Assessing the Risk of Democratic Reversal in the United States: A Reply to Kurt Weyland

ABSTRACT By replying to Kurt Weyland’s (2020) comparative study of populism, we revisit optimistic perspectives on the health of American democracy in light of existing evidence. Relying on a set-theoretical approach, Weyland concludes that populists succeed in subverting democracy only when institutional weakness and conjunctural misfortune are observed jointly in a polity, thereby conferring on the United States immunity to democratic reversal. We challenge this conclusion on two grounds. First, we argue that the focus on institutional dynamics neglects the impact of the structural conditions in which institutions are embedded, such as inequality, racial cleavages, and changing political attitudes among the public. Second, we claim that endogeneity, coding errors, and the (mis)use of Boolean algebra raise questions about the accuracy of the analysis and its conclusions. Although we are skeptical of crisp-set Qualitative Comparative Analysis as an adequate modeling choice, we replicate the original analysis and find that the paths toward democratic backsliding and continuity are both potentially compatible with the United States.

A merican democracy survived the 2020 presidential election and its aftermath. After a four-year assault on the nation's democratic norms, Donald Trump's autocratic inclinations (Lamont, Park, and Ayala-Hurtado 2017;Levitsky and Ziblatt 2018;Mounk 2018;Przeworski 2019) were successfully reined in. However, the probability of observing democratic backsliding in the United States remains an open and important question. Before the election, Kurt Weyland (2020) critically built on the intellectual exercise proposed by Levitsky and Ziblatt (2018) to compare Trump's populism to that observed in other cases.
Expanding the main thesis of his coedited book (Weyland and Madrid 2019), Weyland concludes that the cases in which populist rulers succeeded in eroding democratic regimes were those in which political institutions were weak-a condition that, he argues, is not observed in the contemporary United States. Weyland is to be commended for bringing comparative politics back to the study of American politics and correctly predicting democratic resilience against a direct assault by Trump. However, we take issue with his argument on two grounds: its theoretical premises and its methodological execution. On theory, we highlight the importance of considering the effect of structural constraints when assessing the health of American democracy, which-we argue-remains fragile despite the outcome of the 2020 election. On methods, we show that crisp-set Qualitative Comparative Analysis (QCA)-a method developed by Ragin (2014a) that draws on Boolean algebra and set-theory-is inadequate for the type of predictive exercise developed in Weyland (2020).

