The COVID-19 lockdown paradox: democratic support during democratic restrictions

Abstract

Previous research has shown that positive perceptions of government performance are linked to higher levels of citizens’ support for democracy. However, the policy response to the COVID-19 crisis presented a unique paradox as relative success in preventing the virus spread depended on expanding executive powers, often at the cost of individual freedoms. Exploring this paradox, we investigate whether the link between perceptions of government performance and support for democracy holds in a situation where positive performance essentially means a restriction of freedoms. Using original survey data from seven European countries, we show that notwithstanding the democratic sacrifices, people with positive evaluations of the government’s response are more likely to maintain support for the democratic system. Nevertheless, people weighed responses to the health domain more heavily than to the economic domain, suggesting that the output legitimacy – democratic support link varies across domain-specific evaluations.

Introduction

A long-standing observation in the scholarship of political regime support is that while evaluations of the performance of democratic regimes on the diffuse-specific continuum introduced by Easton (1965, 1975) varies a great deal between countries and over time, support for the abstract ideals of democracy is widespread and stable (Inglehart, 2003; Klingemann and Fuchs, 1995; Dahlberg et al, 2015; Kriesi, 2020). A predominant majority of democratic electorates sign up to the basic principles of democracy even if they do not necessarily agree on what exactly those principles may be (Schedler and Sarsfield, 2007; Ferrín and Kriesi, 2016). Though contemporary democracies may face legitimacy problems because of a growing mass of “dissatisfied democrats” or “critical citizens” (Norris, 1999; Dalton, 2004), the basic challenge appears to revolve around the performance of the system rather than around the basic contours of the game. When citizens are confronted with a choice between their own conceptualization of democracy (in the abstract) and a hypothetical authoritarian alternative, an overwhelming majority continues to support the only game in town.

However, when the COVID-19 pandemic hit the democratic world in the spring of 2020, democratic governments ventured on unchartered terrain. Public consultations, standard institutional procedures, and even parliamentary routines suddenly appeared to be put on hold; emergency powers and executive prerogatives became the new normal. Moreover, the initial success of some authoritarian regimes, such as China, Singapore, and the United Arab Emirates, in halting the spread of the virus and minimizing its death toll via stringent lockdown measures might have risked further alienating the minority of citizens flirting with authoritarian alternatives in normal times. After all, if authoritarian systems appear to provide a superior fix to a global health emergency, democracies’ legitimacy advantages vis-a-vis their authoritarian alternatives might be justifiably questioned.

At the beginning of the crisis, however, democracies seemed to cope reasonably well with this legitimacy challenge: increased trust in elected officials, a high level of satisfaction with the way democracy works, and general approval of stringent lockdown measures conferred on governments a grace period not unlike the ones they tend to enjoy when the public rallies around the flag in wartime (Baker and Oneal, 2001; Groeling and Baum, 2008; Bol et al., 2021; Altiparmakis et al., 2021). However, hints of impatience on part of the public following the strict lockdowns began to surface as the economic situation in these countries turned dire.

In this paper, we take up the challenge of investigating the individual-level determinants of democratic regime support in the context of the COVID-19 pandemic with an eye on what we consider as the lockdown paradox. If the widespread acceptance of the basic rules of democracy stems from the output legitimacy that these regimes enjoy (Scharpf, 1997; Armingeon and Guthmann, 2014), what happens to the performance-legitimacy link when good performance becomes synonymous with restrictions of civil liberties, the empowerment of non-elected experts, and a general marginalization of public debate? Alternatively put, does a general approval of “democracy on hold” come at the expense of democratic regime support or does it reinforce the legitimacy of democratic regimes by underscoring their resilience to unforeseen and unprecedented shocks? This special characteristic of the COVID-19 pandemic, which provides a context of curtailment of civil liberties in advanced democracies, makes it a hard test for the performance-legitimacy link. If the link holds in this least likely context when curtailment of civil liberties could become synonymous with good performance, it is very likely to hold in other contexts and crises as well.

