Is the American Public Really Turning Away from Democracy? Backsliding and the Conceptual Challenges of Understanding Public Attitudes

Abstract

Recent scholarship has warned that the American public is turning away from democracy, with many focusing on the role of polarization in driving these trends. While these studies devote a great deal of attention to conceptualizing and measuring polarization, however, there is much less attention to the concept of democracy itself. As a result, they encounter several problems: First, lack of attention to the categorical and contextual differences between various democratic transgressions can lead to misleading findings. Second, slippage between macro and micro level analysis of backsliding can make it difficult to draw solid inferences about attitudes. Finally, lack of attention to contestation within the concept of democracy obscures a different kind of regime contention reflected in public attitudes. These problems confound efforts to understand whether the American public is truly turning away from democracy and point to the need for greater collaboration among scholars from different intellectual traditions and methodological orientations.

Recent scholarship has warned that the American public is turning away from democracy. Building on work that has identified alarming signs of backsliding in the United States, (Dionne, Ornstein, and Mann 2017; Levitsky and Ziblatt 2018; Ginsburg and Huq 2019; Mettler and Lieberman 2020; Norris 2017), a number of studies have shown a shift in attitudes that suggests a decline in public support for democracy (Foa and Mounk 2016; Wike and Fetterolf 2018). Efforts to understand what may be contributing to the shift have led scholars to focus on the role of polarization, and particularly the rise in affective polarization, a form of estrangement that transforms partisanship into a kind of social identity (Iyengar and Westwood 2015; Mason 2018). Under these conditions, partisan affiliations can become tribal and even sectarian as rival groups become “enemy camps” (Arbatli and Rosenberg 2021; Finkel et al. 2020). This form of polarization, it is argued, can pose dangers for democracy as those embracing polarized partisan identities are more likely to overlook transgressions by co-partisans in many realms, but especially when it comes to upholding democratic principles.

A long line of scholarship has maintained the importance of democratic attitudes to the preservation of democracy (Easton 1965; Eckstein 1961; Mattes and Bratton 2007; Norris 2011 Rose, Mishler, and Haerpfer 1998). Both for the purpose of political legitimacy and as a check on undemocratic leaders, public support for democratic principles is essential for democratic endurance (Booth and Seligson 2009; Claassen 2019; Welzel 2007). Thus, the question of whether the American public is truly turning away from democracy could not be more pressing. The explosion of research on this question, however, has yielded no clear answers, as various studies have produced mixed and contradictory findings. Some have argued unequivocally that, in the context of heightened polarization, democracy takes a backseat to partisan interests (Arbatli and Rosenberg 2021; Graham and Svolik 2020; Svolik 2019). Others have found no significant evidence of a connection and have cautioned that affective polarization is a concept best confined to understanding social interactions (Broockman, Kalla, and Westwood 2020; Touchton, Klofstad, and Uscinski 2020). Some studies have found mixed support, with polarization impacting some democratic principles and not others (Carey et al. 2020; McCoy, Simonovits, and Littvay 2020). And others still have found polarization to have an asymmetric impact with Republicans displaying greater effects of partisan bias than Democrats (Bartels 2020; Carey et al. 2019; Stolle, Gidengil, and Bergeron-Boutin 2019).

Part of the difficulty in making sense of these findings is that while these studies devote a great deal of attention to conceptualizing and measuring partisanship and different forms of polarization, there is much less attention to the concept of democracy itself. Often there is very little justification given for the specific measures of democracy used beyond general references to classical works in democratic theory. As a result, these studies encounter several problems: First, a lack of conceptual clarity regarding different types of transgressions leads to a mix of measures that reveal very different things about democratic attitudes and prospects for backsliding. Without accounting for the categorical and contextual attributes of different types of transgressions, these studies can lead to misleading findings. Second, there is slippage between macro and micro-level analysis of backsliding, as a paradigm intended to understand institutional change is adopted to investigate attitudinal shifts. This can make it difficult to derive solid inferences about attitudes with the treatments used. Finally, a lack of attention to contestation around the concept of democracy itself obscures a different kind of regime contention reflected in public attitudes. This is a form of contention that pits competing democratic principles against each other and may reveal a contest of values rather than an unwillingness to uphold democracy.