THEORETICAL ACCOUNT AND ARGUMENT
Structural and race-based inequality, as well as attitudinal shifts in the American citizenry, arguably are tied to the increasing levels of social and political polarization observed in the United States (Lamont, Park, and Ayala-Hurtado 2017). Mettler and Lieberman (2020) argue that high inequality, racism, and polarization are independently associated with previous crises of American democracy. For the first time, these factors converged in a multilayered crisis of the political system, epitomized by the Trump presidency.
Even before Trump, an American Political Science Association task force (2004) expressed concern about the political effects of increasing inequality in the United States. Since then, further studies have shown that economic inequality in the country translates into political inequality, skewing legislation toward the preferences of the well-off (Kelly 2020;Page and Gilens 2020). Mounk (2018) noted that this bias already is a sign of democratic deterioration if we consider democracy to be a regime in which political decisions follow the preferences of average citizens. Moreover, a series of recent analyses of the politics of race in the United States show that inequality is associated with significant restrictions on Black and Latino access to crucial civil and political rights (Bateman 2018;Baumgartner, Epp, and Shoub 2018;Fraga 2018;Hajnal 2020;Jardina 2019;Michener 2018;Phoenix 2019). In addition, a series of political attitudes-traditionally envisioned as "pillars of democracy"-shifted significantly in the American electorate (Norris and Inglehart 2019;Przeworski 2019). One important shift is the process of polarization, leading to "cultural wars" in the contemporary United States (Carothers 2019).
Eventually, Trump's presidency encouraged researchers to compare the American experience with other cases of populist rule. In an attempt to isolate their common traits, Levitsky and Ziblatt (2018) famously compared cases in which democracy did break after the emergence of populist leaders. They concluded that democracy eroded when elites underestimated the threat of populist outsiders and overestimated existing institutional capacity to contain them. Democracies, they argued, are secure only when systems of mutual tolerance and institutional forbearance prevent aspiring autocrats from reaching office. According to Levitsky and Ziblatt (2018), because Trump was able to overcome those safeguards, the risk of democratic erosion in the United States should not be dismissed.
In contrast, Weyland (2020) compares both cases in which populist leaders have "suffocated" democracy and others in which they have not. He theorizes that democratic survival is contingent on six factors: (1) the levels of political instability; (2) a leader's capacity to operate outside of institutional norms by pursuing paralegal change; (3) the strength of checks and balances in the political system, which prevent rapid change; (4) the incidence of an economic crisis; (5) the incidence of an internal-security crisis; and (6) a country's dependence on natural resources.
Despite its different aims and theoretical emphases, Weyland's (2020) theory replicates one of the main limitations of Levitsky and Ziblatt's (2018) framework-namely, a relative disregard for structural variables that shape both institutional dynamics and democratic survival. In particular, we emphasize the importance of two overlooked predictors of democratic reversal: inequality and attitudinal change in society. Whereas political institutions have remained formally the same in the United States, the depth and nature of social inequality and citizens' attitudinal predispositions toward politics have changed significantly (Piketty, Saez, and Zucman 2018;Przeworski 2019).
Disregarding the effects that these (and other) structural factors have on institutional dynamics fosters endogeneity: democracy in the United States is portrayed as stable and safe given institutional stability, whereas institutional stability is seen as the main consequence of democratic stability. Moreover, because one of the claimed advantages of a set-theoretical perspective is the possibility of accounting for several potential causal variables when comparing few cases, the neglect of structural factors potentially driving the causal dynamics of democratic backsliding is unwarranted.
Scholars already have identified income inequality as a potential driver of democratic erosion in the United States (Lieberman et al. 2019;Mounk 2018;Przeworski 2019). Income distribution also has been shown to correlate with the emergence of populism elsewhere (Pastor and Veronesi 2018) and democratic breakdown (Acemoglu and Robinson 2006). As shown in figure 1, the United States is at its historical peak regarding inequality and ethnic diversity. The country also shows higher levels on both indicators than comparable advanced capitalist societies.
Economic inequality makes it easier for elites to dominate legislative bodies (Gilens 2012;Page and Gilens 2020) and policy (Bartels 2016;Kelly 2020). To the extent that inequality undermines citizenship rights, it may make repression more effective against minorities and the poor (Cole 2018;López 2020). These dynamics slowly erode the legitimacy of democratic institutions while predisposing increasingly large electoral constituencies against "the establishment." Also, these dynamics often precede the polarization that we observe at the institutional level when incumbent politicians and populist leaders seek to advance their career in a socially polarized scenario.
Omitting the effects of inequality is particularly striking for the analysis of democracy in the United States, where race and ethnicity are complexly associated with privilege (predominantly for whites) and with unequal civil, political, and social entitlements for people of color and poor white citizens. In a few decades, the United States crystallized as a highly unequal and ethnically divided polity among advanced capitalist societies. Some scholars point to inequality and broader changes in the country's social stratification (Putnam 2016) as driving consequential attitudinal shifts in the American public (Jardina 2019).
Accounting for these attitudinal traits and their relationship to structural trends in society therefore is relevant (Inglehart and Welzel 2005;Ruck et al. 2020). Foa and Mounk (2016) showed how younger generations in established democracies are more receptive to nondemocratic alternatives, whereas Kwak et al. (2020) argued that those attitudes, when prevalent among new-citizen cohorts, are associated with subsequent reductions in the levels of democracy. Moreover, research recently portrayed the emergence of authoritarian leaders as being facilitated by prevalent resentment among segments of the Even before Trump, an American Political Science Association (APSA) task force (APSA 2004) expressed concern about the political effects of increasing inequality in the United States.
electorate (Norris and Inglehart 2019). Xenophobia and racism are pivotal attitudes today and they appear disproportionately prevalent among the younger and older cohorts in Western societies, including the United States (Inglehart and Norris 2019). Attitudinal change and citizens' predispositions usually are tied to increasing anger, resentment, and animosity against institutions and also are posited as a possible source of observed political polarization in the United States (Miller-Idriss 2020; Przeworski 2019).
In summary, we think social structures (i.e., high levels of inequality and society's racial stratification) and their attitudinal counterparts are fundamental drivers of institutional dynamics. Without accounting for such possible drivers, institutional strength is "explained" endogenously by preexisting institutional strength. We acknowledge US democratic exceptionalism. Yet, we also claim that from a comparative perspective, the United States is better understood as an unlikely democracy, a historical outlier among a more egalitarian and secular set of Western nations (see Mettler and Lieberman 2020). In turn, institutions are at least partially endogenous to social structure and therefore can unravel rapidly in the context of structural change.