Our guiding research question is not merely academic; authoritarian values and the level of democratic regime support have historically been associated with democratization or democratic reversals for that matter (Inglehart, 2003; Brunkert et al., 2019; Claassen, 2020a). In the context of an established democracy, Graham and Svolik (2020) provide evidence from a natural experiment in the US state of Montana to quantify the electoral costs of candidates’ anti-democratic platforms and show that they are fairly limited as far as the local levels of American politics are concerned. Furthermore, Claassen (2020b) shows that erosions of pro-democratic attitudes can happen even in long-established democratic societies, as increases in democracy are associated with a decrease in democratic mood through a negative thermostatic effect. In this regard, our paper contributes to the literature on democratic regime support by bringing under the same theoretical and empirical umbrella three main sets of individual-level drivers of anti-democratic attitudes: output legitimacy, exposure and threat, and political orientations. While much of the literature on anti-democratic attitudes has been focused on the relevance of contextual and institutional factors which are strong predictors of such attitudes across countries and time (Weil, 1989; Anderson and Guillory, 1997; Berggren et al., 2004; Aarts and Thomassen, 2008; Esaiasson et al., 2012), our interest falls on individual-level factors that can explain variation in democratic regime support between individuals within a constant context (advanced democracies in Europe) and at a particular point in time (the COVID-19 crisis). Understanding these individual-level drivers of anti-democratic attitudes may thus provide important cues to the future of democratic stability in Europe.

Additionally, a second contribution we make to the literature on democratic regime support is the extension of output legitimacy to multiple and novel domains of policy performance under a unified framework. Contrary to much of the literature that tends to focus on single dimensions of legitimacy (economic evaluations, corruption, crime, or government performance in general), we use the unique opportunity offered by an original survey fielded in seven European countries in the wake of the first wave of the pandemic (June 2019) to probe policy evaluations on two separate domains. These were presented as trade-offs faced by policy-makers: getting the epidemic under control and keeping the economy on steam. Which dimension weighs more heavily in the minds of citizens as far as the performance evaluation – democratic regime support link is concerned, provides crucial insights to the basis of democratic legitimacy in Europe and beyond.

The rest of this article proceeds as follows. In the next section, we outline our theoretical expectations regarding the micro-drivers of democratic regime support with policy evaluations in the focus. The third section presents our original survey and our empirical strategy. The fourth section presents our findings, and the final section concludes.

Who supports democracy?

The sources of democratic legitimacy are manifold in contemporary democratic societies. The more one moves out towards the diffuse end of regime support, the more certain structural and relatively immutable features of the regimes under study come to play a prominent role. For instance, Cho (2014) demonstrates that the very age of democracies matters: as people become accustomed to living in democracies and learn about its functioning, the more they nurture pro-democratic attitudes¹. In a related argument, Huang et al. (2008) provide evidence that developments within the broad process of modernization – urbanization, economic development, secularization, etc. – contribute to diffuse democratic system support. Within the realms of established democracies, other systemic characteristics of political regimes may be an equally important determinant of pro-democratic attitudes and system support, such as the system’s majoritarian-consensual features as well as the electoral and the party system’s characteristics (Weil, 1989; Anderson and Guillory, 1997; Berggren et al, 2004; Aarts and Thomassen, 2008; Esaiasson et al., 2012). Closer to our theoretical interest, the quality of governance also appears to play a decisive role, though more so for the evaluation of democratic performance rather than for pro-democratic attitudes per se (Ariely, 2013; Magalhães, 2014; Boräng et al., 2017).

These accounts focus on contextual and institutional factors, which are important in specifying where diffuse support for democracies originates from and how it can vary across countries and time. However, they are also inherently limited in predicting differences in individuals’ pro-democratic attitudes within a constant political context precisely because they rely on macro-level, systemic variables. This limitation is especially acute for studies emphasizing long-term, slow-changing structural processes and institutions, such as the degree of entrenchment of democratic institutions, the basic logic of the electoral systems, or the availability of party supply. These are more or less widely shared knowledge and experience by voters and, therefore, unlikely to account for differences between their democratic attitudes within the same country at a single point in time.

Given the focus of our study on studying individual variation in democratic regime support at a specific point in time (the COVID-19 crisis), and considering the above limitation, our theoretical interest is focused mostly on individual-level factors, such as socioeconomic and political attitudes, as the more likely predictors of individual variation between democratic regime support within a constant context. While we acknowledge the importance of institutional factors (and also how they might interact with individual-level factors), our paper does not aim for a direct, systematic investigation of such interactions. Instead, we aim to capture the association between individual-level factors and anti-democratic attitudes above and beyond such institutional differences (i.e., while controlling for them). In light of this, our paper contributes to the literature on democratic regime support by identifying and bringing under the same theoretical and empirical umbrella three main sets of individual-level drivers that have been previously associated with anti-democratic attitudes in normal times: output legitimacy, exposure and threat, and political orientations. Hence, we ask, do these usual suspects explaining regime preference in normal times still play a role during a national emergency? Linking explanations related to output legitimacy with explanations related to threat perceptions and political orientations allows us to map regime support at the individual level and understand the likely sources of friction that might affect the future of democratic stability in Europe. Furthermore, it allows us to expand on previous empirical studies on the legitimacy-democratic support link across multiple and novel policy domains.