These problems confound efforts to understand whether the American public is truly turning away from democracy and point to the need for greater conceptual clarity. I examine each of these problems and offer suggestions for how they may be addressed. My goal is not to resolve the debate about democratic attitudes in the United States. Nor do I provide a comprehensive analysis of the conceptual challenges associated with backsliding. Though I will offer insights into both, the goal here is to address a unique challenge at the nexus of these two areas of inquiry: how should we interpret public attitudes about democratic transgressions? The three main problems identified here operate at different levels, becoming progressively thornier and requiring more of an interpretive lens as we move through the levels. They are offered as independent interventions such that one need not resolve all three to engage with one or the other. Whether it is by clarifying concepts and measures, translating analytical frameworks, or grappling with the instability of meaning in the concepts themselves, engaging with these conceptual and interpretive challenges will help us better understand the American public’s attitudes towards democratic backsliding. Importantly, while this article is meant to help refine analytical tools related to survey research and attitudinal analysis, ultimately it points to the need for greater dialogue and collaboration among scholars from different intellectual traditions and methodological orientations.

Democratic Transgressions

The first problem that arises from the wave of scholarship on polarization and democratic attitudes is that these studies employ a wide array of democracy measures without adequately theorizing what different measures can tell us about prospects for democratic backsliding. Typically done through a survey or candidate choice experiment, respondents are asked whether they approve of acts or positions that contradict democratic principles (refer to the online appendix for a full list of treatments used in the studies referenced). However, though all studies purport to measure democratic attitudes in general, the range of treatments reveals a high level of heterogeneity in the nature of the transgressions examined, both within and across studies.

The problems arise due to the lack of differentiation between transgressions in terms of their categorical and contextual attributes. On the categorical front, these studies conflate several distinct categories of transgressions: violations of the law; violations of democratic norms; violations of democratic ideals; and power-consolidating changes to democratic institutions. Each represents a different kind of threat and public attitudes towards them will tells us something very different about the potential for backsliding. In terms of contextual attributes, these studies elide factors necessary to understand the extent to which different transgressions represent a retreat from the status quo and thus an instance of backsliding. Accounting for both the categorical and contextual differences between transgressions can help us make better sense of public attitudes.

I will discuss the different transgressions used in these studies with a view to classifying them along these lines and developing a sensibility towards the level of threat each represents. The goal of this exercise is not to dictate a particular hierarchy of transgressions. Rather, it is to show, through my own attempt to order them, the benefits of theoretically grounded classifications, and conversely, the dangers of an undifferentiated view of transgressions.

Categorical Differentiation: Types of Transgressions

Violations of the law. This category of transgression suggests an actual violation of the rule of law, the constitution, established procedure, or other formal institutions regulating democratic competition. Examples of such transgressions in the studies examined here include resorting to violence to settle disputes (Bartels 2020; Clayton et al. 2021), forging ballots (Graham 2020), and political interference with investigations (Carey et al. 2020). I also include here questions about the rejection of election results where the question clearly indicates that the candidate has lost (Carey et al. 2020). Some studies ask directly about whether the executive should be constrained by the law (McCoy, Simonovits, and Littvay 2020). Violations of the law represent the most egregious of democratic transgressions. Adherence to the rule of law, it should be noted, is important for the stability of any system of government, but takes on added significance within democratic regimes. This is because the law needs to constrain all citizens, and especially the powerful, in order to avoid power asymmetries that would undermine the possibility of self-governance (Przeworski and Maravall 2003; Holmes 1995). Thus, while having the executive disobey the law would be destabilizing for any system of government, it poses specific dangers for democracy, and public attitudes expressing support for such transgressions would signal a concerning level of democratic decay.

Violations of democratic norms. The term norm has been applied to a broad range of activities in the scholarship on public attitudes. Here I use it to specify informal rules that govern political interactions. Norm violations are not violations of the law but of how things have been done in the past or departures from common expectations of how politics will be conducted (Fishkin and Pozen 2018). Some examples from these studies include abolishing the filibuster (McCoy, Simonovits, and Littvay 2020), election overrides by state legislatures (Broockman, Kalla, and Westwood 2020), and flouting court rulings (Graham and Svolik 2020; McCoy, Simonovits, and Littvay 2020; Stolle, Gidengil, and Bergeron-Boutin 2019). While not as severe as violations of the law, norm violations also represent dangers to democracy. This is because democratic stability requires actors to be able to develop expectations about repeated interactions and laws are often not enough to provide this. Informal rules emerge to supplement codified rules and fill in the gaps where the law is silent or lacks clarity (Azari and Smith 2012; Helmke and Levitsky 2004). For this reason, a violation of norms could potentially damage democracy indirectly by creating uncertainty in interactions and encouraging further norm violations. Even if the act would bring the polity closer to some ideal version of democracy (for example, conferring statehood to Washington, D.C., or eliminating the filibuster to pass voting rights legislation), the norm violation in itself would be destabilizing. According the Levitsky and Ziblatt (2018) a violation of norms damages the “guardrails” of democracy, the unspoken rules that help to keep competition within acceptable limits.

