copartisans as draconian overreach and action against outpartisans as restoring the law and order.

H1: Civilians should be more likely to favor domestic military action to police members of the opposing party than of their own party.

The same processes also suggest that partisan trust of military and police may bias support as well. If one party generally has a more trust in the military—as the Republican Party traditionally has in the United States—they may be more supportive of using domestic deployment, believing military action is likely to support their political preferences. The same may hold for other responses to domestic unrest, like militarizing police or deputizing civilian militias, as both groups are traditionally considered Republican-aligned in the United States.

2.2 Military professionalism as a bulwark

Many have argued that military professionalism—norms about how the military interacts with civilian policymakers and the general public—can act as a bulwark against partisan bias and domestic militarization (Farrell, 2005; Brooks et al., 2022; Brooks and Erickson, 2022). In the American context, professional military norms favor civilian control of the military and a strict division of responsibilities between civilian and military decision-makers (Brooks, 2020; Feaver, 2023; Blankshain et al., 2024). Rather than engaging as an unbiased arbiter, the US military generally avoids any involvement in partisan politics.

In the case of domestic unrest, the relevant norm is that of non-interference, which holds that military forces ought not be deployed to address domestic political unrest, particularly not in support of a partisan agenda (Huntington, 1957; Blankshain, 2020; Brooks et al., 2022). Given the sheer capacity of the US military, its roles are better reserved for protecting US peace and prosperity “beyond the water's edge.” Within US borders, less powerful state and local law enforcement ought to contain domestic unrest. Even if political incumbents and the public would like the military to intervene in a bipartisan fashion, military leaders are urged to stay out entirely.

Although military leaders often present non-interference as a legal constraint, it is primarily a normative one. The governing statutes—most notably the Posse Comitatus Act of 1878—prohibit deploying the military within the territorial United States only when it is not explicitly permitted by other legislation. The President, therefore, has wide authority to do so by invoking a special law (e.g., the 1807 Insurrection Act), by declaring a state of emergency, or at the request of a state governor (Cohn, 2022; Jenne and Martínez, 2022). In the decades following Posse Comitatus, the military was deployed dozens of times to break strikes and contain protests. Only after the rapid modernization and institutionalization of the national security state in the 1940s did the Eisenhower administrations clearly delineate the military's roles from that of local and state forces (Leffler, 1992; Gaddis, 2005). Even in the postwar period, however, the military has been deployed on American soil on numerous occasions—most famously to Little Rock Central High in 1957 and to South-Central Los Angeles in 1992 (Coakley, 2005; Schieps, 2005).

Instead, the norm of non-interference is a form of institutional self-restraint, which “requires individual responsibility and accountability for moral agency, dedication to duty/public service, and commitment to the greater good of the society the institution serves” (Novy, 2017). In democratic contexts, military professionalism particularly values “the subordination of the military to democratic civilian authority, allegiance to the state and a commitment to political neutrality, and an ethical institutional culture” (Ouédraogo, 2014). The logic outlined above, however, reverses the conventional wisdom in bureaucratic politics: intuitions face incentivizes to expand their portfolio of responsibilities and capabilities in order to increase their political influence in the wider bureaucracy (Drezner, 2000). It is more consistent, instead, with the neutrality outlined in professional ethics of other types of civil servants, particularly outside the United States (Pion-Berlin and Martínez, 2017).

Why might military personnel hew to professional norms even (and especially) when elites or the public demand aggressive action? First, military personnel might simply be more knowledgeable about military capabilities and norms. In their daily experience in the ranks, military personnel quickly become versed in the practical challenges of military operations, which may shift attitudes on normative issues. Even early in their service, military personnel receive instruction on institutional norms on which most members of the public are not well-versed. This instruction, notably, need not be based in fact; for example, the authors’ anecdotal experience with military students suggests that military instruction on domestic non-interference often blurs the distinction between normative and legal constraints.

Second, differences in normative beliefs may also derive from different professional and personal commitments built over one's career. Over the course of military careers, institutional norms are internalized and built through training and countless interactions between service members at home and abroad (Joyce, 2022; Tecott et al., 2022). In the US military, most officers who reach ranks where operational decision-making is common (particularly O-6: Army/Air Force Colonel or Navy Captain) take nearly 20 years to do so. As a result, officers typically have nearly two decades to internalize the norm of non-interference before they might be placed in a position to violate it. This socialization process is likely to be strengthened by selection and attrition. Individuals who self-select into the military are likely to be predisposed to internalizing norms related to public service, esprit de corps, and respect for the military. Likewise, the officers who remain in uniform and receive promotions are likely to be those who have accepted and internalize institutional norms, while those who disagree with these norms are likely to retire or remain at lower ranks. Importantly, these professional and personal commitments are likely to be stickier than differences in knowledge. While information can be corrected or shared relatively quickly, professional commitments are likely to be built over a long period of time and more robust to new information.