Among the individual-level factors associated with democratic regime support, policy output legitimacy and performance evaluations have received a great deal of interest in previous literature. People may have vastly different views on how the economy functions, they have different experience with the day-to-day dealings with state authorities and live in different socioeconomic conditions where even the same policies have fundamentally different ramifications on the ground. In the economic domain, for instance, Katz and Levin (2018) show that citizens’ perception of negative economic shocks in Latin America over the period of 1996–2015 was closely related to their level of support for democratic systems over alternatives. Comparing 26 European countries during the Great Recession, Armingeon and Guthmann (2014) show that support for democracy is largely conditional on economic and political performance. Mishler and Rose (2002) and Waldron-Moore (1999) report similar findings on system support in another turbulent economic context: post-communist transition in Central and Eastern Europe. Dalton (2004) utilizes World Values Survey data from the 1990s to arrive to similar conclusions though his findings on the link between democratic norms and economic performance are not particularly strong.²

Economic performance is but one of multiple aspects of output legitimacy. Another chief candidate for linking pro-democratic attitudes with output legitimacy concerns corruption perceptions. In fact, the two concepts are not just empirically but also conceptually linked. After all, one of the alleged advantages of democratic over authoritarian governance is the former’s emphasis on universalism and equality before the law. If, instead, particularism and favoritism prevail in the widespread perception of citizens, they may justifiably withdraw diffuse support from their nominally democratic systems. Accordingly, Linde (2012) uncovers a link between the perception of the impartiality of institutions and preferences for democratic regimes over alternatives in Central and Eastern Europe. These findings are replicated for Latin American countries by Canache and Allison (2005) who use Transparency International’s Corruption Perception Index to reach similar conclusions. Anderson and Tverdova (2003) focus on the evaluation rather than general support for democracy, and confirm the corruption-system support nexus: in mature and new democracies alike, citizens tend to lower their evaluation of democratic performance when they perceive a higher level of corruption.

Beyond such broad and diffuse phenomena as economic performance and systemic corruption, citizens are also inclined to withdraw support from democratic regimes in response to day-to-day experiences of physical insecurity. Ceobanu et al. (2011) focus on personal victimization as a predictor and find a stronger impact on the evaluations of democratic performance rather than on preferences for democratic regimes in the abstract. However, Fernandez and Kuenzi (2010) demonstrate that perception of crime in society is also a strong predictor of democratic regimes in general. Though both studies are conducted in highly crime-ridden parts of the democratic world in Africa and Latin America, the findings combined with other aspects of output performance, discussed above, hammer home a simple message. Citizens’ perception of the government’s performance on various policy dimensions, such as economic performance, corruption and crime³, are important determinants of their loyalty to the broad democratic rules of the game.

Output legitimacy: Economic and health performance evaluations

Applying these insights to the COVID-19 pandemic in Europe, the natural question to raise is whether this time may be different or is it business as usual? Alternatively put, in the context of an unprecedented type of twin-crisis with policy responses taking a heavy toll both on people’s personal liberties and on their economic well-being, does the performance evaluation – democratic support nexus remain intact? Our baseline expectation is in the affirmative. Given the huge stakes of the game – tens of thousands of lives and double-digit losses in economic output – a negative evaluation of the government responses is likely to translate to lower probabilities of democratic system support. This is because individuals who are critical of the crisis management may perceive that the system as such has let them down and simply failed to rise to the occasion. Moreover, if even under an extreme concentration of executive powers democracies are unable to exercise the necessary authority to fight the pandemic and its economic aftermath, how can democracies be expected to stand up to even more existential threats in the future? For these considerations, our baseline expectation regarding the link between the evaluation of government responses to the COVID-19 pandemic and democratic system support, simply follows in the footsteps of the literature on output legitimacy:

H1a: Individuals with positive evaluations on the government’s response to the pandemic and the resulting economic fallout are more likely to support democratic systems compared to citizens with more critical evaluations of the government’s response.

This baseline approach, however, overlooks a crucial aspect of the policy responses to the COVID-19 pandemic. Most of the policy dimensions under study in the extant literature capture valence issues where “good performance” is a rather universally desired good that all society benefits from. A booming economy with low unemployment and stable prices, a generally transparent public sphere with high levels of moral integrity and few instances of corrupt practices, safe cities where crime is more the exception rather than the rule, are unquestionable benefits in the minds of citizens.