Violations of democratic ideals. These actions neither violate the law nor past precedent but rather an aspirational view of how democratic politics should be conducted. Some examples from these studies include an unwillingness to compromise (Broockman, Kalla, and Westwood 2020; Carey et al. 2020; McCoy, Simonovits, and Littvay 2020), an acceptance of constraints on civil liberties (Broockman, Kalla, and Westwood 2020; Graham and Svolik 2020; McCoy, Simonovits, and Littvay 2020), corruption tolerance (Broockman, Kalla, and Westwood 2020) ¹ and conspiratorial thinking (Clayton et al. 2021). Certainly, laws as well as norms are also informed by democratic ideals. But to the extent that ideals do not become codified into law or develop into norms, they reflect a different category and their violation holds more ambiguous meaning for democratic backsliding. Within democracies, there may be high levels of agreement on ideals, but pragmatic limitations on the extent of their application. Take the example of attitudes toward compromise found in many of these studies. While there may be agreement in principle that compromise is beneficial for democracy, and certainly it would be difficult to imagine any system of self-government that could long endure without it, there are pragmatic limits to how much or how often we can reach decisions through comprise rather than unilateral action. It is precisely for this reason that mechanisms of governance such as executive orders – offered in several of these studies as the non-compromising position – are incorporated into the constitution. These ambiguities will limit what we can learn from an unwillingness to uphold ideals.

Power-consolidating changes to democratic institution. This does not involve violations of the established rules or practices, but rather changes to the law or the constitution that consolidate the power of ruling elites (Bermeo 2016; Gerschewski 2021; Haggard and Kaufman 2016; Waldner and Lust 2018). In the studies examined here, some transgressions that fit within this category include support for reducing ballot access and gerrymandering, as well as more egregious acts such as expanding presidential powers, removing presidential term limits, and abolishing Congress (Carey et al. 2020; Graham and Svolik 2020; McCoy, Simonovits, and Littvay 2020; Touchton, Klofstad, and Uscinski 2020). These changes can constitute an endogenous mode of de-democratization that is hard to track given that change operates through and not against democratic institutions. This category of transgression exploits an essential feature of democratic governance: institutions are endogenous to preferences. This means that they can be changed democratically. And this is in fact the most common mode of de-democratization found in counties experiencing significant backsliding (Bermeo 2016; Haggard and Kaufman 2021). Alternately referred to as “autocratic legalism,” these kinds of transgressions can be the most difficult to detect and potentially some of the most dangerous (Scheppele 2018).

Contextual Differentiation: Prevalence in U.S. Politics

In addition to distinguishing the category of transgressions, an evaluation of contextual factors may be important to understanding their implications for backsliding. One important dimension for such an evaluation is the prevalence of different transgressions in the context of U.S. politics. Such considerations are important in determining whether or not the transgression represents an instance of backsliding, understood as movement away from a democratic status quo, or the continuation of a previously undemocratic status quo in a particular realm. ²

To be clear, this is not meant to suggest that the prevalence of a practice makes it democratic, nor that the status quo is satisfactory, only that these contextual factors are necessary to discern changes that would constitute backsliding. For example, attitudes supporting racial exclusion would be undemocratic under any circumstances. But if we were to detect a preponderance of attitudes in support of segregation in the United States in 1968, we would not consider this evidence of backsliding but rather the continuation of prevalent undemocratic practices. Detecting such attitudes today, however, would certainly indicate a retreat from the status quo and an instance of backsliding. Thus, context plays a role not in setting the standard for what is or is not democratic, but for measuring change. Consequently, such contextual factors are important in understanding whether or not public attitudes are reflecting the status quo, or countenancing a form of democratic decay.