If the norm of non-inference strongly shapes normative attitudes, we should expect military professionals to hold different views than the public. In particular, we should expect military officers to be less supportive of the domestic use of military force than members of the public.

H2: Military officers—regardless of partisanship—should be less likely to support the domestic use of the military than civilians.

The mechanisms explored above also suggest differences between subgroups. If greater knowledge of the military grounds such differences, then veterans and individuals with friends or family in the military should also be less supportive of domestic military intervention. If socialization to professional norms grounds subgroup differences, more experienced officers should be more likely to oppose domestic military deployments than their junior colleagues.

3. Research design

We test our theory by enumerating two online surveys.⁴ The first survey, enumerated among a representative sample of 800 registered voters in the continental US, examines civilian beliefs and attitudes toward civilian oversight and military participation. The second survey enumerates a matching questionnaire to military officers studying at the US Naval War College.

3.1 Survey methods and sample

The civilian survey was enumerated in March 2022 among an online panel of 800 US voters. Qualtrics recruited the panel based on gender, age, and educational characteristics representative of the wider population of American results. The survey was designed to take approximately 15 min and respondents were provided with a small monetary incentive (less than $5) to participate. Tables A1–A2 in SI Appendix offer an overview of descriptive statistics.

The military survey was enumerated in May and September 2022 among an elite sample of American military officers assigned to a one-year JPME (Joint Professional Military Education) course at the US Naval War College. The military survey was distributed by college administrative personnel. Participation was entirely voluntary, without any monetary incentives or additional encouragement from superior officers or civilian course instructors. In all, approximately 20 percent of students enrolled in the 2021–22 and 2022–23 cohorts responded to the survey. At the request of institution officials (to preserve anonymity), we did not ask about officers’ demographic characteristics such as age, race, or education. We did ask about service branch and length of service.

Data from the institution, however, suggest that USNWC students are quite representative of mid-career, leadership-track military officers. Nearly every leadership-track officer attends a JPME institution at some point in their career, regardless of branch or specialty. The Naval War College, moreover, draws from an unusually wide variety of military professionals. Our survey includes not only officers in the US Navy (about 40 percent), but also from the US Army (about 25 percent) and the other four services, including the newly formed US Space Corps. Because the institution includes both intermediate-level and senior-level courses, its students range from O-3 (an Army/AF/Marine Captain or a Navy Lieutenant) to O-6 (an Army/AF/Marine Colonel or a Navy Captain) and have served as few as 5 or as many as 25 years in their respective services. Most have had combat deployments, and many have had several. Students in JPME courses, like military officers more generally, are disproportionately male (80–90 percent) and highly educated; essentially 100 percent have a Bachelor's Degree and many have Master's Degrees. The officers in our survey are similar to the US population in terms of partisanship, though Republicans are slightly over-represented relative to the general population.

Aside from demographic questions, the contents of the two surveys were identical. The first section asked a few basic political views—including, importantly, partisan identification. The second section asked a host of opinions on the proper role of the military in politics. The third posed the two hypothetical vignettes that are the primary focus of this paper. The survey was constructed to reduce social desirability bias, particularly among the military professionals. The solicitation and instructions assured respondents that their participation and responses could not be attributed to them and that there were no “correct” answers. There is no evidence of systematic attrition based on respondent-level traits.

3.2 Vignettes on domestic military action

The centerpiece of the survey included two hypothetical vignettes (summarized in Table 1) posing plausible scenarios in which the military might be called into partisan domestic unrest. These scenarios are modeled on the two highest-profile cases in recent American political history: the January 6 Capitol Attack and the Portland riots of Summer 2020.⁵

Table 1. Summary of vignettes

The first vignette (Capitol Protest) is modeled after the January 6, 2021 Capitol Attack. After a closely contested 2028 Presidential Election, the party of the losing candidate alleges widespread fraud with some support from media watchdogs. A large crowd of protesters supporting the losing candidate surrounds the Wisconsin State Capitol in an attempt to prevent the certification of the decisive electoral votes. The crowd is said to be becoming increasingly aggressive, threatening to harm both Capitol Police and the state officials they are protecting.