Good government performance in fighting the COVID-19 pandemic is a different story. On the one hand, given the unprecedented nature of the problem pressure, citizens had little access to past experience to form an unambiguous, or even a coherent opinion on what good policy performance might look like. As a default, they tended to fall back on expert advice and accept lockdown measures as the necessary price to pay for getting the pandemic under control (Altiparmakis et al, 2021). The acceptance of these measures, however, also presented citizens with a latent dilemma as far as their regime support is concerned. If they were to draw on the regime’s output legitimacy as a source of diffuse system support for their democracies, good performance would stand in conflict with some of their core democratic rights. Restrictions on freedom of movement, the banning of large political gatherings, mandated use of masks and other regulations collectively implied a significant curtailment of citizens’ civil liberties, pitting some of the core principles of democracy against the imperative of defeating the pandemic. Caught in this dilemma, many citizens may infer from a positive reading of stringent lockdown measures (good performance evaluation) that, in some contexts, democracy may not be the only game in town after all. Building on this lockdown paradox, we expect a divergence between the impact of performance evaluations on the two policy dimensions under study on citizens’ propensities of democratic regime support. While positive evaluations of the economic responses continue to reinforce citizens’ commitment to democracy, a positive evaluation of the lockdown measures have a weaker impact or none at all.

H1b: Positive evaluations on the economic response to the pandemic have a larger positive impact on citizens’ democratic regime support than evaluations of the health response.

Alternatively, the impact of domain-specific evaluations is a function of salience rather than of the degree of sacrifice in terms of core democratic rights. In the initial stages of the pandemic with the airwaves filled with the number of infections and death counts, the health aspect is likely to weigh heavier in the minds of citizens than the economic one due to the sheer novelty of the nature of the emergency. According to this logic, health evaluations are likely to exert a larger impact on democratic regime support than economic evaluations:

H1c: Positive evaluations on the health response to the pandemic have a larger positive impact on citizens’ democratic regime support than evaluations on the economic response.

Exposure and threat: socioeconomic exposure, egotropic, and sociotropic threat perceptions

While important drivers of democratic attitudes in themselves, policy evaluations are not formed in a vacuum. Citizens have varying levels of personal exposure to the pandemic either via their socioeconomic exposure (such as age, income, or having children) or via their more diffuse subjective assessment of threat. This logic bears resemblance to the aforementioned link between the experience of personal victimization by crime and support for democracy (Ceobanu et al, 2011). It is one thing for individuals to form a negative assessment of policy output in a given domain, but it is another to feel (or to have been) directly victimized by it. Carlin and Singer (2011) build on this insight and include multiple victimization items in their models and arrive at mixed findings: personal economic pain, victimhood of crime, corruption and discrimination have varying influence on support for different aspects of polyarchy⁴. Additionally, previous experimental research focused on citizens’ policy preferences in the first wave of the COVID-19 pandemic shows that socioeconomic exposure and subjective threat have different levels of influence in determining preferences for mild or strict lockdowns (Oana et al., 2021). We adopt here a similar approach by acknowledging and differentiating between multiple sources of exposure to COVID-19 and, therefore, distinguishing between the socioeconomic exposure of respondents and the subjective threat they perceive. Specifically, we hypothesize that citizens who are more exposed to the pandemic and the resulting economic fallout, both in terms of their socioeconomic characteristic and their diffuse, subjective threat perceptions, are more likely to be skeptical of democracies’ merits in fighting this unprecedented challenge. In what regards socioeconomic exposure, we consider the effect of age, having children, and income as likely to have an impact insofar as older individuals, those having children, and those suffering major income losses are less likely to support democratic regimes compared to the rest of the population. In what regards diffuse, subjective threat exposure, we consider citizens’ egotropic threat perceptions, that is, perceptions of the crisis as being threatening for their personal health and financial situation.

Nevertheless, exposure to the crisis does not necessarily require personal involvement in it. Perceptual threats do not only reflect what is considered more threatening personally (egotropic threat), but also reflect sociotropic considerations based on what is considered more threatening for the broader community or the country one is a part of. Previous research has devoted much attention to studying whether political preferences are based on one’s egotropic considerations, or whether, these preferences follow sociotropic concerns based on what is desirable for the whole country (e.g., Kinder and Kiewiet, 1981; Joslyn and Haider-Markel, 2007; Mansfield and Mutz, 2009). For instance, the economic voting literature has shown that it is country-level sociotropic factors that play a more important role in shaping public attitudes towards rewarding/punishing incumbents (Duch and Stevenson, 2008). More specifically, in the context of the COVID-19 pandemic, Oana et al. (2021) show that sociotropic threat concerns are a significant predictor of preferences for mild or strict lockdown measures alongside personal exposure. Along these lines, we expect sociotropic threat perceptions, that is perceived threat to the national public health and to the national economy, to exert a positive effect on people’s skepticism towards democratic ideals.