At the margins, prevalence may not be as useful in differentiating between transgressions; violations of the law typically will not be prevalent, while violations of ideals may be quite common. However, within the categories of norm violations and power-consolidating institutional changes, there can be significant variation, and this can help us understand the significance of public attitudes towards transgressions in these arenas. Take for example the issue of partisan gerrymandering which comes up in several studies as a violation of electoral fairness (Graham 2020; Graham and Svolik 2020; McCoy, Simonovits, and Littvay 2020). While this can be considered an instance of changes to democratic institutions that might consolidate the power of ruling elites—a category of transgression that poses one of the greatest dangers to democratic stability—we know that the practice of gerrymandering is not uncommon in U.S. politics. It goes at least as far back as 1812 when Governor Eldrige Gerry notoriously crafted a salamander shaped district for partisan gain. There is caselaw on gerrymandering going back to the 1940s (Colegrove v. Green 1946), and the courts have offered no conclusive indication of when partisan bias represents a constitutional violation. That the result of partisan gerrymandering could damage the vitality of electoral competition and by extension democratic stability is not in doubt. However, given the prevalence of this practice, and the fact that there has emerged no definitive political consensus on the matter, it represents the status quo, and public attitudes in support of it cannot be offered unproblematically as evidence of democratic decay. Moreover, it cannot be put on equal footing with more egregious forms of power-consolidating institutional change such as shutting down Congress which also appears in some studies (Stolle, Gidengil, and Bergeron-Boutin 2019; Touchton, Klofstad, and Uscinski 2020).

One can also use contextual factors to distinguish between different types of norm violations. Though norm violations by definition are departures from past practices, there can be degrees of variations in this as well. Take for example the question of judicial deference which appears in several studies (Carey et al. 2020; Graham and Svolik 2020; McCoy, Simonovits, and Littvay 2020; Stolle, Gidengil, and Bergeron-Boutin 2019). Though constitutionally there is ambiguity regarding whether the president is bound by the decisions of the Supreme Court, there is such a strong norm regarding the matter that its violation could bring about a constitutional crisis. ³ Compare this to court packing as a strategy of partisan gain or doing away with the filibuster to advance a partisan agenda, which also appear as norm violations (McCoy, Simonovits, and Littvay 2020). Looking back in U.S. history reveals many instances of court packing and that the filibuster itself emerged to replace a norm of majoritarianism in the Senate (Koger 2010; Wawro and Schickler 2007). Given this ambiguity, it is reasonable for us to conclude, and for respondents in these studies to determine, that such changes are part of the ebb and flow of U.S. politics rather than instances of democracy-eroding norm violations. With regard to judicial deference there is decidedly less ambiguity.

Navigating Diverse Measures

Bringing together both the categorical and contextual factors at work in these measures, a clearer picture emerges regarding the level of threat various transgressions present to democracy. In figure 1, a visualization is offered combining these two dimensions and placing the treatments used in these studies in terms of the level of threat they represent. Those actions deemed most egregious according to this classification are centered and enlarged and those that are less threatening are reduced and located further out in the periphery. All else being equal, violations of the law would typically be at the center and violations of ideals farthest out, with violations of norms and power-consolidating changes falling somewhere in between. Factoring in contextual considerations, we may further decide to move acts within these categories based on their prevalence.

Figure 1 Mapping Transgressions

Note: The level of threat of a given act is indicated by the text size and its location either closer to the center (more threatening) or periphery (less threatening). Specific acts are marked with a star (*) if any study found support for partisan bias effecting evaluations of transgressions.

Classifying transgressions in this way can help us make sense of the findings within these studies. For example, in the study reporting the strongest findings, Graham and Svolik (2020) examine partisan support for candidates engaged in a range of transgressions that would fall under 1) violations of norms (prosecution of journalists, judicial deference); 2) violations of ideals (executive orders, protests bans for far right/left groups); and 3) power-consolidating institutional changes (gerrymandering and ballot access). Considering the categorical and contextual attributes of these trangressions, we might order them in terms of threat as follows: judicial deference > prosecution of journalists > ballot access > protest bans > executive orders > gerrymandering. Importantly, the study reports the strongest findings for items considered least threatening and the weakest findings for items posing the greatest threat: “Respondents most severely punish candidates who want to prosecute journalists (16.1%) and ignore court rulings (14.1%). Respondents are least sensitive to candidates who endorse gerrymandering (by 2 seat, 10.6%) and suggest that the governor ban protests or rule by executive order (10.2 and 10.5%, respectively)” (Graham and Svolik 2020, 402). While there is a positive finding for all items, this suggests that the overall threat represented in these findings may not be as severe as it would appear based on an undifferentiated view of transgressions.