The second vignette (Downtown Riot) is modeled after the unrest following Black Lives Matter protests in Portland during Summer 2020. A statue-related protest goes awry when a police officer shoots a protester, unleashing a week of escalating protests against partisan police conduct. Amid the protests, protesters have burned buildings and injured bystanders and police fear more violence ahead.

The scenarios have two features in common that are critical to interpreting their results. First, in each case, deploying uniformed military would likely be legal under US law. As noted earlier, the Posse Comitatus Act forbids the military to be deployed for domestic policing only in some cases, and only when state officials object to the deployment. The first scenario is legally clearer: local law enforcement officials are overwhelmed and the governor him/herself is under threat, so there is every reason to think the governor is requesting aid. In any case, given that the voting process itself is under threat, Section 253 of the Insurrection Act could almost certainly be invoked. The second case is less clear, but given that local law enforcement is requesting aid, the intervention would be legal merely with approval from the state legislature or governor.

Second, each scenario nevertheless puts law enforcement in the normatively difficult position of intervening on one side of partisan unrest. The vignettes and questions explicitly state that the deployments would be intended “to protect the State Capitol [/ downtown neighborhood] from the protesters.” Note that this is not an impartial mission meant to calm the violence or to provide security for state officials, bystanders, and protesters alike. Instead, supporting military deployment means curbing protesters with a particular partisan agenda and potentially arresting many of them. Any force that intervenes could be perceived as acting in favor of the opposing party's political agenda. This is precisely the sort of intervention, in other words, that the norm of military non-interference is intended to prevent.

The vignettes then ask whether respondents would support three potential policies for containing the protests. The first and most important question asks whether respondents would support the (hypothetical future) president deploying, with the governor's permission, a unit of US Army's 101st Airborne Division (who were famously deployed to Little Rock Central High School in 1957) to stop the protesters. This outcome directly tests the norm of military non-interference into partisan unrest. The second and third questions, meanwhile, pose two policies that are pro-“law and order, ” but do not violate norms separating the roles of military, local law enforcement, and partisan activists. The second question asks whether respondents would support arming police with lethal, “military-style” equipment, such as assault rifles and body armor—as opposed to only nonlethal equipment, such as riot shields and tear gas, or no additional equipment at all. The third asks whether respondents would support the governor encouraging armed “neighborhood watch groups” to assist law enforcement in containing the protesters. This phrase is intentionally ambiguous on the precise identity of the armed groups, but they presumably could include either radical militias (e.g., the Proud Boys) or local amateurs.

In each scenario, we randomly assign the partisan identity of the protesters. In the Capitol Protest vignette, the protesters could be Republicans objecting to a Democratic state administration calling the election for the Democratic Nominee, or Democrats objecting to a Republican state administration calling the election for the Republican Nominee. In the Downtown Riot vignette, the protesters could be Democrats demanding the removal of a statue or Republicans objecting to the removal of a statue. This treatment is intended to test whether antipathy toward outpartisans biases support for military action.

Therefore, we analyze treatment effects in the context of respondents’ partisan identification, comparing support for military action when protesters are of the same party versus the opposite party of respondents.

Given that the treatment was randomly assigned, there should be no correlation between this copartisan/outpartisan treatment and respondents’ partisan identity. In our main analysis of partisanship, we include independents who say they lean toward one party, as previous research in American politics shows that these “leaners” tend to be as reliable partisan voters as self-identifying partisans. Indeed, we find little difference in the preferences or behavior of “leaners” and partisans in our survey.

4. Polarization and public support for military policing

Contrary to our expectations, we find very little evidence that partisanship substantially influences attitudes toward military interventions in domestic unrest. Instead, we find American voters are quite supportive of the military intervening in domestic unrest regardless of the partisan consequences.

Figure 1 visualizes respondents’ attitudes when asked whether they would agree with deploying the military in each of the two fictional scenarios. In each vignette, respondents were remarkably supportive of the military being deployed to quell the protesters regardless of partisan identification. In the first vignette, 61 percent of respondents somewhat or strongly agreed with deploying the military against protesters at the Wisconsin State Capitol against 24 percent who somewhat or strongly disagreed. In the second vignette, 72 percent of respondents supported deploying the military against protesters in Downtown Pittsburgh against just 15 percent opposed. Seventy-nine percent of respondents supported deploying the army in at least one of the two scenarios, and just 11 percent opposed deploying the army in both cases. Respondents of both parties were significantly more likely to support deploying the Army than heavily arming the police (47 and 60 percent, respectively, in the two vignettes) or deputizing armed volunteers (38 and 37). This is consistent with respondents’ relatively high trust in the military: 73 percent of Republicans and 68 percent of Democrats said they trust the military “most of the time” or “just about always” (Figure 2).