H2a: Citizens with higher levels of socioeconomic exposure to the public health emergency and the economic fallout are less likely to support democratic regimes.

H2b: Citizens with higher levels of egotropic threat perception to the public health emergency and the economic fallout are less likely to support democratic regimes.

H2c: Citizens with higher levels of sociotropic threat perception to the public health emergency and the economic fallout are less likely to support democratic regimes.

Political orientations: winner-loser partisan status and ideological leanings

Beyond their policy evaluations and personal exposure, citizens’ political orientations in the context of the pandemic are likely to serve as yet another powerful source of diffuse support for their democratic regimes. In fact, the core insight behind the link between policy evaluations and system support lies in a sense of helplessness by critical citizens. It is not merely particular political actors – who may be voted out at the next elections – that are to blame for botched policy responses but the system as such is at fault. Such broader, anti-systemic sentiment, however, is likely to be also driven by citizens’ degree of attachment to elected politicians. First, by having participated in the previous election, individuals already placed some degree of confidence in the system. Second, by having their preferred politicians in policy-making positions, negative policy evaluations may be counterbalanced by other benefits that an individual derives from having their partisan choice in power, such as ideological alignment, tangible benefits in other policy domains, or just the sheer utility derived from being in the “winners’s camp” (Di Tella and MacCullogh, 2005). In fact, the empirical literature has demonstrated a close link between such winner-loser status and citizens’ evaluation of democracy (Anderson and Guillory, 1997; Curini et al, 2012; Henderson, 2008). In the context of the policy response to COVID-19, the importance of being in the “winners’ camp” may be especially pronounced. Facing deep uncertainty on the future course of the pandemic and on the appropriate policy response, one of the few sources of citizens’ reassurance that the system is able to address for this novel challenge is that their preferred partisan choice is in power. We thus expect a link between citizens’ prior vote choices and their democratic regime support:

H3a: Citizens who have voted for government parties at the last elections are more likely to support democratic regimes over authoritarian alternatives compared to opposition voters or non-voters.

Such government-opposition dichotomy, however, is not the only relevant dimension for conceptualizing “winner-loser status.” Though opposition voters may be “political losers” in the sense of feeling unrepresented by government parties, some of them may have paradoxically emerged as “ideological winners” from the policy debate surrounding COVID-19. From this ideological perspective, the policy response to COVID-19 can be seen as a quintessential left-wing moment. The extreme salience of healthcare issues, new responsibilities of the state in implementing track and trace systems, encouraging vaccine development, procuring protective equipment and other medical devices, the unprecedented degrees of fiscal largesse including generous furlough schemes and wage guarantees are all part and parcel of a left-wing policy package and touch upon policy domains that the left traditionally owns. Likewise, the COVID-19 policy debate unleashed a pan-European wave of transnational solidarity and common policy solutions, culminating in an unprecedented step in fiscal integration and risk-sharing in the form of the Recovery Fund (Kriesi et al., 2024; Oana and Truchlewski, 2024). Just like the national-level policy responses marked an ideological victory for the left, a step towards fiscal integration marked an ideological defeat for voters of Eurosceptic parties.

H3b: Left-wing voters and pro-European voters are more likely to display pro-democratic attitudes than centrist/right-wing voters and Eurosceptics.

Data and measurement

To test our hypotheses, we rely on data from an original survey fielded in seven European countries (France, Germany, Italy, the Netherlands, Spain, Sweden, and the United Kingdom) between the 5th and the 22nd of June 2020 when relaxation measures were gradually taking place in most of Europe. National representative samples were collected using a quota sample design based on gender, age classes, macro-area of residence, and education. Overall, we have a sample of 7,579 respondents, with national sample sizes varying between 1,033 and 1,169.⁵ All seven countries are consolidated Western democracies (Voeten, 2017). The choice of countries in which the survey was conducted was guided by the aim of capturing dynamics across both Southern and Northern European Member States. Furthermore, it also aims to capture dynamics across countries that, while severely hit during the first wave of the pandemic, varied in terms of the stringency and nature of their lockdown measures, with some implementing stricter and/or earlier policies than others. Generally, the choice of adopting a cross-national research design allows us to consider the relationship between performance evaluations, personal exposure and threat perceptions, and political orientations and democratic support more widely. This is given not only because of the uneven distribution of the pandemic, but also due to the heterogeneity in political factors, the conditions of the national economies before the crisis, and the different government responses to the crisis (Hale et al. 2020).