Such a heuristic can also help us make sense of findings across studies. Looking at how these studies assess attitudes towards executive constraint, for example, we see that authors operationalize this in ways that reflect different categorical attributes: McCoy, Simonovits, and Littvay ask if respondents would support impeachment for criminal misconduct—a violation of law; Graham and Svolik ask if they would support leaders who rule by executive order—a violation of ideals; and Touchton, Klofstad, and Uscinski ask about increasing executive power through the constitution—power-consolidating institutional change. Combined with this categorical delineation, contextual factors may lead us to conclude that a willingness to increase the president’s constitutional powers would pose a greater threat than both a reluctance to impeach for criminal activity, which enjoys some ambiguity depending on legal interpretations of the terms of impeachment, and the use of executive orders, which is highly prevalent. This means that even if these studies were all accurately measuring public attitudes for the items they identify, they are not all measuring the same thing. Again, across these studies, we find the strongest findings for items representing the weakest threat—executive orders (Graham and Svolik 2020, 402)—and no significant findings for the item representing the most serious threat—executive aggrandizement through constitutional changes (Touchton, Klofstad, and Uscunski 2020), suggesting that the public’s democratic commitments may not be as weak as it might at first appear.

It should be stressed that this exercise is not meant to impose an artificial uniformity either within or across studies. It is understandable that scholars would want to speak to different issues and it is important that they pursue whatever questions they think worthwhile. But just as important is that they precise the theoretical claims to match the specific treatments. The problem here is not that these studies are measuring different things but that they all purport to be measuring the same thing, which can lead to misleading interpretations, exaggerating the threat in some areas and downplaying it in others.

Certainly, scholars may disagree on the categorical and contextual attributes of specific transgressions. Some may develop different classifications altogether depending on the theory of democratic breakdown they adopt. The purpose of the classification offered in this analysis is not to dictate a specific hierarchy of transgressions, but rather to demonstrate the benefits of adopting a theoretically grounded and systematic approach to conceptualizing democratic transgressions and choosing measures to reflect that. Of course, many acts will blur the lines. But, imperfect as our classificatory schemes may be, attending to the categorical and contextual differences across transgressions is important in understanding what our treatments are actually measuring and what it might mean for democratic backsliding.

Macro versus Micro Backsliding

A second and related area of difficulty we encounter has to do with the level of analysis in these studies, and specifically the slippage that takes place between a macro view of backsliding and the micro-level of analysis offered through survey experiments. The macro view is what is typically utilized in institutional analysis of backsliding, and it is meant to assess in aggregate whether a polity is moving away from democracy. However, in order to apply the concept of backsliding to the study of public attitudes, scholars have had to focus on discrete acts and query respondents about them in isolation. This micro-level analysis is in many ways incommensurable with the nature of backsliding, which often consists of acts that are on their face democratic.

The problem stems from the fact that the focus of many of these studies is democratic backsliding, typically understood as a process that moves a country away from democracy but does not cross the line into a different regime type (Haggard and Kaufman 2021; Waldner and Lust 2018). According to Waldner and Lust this mode of backsliding “makes elections less competitive without entirely undermining the electoral mechanism; it restricts participation without explicitly abolishing norms of universal franchise seen as constitutive of contemporary democracy; and it loosens constraints of accountability by eroding norms of answerability and punishment” (Waldner and Lust 2018, 95). Though backsliding can also refer to movement across regime types or degradation of autocracy, democratic backsliding typically refers to changes that remain within the bounds of democratic politics. This has been alternately conceptualized as a form of autocratization (Lührmann and Lindberg 2019), de-democratization (Agh 2015), regression (Tomini and Wagemann 2018), erosion (Haggard and Kaufman 2021), or a general loss of quality in democracy (Erdmann and Kneuer 2013).

The nature of democratic backsliding presents certain challenges when trying to understand attitudes. Since the individual events that would constitute backsliding may not be explicitly undemocratic, they are hard to identify. This has made backsliding notoriously difficult to measure (Lueders and Lust 2018; Waldner and Lust 2018). Bermeo has described a “vexing ambiguity” in understanding backsliding within discrete institutions, and notes that, in isolation, many events may not appear to be threatening (Bermeo 2016, 19).