Figure 1. Support for military deployment, by respondents’ and protesters’ party.

Most striking, however, was how little effect partisanship has on support for deploying the military. First, contrary to Hypothesis 1, respondents’ support for deploying the Army was not significantly affected by the partisan valence of the scenario. In each vignette, we randomized the partisan identity of the protesters: whether the Republican had won the election and Democratic protesters were threatening to breach the state capitol, or vice versa, in the first vignette; and whether the out-of-control crowd of protesters were Republicans or Democrats in the second vignette.

Figure 2. Trust in government institutions by party.

Figure 3 summarizes how the identity of protesters affected respondents’ support for deploying the military (as well as militarizing the police and deploying armed volunteers). In the first Capitol Protest vignette, there was a very small partisan effect. When the protesters seeking to overturn the election were from the opposite party as the respondent—the respondent was a Democrat and the protesters were Republican, or vice versa—64 percent of respondents supported deploying the Army. However, when the protesters were from the same party, nearly the same proportion of respondents, 60 percent, supported doing so. Even when using the more fine-grained (5-point scale) versions of the responses, the treatment effect (i.e., the interaction effect between the treatment and respondents’ partisan identities) was barely statistically significant at the ${\rm p}< 0.05$ level and only at the ${\rm p}< 0.10$ level with demographic controls (see Appendix Figure A2). In the second Downtown Riot scenario, there was virtually no difference at all: 73 percent favored deploying the Army regardless of whether the protesters were from the same or opposite party, and there was essentially no observable partisanship difference on the 5-point scale.

Figure 3. Support for government actions, by treatment (US Public Survey).

Second, we find very little evidence that respondents responded to the vignettes based on their partisan context. We deliberately wrote these vignettes to evoke two recent events with high partisan salience: the January 6 protest at the US Capitol and the Black Lives Matter protests in Portland. Given this context, we should expect Democrats to be more critical of protesters—and more willing to call out the Army—in the first scenario (Capitol Protest), which mirrored events with Republican protesters. Likewise, we should expect Republicans to be more critical of protesters and supportive of military intervention in the second scenario (Downtown Riot). Figure 4 tests this expectation by assessing overall support for deploying the military (and other outcomes) among Democrats and Republicans. Though Democrats were slightly more likely to favor intervention in the first case (66–57 percent) and Republicans in the second (74–72 percent), these effects were small and inconsistent across vignettes and model specifications.

Figure 4. Support for government actions, by party (US Public Survey).

We find these results surprising given that the vignettes were specifically designed to amplify partisan stakes. In the first scenario, a presidential election hangs in the balance, with accusations of fraud and a credible threat of violence. Yet support for deploying the military (and, to a slightly lesser extent, militarizing police and deputizing armed volunteers) was generally bipartisan and unaffected by protesters’ ideological worldview. Moreover, respondents understood the partisan stakes. More than 90 percent of respondents correctly identified the party of protesters in attention check questions (Appendix Figure A4).

The results are also remarkably consistent across subgroups. First, even among respondents who identify as strong partisans (i.e., “strong Democrat” or “strong Republican”), treatment effects are relatively small and statistically insignificant (Appendix Figure A5). There is little evidence, in other words, that even political extremists are particularly biased in their support for deploying the military. Second, trust in the military appeared to have very little correlation with the effect of partisanship (Appendix Figure A3). While low-trust military respondents were less likely to favor deploying the military into partisan unrest, they were no more (or less) biased against outpartisans in doing so. Third, the results are very similar when looking at the two other outcome questions: militarizing the police and deputizing armed volunteers to contain protesters (Figure 3). In both scenarios, respondents are only slightly more likely to support militarizing police or deputizing militias against protesters of the opposing party than they are against protesters of their own party.

Taken together, these results suggest that the majority of Americans hold durable, consistent beliefs supporting the domestic use of military force. It also suggests that the partisan-focused coverage of the January 6 Capitol Insurrection and the 2020 Portland Protests might have obscured a broader bipartisan consensus toward law and order. Respondents were comfortable with the military being used to restore domestic order, regardless of the partisan implications. They were also broadly comfortable with militarizing the police and, to a lesser extent, calling upon armed civilians to prevent disorder. Our findings are consistent with recent work showing that, when measured correctly, support for political violence is far lower than many commentators fear (Lelkes and Westwood, 2017), and that affective polarization may have less effect on support for democratic norms than prior scholarship suggested (Broockman et al., 2022).