Regarding the measurement of the dependent variable, instead of using specific democratic performance evaluations as a measure of specific support (Inglehart and Welzel, 2005), we adopt a diffuse measure to tap into respondents’ general preference for their political regimes. Similar to questions used by other surveys (e.g., World Values Survey, the European Social Survey and Barometer Series), we ask respondents whether democracy is always preferable, or whether deviations to authoritarianism can be made in certain circumstances/the regime does not matter.⁶

Figure 1 depicts the distribution of the regime support variable across the seven countries in our sample. Unsurprisingly given that these countries are consolidated democracies, people who think democracy is always preferred far outnumber those who are open to authoritarian alternatives. However, a non-trivial share of people is open to autocratic alternatives with some variation across countries. For instance, roughly 30 per cent of Italian and French respondents are open to autocratic alternatives, making them the most skeptical towards democracy among the countries in our sample. In comparison, the broadest consensus in terms of democratic regime support is found in Germany and Spain.

Figure 1. Distribution of support for democracy by country.

To measure respondents’ evaluations of the government’s response to the pandemic and the resulting economic fallout, we rely on two questions. The first question is about the overall effectiveness of the measures taken by the government in the public health and the economic domains, respectively.⁷ Second, in addition to overall effectiveness, another important aspect of evaluating governments’ responses concerns how respondents view the timeliness of the measures. Therefore, we introduce two additional questions asking respondents whether the government acted too early or too late with regard to the health and economic measures. To measure respondents’ agreement or disagreement with the government’s response from this timing angle, we create two dummy variables (for the health and for the economic aspect) differentiating those that believe the government acted early from those that believe that the government acted at the right time/late⁸. The reason why we combine “government acted at the right time” and “government acted late” into one category is that both categories imply a general agreement with the measures adopted by the government. Both categories indicate that respondents think the government either did the right thing at the right time or should have done it earlier. “Government acted early,” on the contrary, implies a critical attitude towards the measures as respondents believe such measures were not needed at that particular point in time.

In order to test our hypotheses regarding socioeconomic exposure and subjective threat to the public health emergency and the economic fallout, we utilize several different indicators. For socioeconoomic exposure, the first one is age, recoded into three categories: under 35, between 35 and 60, over 60. Differentiating between the three generational cohorts is crucial, because the risk for severe illness with COVID-19 rises with age, with older adults at highest risk as clinical data have shown since the outbreak of the pandemic. In Europe, roughly 94% of fatalities are concentrated in the population over 60 years of age (Natale et al, 2020). Therefore, citizens in the older cohort face higher levels of exposure to the pandemic than their younger counterparts. The second indicator to capture socioeconomic exposure is self-reported income change due to the pandemic. We differentiate between those respondents that experienced no/minor loss in work income and those who experienced major loss in work income or were unemployed. We consider being unemployed as an equally dire situation with suffering major work income losses because the unemployed have no work income to start with and the pandemic made the prospects of finding one even bleaker. Additionally, we also control for whether respondents indicate as having children in their household as the lockdown regulations in many of the countries included in our sample put parents in particularly difficult situations having to juggle home schooling, working remotely or not working at all. For operationalizing subjective threat perceptions of the economic/health aspects at the egotropic or sociotropic levels, we use four variables originally ranging from 0 (no threat) to 10 (a major threat), recoded into three levels: a minor threat, a mild threat, and a major threat, for simplicity.

Finally, with regard to political orientations we rely on three indicators. For our argument on “political winners and losers,” we utilize the vote recall of respondents. We then classify respondents into government voters on the one hand, and opposition voters (and non-voters) on the other hand. For our argument on “ideological winners and losers,” we firstly took respondents’ left-right positions into account and recoded a variable with 11-level left-right scale into three camps, i.e., left, center and right. Another ideological indicator we rely on is Euroscepticism. More specifically, we use the question on vote recall to distinguish respondents who voted for mainstream parties from those who voted for Eurosceptic parties.⁹ Finally, we also control for political sophistication as citizens might have a more or less clear idea of what democracy means, which might, in turn, affect their preference for it.¹⁰