The solution for many scholars is to move to the macro level when determining instances of backsliding. That is, while the analysis of institutional change may focus on specific events, the determination of whether this constitutes backsliding is made at the systemic level, taking into consideration what is happening in other arenas. This typically yields aggregate measures of backsliding assessed across multiple dimensions (Lührmann and Lindberg 2019; Skaaning 2020). In fact, most studies of backsliding have measured it at either the national level or globally, with some further specifying “episodes” rather than discrete observations as the appropriate level of measurement (Lührmann and Lindberg 2019; Pelke and Croissant 2021). Waldner and Lust offer a strategy for coding backsliding that would look across the political system, and require degradation in at least two out of the three dimensions of democratic governance they identify: competition, participation, and accountability (Waldner and Lust 2018, 95). This macro-level analysis gives greater confidence that micro-level findings are meaningful indicators of backsliding.

However, this is part of the difficulty in applying the concept of backsliding to understanding public attitudes. It is one thing to evaluate backsliding through the lens of institutional change and draw conclusions about whether, in aggregate, the observed changes move a polity farther from democracy. It is another thing entirely to evaluate public attitudes about discrete acts that are on their face democratic but may combine to constitute backsliding. This is effectively what happens in many of these studies when respondents are asked to make determinations about actions in isolation. When asked about gerrymandering, while respondents may recognize it as an undesirable practice, there is no reason to conclude that it could compromise democracy. When asked about the use of executive orders, while they may recognize that interbranch compromise is preferrable, they can express support for this and see no conflict with upholding democracy. As discrete acts, it becomes difficult for respondents to evaluate them as part of a pattern of backsliding. And because these acts are on their face democratic, it is difficult for researchers to draw solid inferences about respondents’ attitudes toward democracy or support for backsliding.

For that reason, measuring public attitudes about various transgressions may not reveal an un/willingness to uphold democracy per se. It may be used to assess overall systemic stability—the greater the support for acts that may contribute to backsliding, the more unstable the regime overall. But this would not necessarily reflect the public’s democratic commitments. If instead we want to assess support for backsliding, a different approach is necessary to bring the macro perspective into the micro-level analysis of attitudes. One strategy would be to design treatments that connect discrete acts to a broader context of democracy erosion. In other words, we would need to prime respondents to think about the potential for backsliding. Instead of allowing respondents to assume a situation of democratic stability, we could portray a situation in which some of the pillars of U.S. democracy are fraying. Instead of asking in a general sense about judicial deference, we could indicate that defying the courts might bring about a constitutional crisis. Instead of asking in general about ballot access and redistricting, we could specify that changes to ballot access would reverse some of the major gains of the civil rights movement.

What allows us to make determinations of backsliding in the institutional realm is an understanding of the broader context and the way in which seemingly innocuous acts can contribute to an overall pattern of democratic decay. Similarly, if we want to make determinations about support for backsliding in the attitudinal realm, the treatments used must provide the relevant context for the actions as well as the stakes. Such questions can be constructed in a way that avoids response bias associated with the desirability of democracy, but still indicate the interconnectedness of the specific act with broader patterns of decay (Kiewiet de Jonge 2016; Zaller and Feldman 1992). Of course, this does not resolve the problem that in the real world, where respondents do not have this context constructed for them, they may support actions that further backsliding. But these limitations are inherent to the backsliding paradigm. It is a concept derived to understand an in-between stage, a form of political change that remains within the realm of democratic politics but moves us farther away from democracy. However we choose to approach these questions, there will necessarily be trade-offs.

Democracy and Contention

The final area of concern and that which can be the most difficult to tackle is the question of contention within the category of democracy itself. Many of these studies of public attitudes proceed as if there were a stable notion of democracy within the public sphere that we can measure attitudes against, ignoring sometimes fundamental disagreements about what constitutes a democratic act or which is the more democratic course of action. This is not a question of whether they employ the appropriate benchmarks for democracy. Rather the problem is that they assume a fixed status for democracy, which belies the fundamental ambiguity that exists within this category. For any given question, there may be multiple and competing democratic principles that actors can appeal to in order to justify a preferred course of action. And there is no natural hierarchy that would help us determine which principles should be prioritized.