METHODOLOGICAL SHORTCOMINGS
The adoption of QCA to estimate the likelihood of democratic decay in the United States also is analytically problematic. Recent work strongly discourages the use of QCA for causal analysis (Baumgartner and Thiem 2020; Bowers 2014; Lucas and Szatrowski 2014; and see Ragin 2014b for a defense). Even if effective for other analytical purposes, set-theoretical approaches are deterministic and thus inadequate for assessing the likelihood of a future event. Yet, we still think it is useful to replicate the original analysis using QCA while adjusting for debatable coding decisions and omitted variables. We demonstrate that coding corrections-and the inclusion of three relevant omitted variables discussed previously-lead to very different Boolean reductions. Bracketing the inadequacy of QCA for this purpose, the results we obtained are ambiguous regarding the immunity of the United States to democratic recession.
Set-theoretical approaches can be interpreted as being deterministic in nature (Mahoney 2000). Therefore, the method is not appropriate for estimating the probability of an unobserved outcome (i.e., the likelihood of democratic breakdown in the United States). Second, according to the previous theoretical discussion, the exclusion of important structural causes of democratic breakdown should be reconsidered. Third, assuming that the research goal is to estimate the determinants of democratic breakdown and continuity, instead of the probability of "suffocation," other methodological fallouts deserve attention.

The United States is included in the analysis.
By entering a Democratic Suffocation score for the United States, the idea that the Boolean reduction could predict that same score becomes redundant. Moreover, except for the United States, all Omitting the effects of inequality is particularly striking for the analysis of democracy in the United States, where race and ethnicity are complexly associated with privilege (predominantly for whites) and with unequal civil, political, and social entitlements for people of color and poor white citizens. cases in the sample have some form of alleged institutional weakness. In a Boolean-logic application, this means that institutional weakness becomes a necessary condition for any outcome that is observed elsewhere while not being currently observed in the United States.
2. The measure of weakness correlates with systems of government and region.
The original analysis includes three indicators of institutional weakness: High Instability, Paralegal Change, and Easy Change. The latter indicator denotes the separation between presidential systems (which display more constraints on executive power) and parliamentary or semi-presidential systems. We find it counterintuitive to classify parliamentary regimes as institutionally weak by default, which this coding scheme implicitly does. Researchers have long touted the benefits of parliamentarism for democratic survival (Linz and Valenzuela 1994). Those who question the superiority of parliamentary regimes claim that there are no meaningful differences between systems in terms of stability (Cheibub 2007). Furthermore, case selection accounts for an almost perfect correlation between "Easy Change" (i.e., parliamentary systems) and being a European case (r=0.9), Bulgaria is the only European case not included in the set despite its parliamentary system.

Attributes of institutional strength and signs of democratic erosion are interchangeable.
The set of cases with "High Instability" consists of countries that recently experienced irregular removals of a president or coup attempts (Weyland 2020, online appendix). "Paralegal Change" denotes the set of cases in which leaders of the executive are able to coerce other branches of government (i.e., where the separation of powers is constrained). The coding of the latter two attributes creates a problem of endogeneity because clear signs of democratic erosion are used to estimate the likelihood of suffering democratic erosion. At best, the assessment implies the logically consistent but theoretically inconsequential notion that democracies that broke recently-or are in the process of breaking-are more likely to break in the future.