5. Military professionalism and the norm of non-interference

Meanwhile, we find strong evidence that professional norms have a significant effect on attitudes regarding the domestic deployment of military force. Consistent with Hypothesis 2, military officers are much less supportive of deploying the military into situations of domestic partisan unrest than are the public.

Figure 5 shows support among the civilian and military samples for deploying the Army to quell the two fictional sets of protests. While large majorities of the civilians favor deploying the military in both scenarios, large majorities of military professionals oppose doing so. Combining the two scenarios, less than 25 percent of the military officers agreed with deploying the Army to quell the protesters, as opposed to more than 65 percent of civilian respondents. More than 75 percent of military officers said that they disagreed, with more than 50 percent answering that they “strongly disagreed” with doing so, as opposed to 19 and 10 percent among the public. These 40+ percentage point differences in support are statistically significant at any level, but more importantly they are extremely large. Support among military officers is especially low when considering that the vast majority of the military officers in our sample are college-educated men, who are even more supportive of deploying the military in both scenarios than the rest of the public (see Figure 7 below).

Figure 5. Support for military deployment, public versus military officers.

As with the US public survey, partisan cues appeared to have no significant effect on these military officers’ attitudes regarding military deployment. In both scenarios, respondents supported deploying the military—as well as militarizing the police and deputizing armed volunteers—at nearly identical rates regardless of whether protesters were from the same or opposite party (see Appendix Table A6).

Our surveys also provide information on the nature of military officers’ normative disagreement with civilian voters. The issue does not appear to be institutional trust: 93 percent of military officers say that they trust the US military most or all of the time, far higher than any subset of the public. Nor do they show any weaker of a preference for law and order: military officers were just as likely to support higher defense spending and more frequent military intervention overseas as civilians. Military respondents were also far more likely to trust the police.

Instead, the disagreement centers around the proper role of the military in American politics. The clearest evidence is in comparing military and civilian attitudes across the three outcomes in the vignette experiments (Figure 6). Military officers are far less supportive than civilian respondents in deploying military personnel (23 versus 67 percent) or deputizing civilian volunteers (6 versus 37 percent) to quell protesters. However, they are more likely to agree with arming the police with military-style equipment for the same mission (64 versus 53 percent). This variation suggests that military officers have a strong sense of proper institutional roles. Military officers hold just as strong a preference for law and order as the public, but instead prefer that the police enforce it rather than the uniformed military or deputized civilians. Military respondents’ preferences also manifested in other questions about the proper role of the military in our survey. While the military and civilian respondents held similar views on most civil-military issues, they held starkly different opinions on military leaders “going public” when civilian leaders ignored their advice. While a plurality of the public were comfortable with military leaders voicing their disagreements in the press, military officers were nearly uniformly opposed.⁶ These results mirror findings elsewhere that military veterans in Congress are more likely to vote for institutional constraints on the military (Lupton, 2017).

Figure 6. Support for government actions, public versus military officers.

Mechanisms

Are these differences between military and civilian respondents primarily rooted in different information about the norms and capabilities of the military, or in different professional and personal commitments?

On one hand, anecdotal evidence suggests that military officers did have misconceptions about the legality of domestic military deployments. During the piloting and fielding of the survey, the authors received multiple emails and comments from military officers disputing the premise of the question, claiming that domestic military deployments are by their nature illegal under the Posse Comitatus Act of 1878. As noted above, this is incorrect: Posse Comitatus has several large exceptions, and the scenarios were crafted with these exceptions in mind.

However, our evidence instead suggests that these differences are grounded in military officers’ personal and professional normative commitments, rather than different knowledge. First, members of the public with greater exposure to the norms and capabilities of the military—namely military veterans and those with friends and family in the military—were no less likely to support deploying the military into domestic unrest than others in the general public. If military officers’ beliefs about military deployment were based primarily in their knowledge (or misconceptions) about military norms, laws, and capabilities, we should expect military veterans and friends and family to be less supportive of deploying the military into domestic unrest than the general public. Instead, both groups were generally very supportive of such deployments (Figure 7). In our sample, military friends and family were actually more supportive than other members of the public to favor deploying the military veterans (60 versus 67 percent). Neither difference is statistically significant at the ${\rm p}< 0.05$ level.

Figure 7. Support for military deployments, by sub-sample.