Take an issue that has been at the center of public debate and appears in many studies examining public attitudes on democracy: ballot access. In these studies, this appears as questions about voter ID laws (Carey et al. 2020), purging voter rolls (McCoy, Simonovits, and Littvay 2020), the length of time polling places are open (Graham and Svolik 2020), as well as general questions about voting access (Stolle, Gidengil, and Bergeron-Boutin 2019). Though in these studies there appears to be a clear democratic position, one aimed at increasing voter access, we know that in the public sphere, these issues are debated through different democratic lenses, pitting those seeking to make voting more accessible against those who fear that ease of access will translate into corruption. This is precisely what happened with the issue of mail-in-voting that captured public attention during the 2020 election. The circumstances of the pandemic in the context of the 2020 election intensified polarization around these questions and heightened the urgency of resolving them, but the contours of the debate were familiar, with one side identifying voting access as the core democratic principle and the other emphasizing electoral integrity. ⁴

It may be tempting to dismiss appeals to electoral integrity as insincere or illegitimate, perhaps even revealing undemocratic intent at their core. We know of course that actual instances of electoral fraud are rare, and thus such claims may be seen as merely masking partisan motivations. But whatever the intent of the elites making these claims, for the purpose of understanding public attitudes, the important thing is that these actors do make appeals to democratic principles and do so convincingly to their audiences. Certainly, there may be an element of “partisan motivated reasoning” at work here (Bolsen, Druckman, and Cook 2014; Edelson et al. 2017), but that could be true on both sides of the debate. The push to increase access to voting is no doubt a noble one in principle, but in practice, we know that parties only target their own voters for such efforts, voters who they can identify with ever greater precision thanks to advances in voter databases and micro-targeting efforts (Panagopoulos 2016). Therefore, the push to increase voter participation surely has partisan implications as well, and this can be seen as playing a role in support for voting access. In both cases, partisanship is enmeshed with democratic ideals in ways that are difficult to tease apart.

In dealing with the question of ballot access then, the problem is not that these studies choose to focus on one set of democratic principles and not others, but that they proceed from the assumption that there is a single clear democratic position, dismissing the contestation that takes place through democratic claims-making. Even more problematic is that some attempt to settle this contest by imposing a specific interpretation of events. For example, many introduce questions of ballot access as measures of voter suppression. More than just a label, this suggests that the controversy over this interpretation is immaterial. Whether or not we share this interpretation (I do), it is not helpful for our analysis to ignore differing interpretations, or to assume that our social scientific determinations of what the correct position is constitutes the relevant benchmark.

All this points to the need for alternative methods of grappling with regime contention in the United States, and a recognition that there are in fact multiple modes of contention taking place. Certainly, we can find traditional forms of regime contention and actors who would seek to establish autocratic practices and institutions. But alongside that, there is an equally consequential struggle over the very meaning of democracy. This is not a struggle of democrats versus autocrats, or even democrats versus partisans, but rather between democrats appealing to one set of principles and democrats appealing to others. This kind of regime contention in many ways is more difficult to grapple with than traditional forms because at its core there is a contest of values that are not subject to scientific adjudication but are fundamentally political questions of what kind of democracy we wish to achieve.

Though the issues underlying the contention are not themselves resolvable in this manner, however, there are tools that may help introduce some conceptual clarity. Daniel Slater has identified similar forms of regime contention in other democracies, leading to what he describes as a kind of democratic “careening” rather than backsliding (Slater 2013). Democratic careening is defined as “political instability sparked by intense conflict between competing visions of democratic accountability” (Slater 2013, 731). Slater notes that there has been a tendency to collapse this form of regime contention into more traditional forms, obscuring the fact that they operate according to very different logics. Slater is especially interested in the question of accountability, which has been at the heart of regime contention in Thailand. There, he argues that careening has led to a struggle, not between democrats and autocrats, but between those advancing populist visions of accountability resting on inclusivity and those advancing a more oligarchical vision emphasizing constraints. This accountability dimension in Slater’s formulation rests orthogonal to the democracy/autocracy dimension (Slater 2013, 739–40).

While the specifics will differ, this mode of conceptualization may also be fruitfully deployed to add much needed complexity to the analysis of regime contention in the United States. Most important is that it opens up the possibility of understanding the competition over democratic principles as separate axes of regime contention, related to but not reducible to the competition over democracy and autocracy as systems of governance. We can envision many such dimensions, some of which may rest closer to the democracy/autocracy axis depending on how closely they align with those structures (see figure 2).

Figure 2 Dimensions of Regime Contention

Such a scheme might also allow us to move past political democracy as the only sphere of contestation. If we understand democracy as a system designed to resists or ameliorate structures of domination, we can consider multiple modes of domination. In the context of the United States, we may wish to think about the ways in which different acts impact not just the domination of one political group over another (political democracy), but also relations of domination between racial groups (racial democracy) or economic groups (economic democracy). The same act may register differently given these different considerations. Take again the question of ballot access, still examined as a different dimension of contestation, but one that will align with democracy differently depending on the sphere. Given the differential impact of restrictions on ballot access for racial minorities and lower income groups, we can make a claim that electoral inclusivity aligns more closely with racial and economic democracy because of its importance in resisting those structures of domination (see figure 3).