Our empirical analysis uses logit regressions with fixed effects to control for between-country differences. While in the previous section of the paper we recognized the importance of institutional factors in explaining democratic regime support, given that our study relies on a single point in time and seven countries, it does not allow for a direct, systematic investigation of such factors. In light of this, we control for such institutional factors via the country fixed effects which capture these between-country differences, while focusing the paper on the individual-level factors that affect anti-democratic attitudes above and beyond these institutional differences. Furthermore, while we acknowledge that under certain circumstances, hierarchical models are recommended when there is a nested structure of the data (respondents nested in countries), this both falls outside of the theoretical scope of our paper, as our emphasis is on micro-level determinants of regime support and is empirically unfeasible due to the small sample size of countries (7 countries only)¹¹. Nevertheless, as a robustness test, we also present results by individual countries in the online Supplementary Material Table A4 and engage with any discrepancies in the results section of the paper.

Results

Table 1 shows the results of four logit models in which we analyze how the three groups of variables discussed above impact on democratic support during the COVID-19 crisis. Model 1 estimates a model with only policy evaluation as predictors, Model 2 estimates a model with socioeconomic characteristics, threat perceptions, and political sophistication, Model 3 includes only political indicators, and Model 4 includes the full batch of independent variables. Most of the independent variables are significant in separate models and remain significant in the full model (Model 4) except political sophistication, voting for the opposition, and sociotropic health threat perceptions. In the following, we interpret the results using the fully specified Model 4 followed up by plots of predicted probabilities for the statistically significant predictors.

Table 1. Logit models for democratic regime support¹²

Note: *p<0.1; **p<0.05; ***p<0.01

Let us begin with the perceived overall effectiveness of the government’s response. Figure 2 plots the predicted probability of supporting democracy by perceived effectiveness in the health and economic domain, based on Model 4 in Table 1. We can see that the baseline hypothesis is confirmed as the expected upward trend suggests a consistently positive association between the two measures. There is a positive and statistically significant association between policy evaluations in the health domain and support for democratic rule. A one unit increase in the effectiveness scale increases the odds of being supportive of democracy (versus being open to an authoritarian alternative) by a factor of 1.08¹³. People with positive evaluations on the government’s response to the pandemic are marginally more likely to maintain support for the democratic system compared to citizens with more critical evaluations of the response. On the contrary, the coefficient of the perceived overall effectiveness in the economic domain is not significant and negatively signed in both the baseline Model 1 and the full Model 4 (decrease in the odds of support by a factor of 0.9). This result is also fairly robust when looking at individual countries (see online Appendix Table A4), as perceived effectiveness in the health domain being significantly associated with democratic regime support in four of our seven (Sweden, Germany, Netherlands, and France) countries and having positive coefficients in all others. The contrasting result lends clear support to hypothesis H1c which anticipated that health evaluations would exert a larger impact on democratic regime support than economic evaluations, indicating that the impact of domain-specific evaluations is a function of salience of the respective domains. It appears that the temporary sacrifice in basic civil liberties does not reduce citizens’ support for democratic regimes if they see the measures as effective responses to the pandemic.

Figure 2. Overall effectiveness evaluations effects.

When it comes to the timing of government response, the effects of evaluations in both the health and the economic domains are statistically significant and substantively large (see Figure 3 and Model 4). Those who consider the measures too early appear to be less inclined to support democracy. In contrast, those who consider measures timely or too late are more likely to support democratic rule. Again, the effect is more pronounced in the health domain, supporting H1c. Figure 3 illustrates the results graphically. We can see that those who consider the measures too early have a much lower predicted probability of supporting democracy compared with those who consider the measures timely or late, and the difference between the two groups is much more prominent in the health domain (about 0.15) than in the economic domain (about 0.05). As with effectiveness perceptions, this result is also fairly robust across individual countries with negative associations between those that consider the health measures too early and support for democracy in all seven countries (see Appendix Table A4). According to these results on the effectiveness and timeliness of the measures, citizens who evaluate a government’s response positively tend to be less open to authoritarian regimes compared to citizens who negatively evaluate the government’s response.

Figure 3. Timing effect.

Turning to the impact of threat perceptions, the models on socioeconomic exposure offer somewhat less clear-cut results. On the one hand, contrary to our expectations in H2a, major losses in income and unemployment due to the pandemic are not significantly associated with democratic regime support. On the other hand, age and having children do show significant associations, but in opposite directions. While respondents having children are significantly less likely to support democracy¹⁴ as expected by H2a, older age cohorts are more likely to do so. Figure 4 shows the predicted probabilities of supporting democracy among respondents who differ in their age. In the graph on the left, the results show that contrary to our expectations, older age groups, despite being the most vulnerable, are also the ones more inclined to support democracy. Ceteris paribus, people in the oldest group (60 and above) have the highest predicted probability of supporting democracy (0.85), while their younger counterparts (less than 35 years old) have a significantly lower predicted probability at 0.73. One possible reason why this result contradicts our hypothesis is that other effects might overwhelm the socioeconomic exposure effect, such as an age-related cohort effect. Recent studies on democratic (de-)consolidation argue that younger generations are more open to alternatives to democratic governments, such as expert or military rule, and that these younger cohorts, especially in established democracies, value democratic principles or practices less than the older generations (Denemark et al, 2016; Wuttke et al, 2022).