Criteria for case inclusion and coding are inconsistent.
We believe that the inclusion and coding of some cases could be reconsidered. Bulgaria, as mentioned, is coded as nonparliamentary. Ecuador is coded as a member of the "Hydrocarbon Windfall" set, which comprises countries dependent on natural resources.
However, it is only coded as such when ruled by Correa and not when ruled by Moreno, Bucaram, or Gutiérrez. Moreover, Weyland (2020) notes that Moreno's administration is not a case of populist rule but is included nonetheless because it illustrates "the impact of the resource curse." Moreno could never be a positive case of autocratic populism because he is not a populist. His inclusion thus violates the possibility principle (Mahoney and Goertz 2004).
5. Reduction strategies do not follow the QCA protocol.
Boolean reduction is not properly executed. First, the reduction makes use of only the logical AND (∧), dismissing the logical NOT (¬). Second, the analysis assumes symmetrical causation (i.e., the configurations for breakdown are the reverse mirror of those of continuity). In doing so, the analysis quickly concludes that democracy in the United States is "unlikely" to breakdown without estimating how the paths to democratic continuity would appear.
Due to excessive concern with the event of a populist subverting democracy in self-serving ways, cases in which democracy eroded because of a backlash against a populist leader are coded as cases of democratic continuity (e.g., Zelaya in Honduras and Bucaram in Ecuador). Those cases more properly would be coded as nonmembers of the set of Democratic Survival, particularly when the outcome of interest is NOT-Suffocation.

REPLICATION AND RECONSIDERATION
We now focus on an alternative version of the data (López 2021) (the full dataset, all results, and R codes are in the online appendix), which recalibrates the outcome set, excludes the United States from the sample, removes endogenous measures, recodes the cases of Ecuador and Bulgaria, and adds case memberships for three sets: unequal countries, 1 ethnically divided countries, 2 and countries with attitudinal complexes in the public that are considered more compatible with authoritarianism. 3 Table 1 displays the number of predicted paths toward both democratic breakdown and NOT-Breakdown. We assess each added variable separately and in combination.
As long as we do not consider the possibility of economic and security crises occurring in the United States, Weyland's (2020) argument resists each adjustment separately because NOT-Membership in "High Instability," "Easy Change," and "Hydrocarbon Windfall" continue to predict "NOT-Democratic Suffocation." However, inconsistencies do emerge. For instance, the recalibration of the outcome portrays membership in the "High Instability" and "Paralegal Change" sets as associated with NOT-Suffocation (see the online appendix for all reductions). When considering the emergence of economic and security crises, removing endogenous measures is sufficient for the US membership to match one path toward Democratic Suffocation and no path toward NOT-Suffocation. The results display further inconsistent configurations leading to either democratic breakdown or survival. For instance, membership in the "Economic Crisis" set is not only a necessary condition for breakdown but also part of one path toward survival. Moreover, the latter fits the other US set memberships, suggesting When considering the emergence of economic and security crises, removing endogenous measures is sufficient for the US membership to match one path toward Democratic Suffocation and no path toward NOT-Suffocation.
that American democracy would benefit from a major economic crisis. Some scholars might claim that the United States is already a member of the economic and security crises sets due to the repercussions of the COVID-19 pandemic and the insurrection by Trump loyalists. In that case, the only full match for the United States is consistent with a Democratic Suffocation outcome. We do not take these results at face value but rather as evidence that QCA reductions failed to clearly differentiate case configurations leading to either democratic breakdown or survival-even when specified in ways that we consider more theoretically and methodologically adequate.

Kurt
Weyland is an influential figure in the field who should be commended for bringing comparative politics back to the study of American politics. He is theoretically correct in claiming that populists who succeeded in suffocating democratic institutions had more working in their favor than their autocratic aspirations. Notwithstanding several insightful aspects of his analysis, we believe that structural causes of democratic breakdown are critical for understanding the predicament that the United States now faces. The record levels of inequality and ethnic polarization, as well as the attitudinal changes associated with them in contemporary American life, suggest that the likelihood of democratic reversal in the United States is much higher today than it was in the not-so-distant past. The upshot of this analysis seems clear: when accounting for these variables, the future of American democracy remains uncertain.

DATA AVAILABILITY STATEMENT
Replication materials are available on Harvard Dataverse at https://doi.org/10.7910/dvn/em2gao. 2. We imputed the Ethnic Fractionalization Index (EFI) with a reference year of 2013 and coded all cases with EFI>0.5 as members of the set of ethnically divided countries. 3. We count as members of the "Traditional Values" set those countries that scored below zero in the Self-Affirmation/Traditional Values Index in the wave of the World Values Survey most proximate to the emergence of the populist ruler.