Figure 3 Spheres of Regime Contention

However we may wish to populate the dimensions of regime contention relevant to a specific context, it is important to stress that this is at best an analytical tool and not a means of resolving the contest of values at the heart of these controversies.

This recognition need not dissolve into a nihilistic view, or the idea that we are helpless in the face of contrasting visions of democracy. U.S. history is rife with such forms of regime contention. At each stage, the direction that has been taken, either toward more inclusionary democratic reforms or toward exclusionary constraints, has been determined not by reference to some absolute value of democracy, but by the ability of different groups to form viable coalitions that advance their preferred visions of democracy and of democratic peoplehood (King and Smith 2011, 2014; Mettler and Lieberman 2020; Smith 2003, 2015). Those seeking to better understand these dynamics of regime contention would benefit from greater engagement with a significant literature on American political development that has returned to these themes time and again to show how such ideational struggles have been resolved historically in favor of greater or less inclusion. A long line of work has shown the importance of coalitions in moving democracy both towards inclusionary expansion (King and Smith 2011; Mickey 2015; Noel 2012; Rhodes 2017; Schickler 2016; Teele 2018) and toward exclusionary retrenchment (Ahmed 2013; Bateman 2018; Keyssar 2009; Rana 2011; Valelly 2009). These works show the contingency of the struggle for democracy, and demonstrate that, while there are no assurances that inclusionary visions will prevail, inclusive coalitions can be forged under even the most hostile circumstances (Gilmore 2008; Krochmal 2016). This might also help us better understand how coalitions might be formed today to fight for more inclusive visions of democracy.

The challenge of understanding different forms of regime contention in the United States also points to the need for different approaches to understanding public attitudes. Rather than just looking at attitudes and asking if they are democratic, we need to better understand the construction of different attitudes as democratic. Here we can tap into work that has shown the heterogeneity of ideas and practices associated with democracy in order to understand the variety of meanings ordinary people attach to this concept (Bratton 2010; Schaffer 2000). We can also benefit from work that has utilized ethnographic methods to study public opinion (Cramer Walsh 2009). Such work can help us better understand the political associations that allow for the denial of ballot access to be constructed as democratic, political violence as patriotism, and the rejection of election results a defense of democracy. In addition, a significant body of scholarship has shown how the politics of ideational construction has shaped critical junctures in the history of U.S. democracy (Hattam and Lowndes 2008; King and Smith 2014; Lieberman 2002; Orren and Skowronek 2004; Skowronek 2006). This approach, which sees ideas as essentially unstable, shaped and reshaped by political forces, in some way runs counter to an approach that treats ideas (via attitudes) as stable enough to measure through survey research. But there can be fruitful dialogue between the two insofar as both seek to understand how ideas about democracy can impact democratic outcomes. Indeed, this kind of dialogue, across intellectual traditions and methodological perspectives, may be the key to understanding regime contention in the United States.

Conclusions

The preceding discussion is not meant to diminish the contributions of scholarship seeking to understand the public’s democratic commitments. There may not be a more pressing question confronting U.S. political science today, and each of these studies represents important efforts to grapple with complex issues. However, additional attention to conceptual clarity would enhance the contributions of these studies in significant ways. First, researchers need to adopt theoretically grounded and systematic approaches to conceptualizing democratic transgressions and choose measures to reflect that, differentiating transgressions in terms of their categorical and contextual attributes. Further, a process of translation is necessary when moving from the macro-level study of backsliding in the institutional realm to the micro-level analysis involved in the study of attitudes, contextualizing discrete acts so that they are legible as instances of backsliding. Finally, research must be attentive to the ways in which competing democratic principles inform public attitudes, leading to different forms of regime contention that are relatable to but not reducible to the contest over democracy and autocracy. Throughout this article, I have offered suggestions for refinements that may improve the inferences we can draw from such studies. It must be stressed, however, that the problems noted here cannot be resolved exclusively through technical fixes. That is, it will not be enough to refine measures or devise new survey techniques. What is clear from this discussion is that advancing our understanding of the state of democracy in the United States will require sustained scholarly engagement across multiple research traditions and methodological orientations.