Note: These percentages combine the outcome questions in both vignettes.

Second, among the military officers in our surveys, longer-tenured officers were less likely to support deploying the military than their junior colleagues. If military officers’ beliefs were likely to be rooted in mere exposure to military norms or legal misconceptions conveyed early in military training, then relatively shorter-tenured officers should hold similar beliefs to longer-tenured officers. Instead, officers who had served more than twenty years were significantly less likely to support deploying military forces than those who had served less than 20 years: 13 versus 28 percent in vignette 1, 18 versus 33 percent in vignette 2. These differences are statistically significant, despite the small number of observations. By contrast, longer-tenured officers were no less likely to support militarizing police forces or deputizing armed volunteers. Military officers increasingly internalized martial norms surrounding the domestic use of the military over years of socialization and selection into higher ranks.

Taken together, our evidence suggests that military officers’ opposition to domestic military deployments is not based primarily on knowledge (or misconceptions) about military norms and capabilities. Instead, it is more likely to stem from differences in priorities and professional commitments built over career-long socialization and promotion patterns.

Real-world implications

Both empirical patterns—the lack of partisan bias among the US public and the normative divide between civilians and military—were observable during the Black Lives Matter protests in Summer 2020, with important implications for their outcome. During the Summer of 2020, public polls showed consistent, bipartisan support for deploying military forces to quell protests. In a Morning Consult poll in the first week of protests, before any significant violence or property damage, 58 percent of Americans (including 48 percent of Democrats) agreed that President Trump should deploy uniformed military personnel to contain protesters—only modestly fewer than the 71 percent who favored deploying the National Guard.⁷ When violence escalated in Portland, Trump threatened to deploy the military and National Guard but appeared to face significant push-back from the military. While uniformed military largely stayed silent, retired military officers spoke out—most notably Trump's former Secretary of Defense James Mattis, who publicly warned the White House that deploying the military would be a “threat to democracy.”⁸ When unrest escalated in Portland, President Trump ultimately opted to deploy federal law enforcement (Customs and Border Patrol and US Marshals) rather than the military. Even then, however, he faced significant criticism from the Department of Defense. When federal agents deployed in “uniforms that made them look military in appearance,” Secretary Mark Esper strenuously objected, reportedly saying, “We want a system where people can tell the difference.” At the same time, DOD spokesman Jonathan Hoffman went out of his way to declare publicly that the military would not be deployed to Portland because the Posse Comitatus Act expressly forbade it—again, blurring the line between a political and a legal constraint.⁹ These events suggest an environment in which the primary obstacle to deploying the military was not concerns about public opposition—even among Democrats—but instead military resistance.

6. Conclusion

In highly polarized environments, when do American civilians and military officers support and oppose the use of the military to contain protests—and what do those findings say about? Our survey offers two major conclusions.

First, even in contexts with high partisan stakes, partisan antipathy appears to have very little effect on public attitudes regarding the military's role in domestic law and order. Large majorities of the public, regardless of party, support deploying the military to quell potentially violent protests. More surprisingly, these majorities remained about equally strong regardless of the partisan implications: members of both parties were about equally likely to favor domestic military actions against Republican protesters and against Democratic protesters.

Our findings point to a deep public consensus in favor of law and order that overrides even partisanship. Despite public concerns over the threat that affective polarization poses to national security, we find little evidence that partisan bias infects attitudes on the military, protests, and political violence.

Second, even when the public generally supports military interventions into domestic politics, military professionalism may act as an effective bulwark against such militarization. Contrary to the attitudes of the general public, our survey of military officers shows widespread opposition to deploying the military into partisan protests. The norm of non-interference runs deep among officers, particularly those who have served for a long time—and much more so than among veterans or military family members. This is consistent with warnings from policymakers and journalists that the public is becoming increasingly uncritical of the military, willing to outsource the hard work of security as long as they do not have to bear the cost. As James Fallows of The Atlantic put it, “the American public and its political leadership will do anything for the military except take it seriously.”¹⁰ On this issue, the military does appear to take its role very seriously.

Given the nature of our findings, future work should examine how nonpartisan institutions become drawn into partisan politics. At what point in the decision-making process do such institutions risk losing political legitimacy and popular support? How do political actors who are seen as non-partisan become more partisan over time? What political conditions drive voters to view the military through a partisan lens? Are public perceptions bound by policy domain (e.g., domestic versus foreign security) or issue area, or are other factors driving public attitudes? Insights into these questions will not only advance research on political polarization, but will also shape policy responses to future partisan unrest.