Figure 4. Age cohort.

Proceeding to the effects of subjective threat perception, we now test whether citizens with higher levels of subjective threat perception to the public health emergency and the economic fallout are less likely to support democratic regimes (H3b). Additionally, we examine which of the two competing perspectives on subjective threat, if either, holds: is it egotropic considerations that move pro-democratic sentiment or on the contrary, is it sociotropic concerns that drive people’s preferences on regime types? Figure 5 provides some answers to these questions. Starting with the egotropic considerations, we can see from the upper panels of Figure 5 that those who perceive a higher level of personal threat (either in the health or the economic domains) have a lower predicted probability of supporting democracy. When comparing the health domain with the economic domain, we find stronger effects for the latter when it comes to egotropic threat. The estimated probability of support for democracy is lower (0.77) among respondents who think the pandemic poses a major threat to household finances compared to those who only perceive a minor financial threat. The result of the health domain is less clear-cut than that of the economic domain but still presents a downward trend. Those who see only a minor personal health threat have a higher probability of supporting democracy (0.83) than those who see a mild and a major threat (0.79 and 0.80, respectively). This lends support to our hypothesis 3b, confirming that egotropic considerations indeed move pro-democratic sentiment, and citizens with higher levels of subjective threat perception are less likely to support democratic regimes.

Figure 5. Subjective threat effect.

On the contrary, the effect of sociotropic threat on democratic support does not mirror that of the egotropic one. As can be seen from the lower panel of Figure 5, the probability of supporting democracy is lower but not significantly so among those who see the pandemic as a minor threat to public health compared to those who see it as a mild public health threat and those who see it as a major threat (around 0.78 compared to 0.81 and 0.80, p > 0.05). In the economic domain, the results are similar but more clear-cut. Those who perceive a major threat to the national economy have the highest support for democratic regimes (0.81), while those who perceive a minor economic threat have the lowest level of support (0.70). This means that, unlike the threat to one’s personal health and finances (egotropic concerns), a perceived threat to the general population and the national economy (sociotropic concerns) has a positive effect on people’s support for democracy.

Last but not least, with regard to “winner-loser status,” we can see from the first plot in Figure 6 that the difference in the predicted probability between groups (political winners and political losers) is rather small. On average, respondents appear to be advocates of democracy (probability above 0.80), irrespective of whether they voted for opposition parties or governing parties. Therefore, our H3a is clearly disconfirmed. However, when turning to the ideological winner-loser dichotomy, we can see from the second and third plots in Figure 6 that a strong pattern emerges. The differences in predicted probability between ideological winners and losers are substantively large. Right-leaning respondents have a much lower probability of supporting democracy (less than 0.75) compared with those who are on the left (0.88) and at the center (0.80). Similarly, those who voted for Eurosceptic parties have a lower probability of supporting democracy (0.76), than those who voted for mainstream pro-European parties (0.83). These results suggest that political winner-loser status matters less than ideological winner-loser status both in terms of Euroscepticism and in terms of left-right ideology. Furthermore, the results are strongly confirmed in our robustness tests for all individual countries (see Appendix Table A4). Hence, H3b receives strong empirical support from the data.

Figure 6. Winner-loser effect.

Conclusions

A consolidated democracy must rely on the support of the vast majority of its citizens. The current public health crisis gives rise to a unique paradox concerning democratic governance: if output legitimacy leads to diffuse support for democracy (Scharpf, 1997), what happens when higher output performance is inevitably associated with the restriction of democratic rights? The first wave of the COVID-19 pandemic witnessed restrictions of civil liberties and policymaking without essential democratic debate. Does the temporary suspension of democratic governance aiming at saving lives reduce democratic regime support, or does it strengthen the legitimacy of democracy by underlining their resilience to unexpected shocks? Besides, do the usual suspects explaining regime preference in normal times still play a role during a national emergency?

With these questions in mind, this study set out to explore the individual-level determinants of support for democratic governance during the COVID-19 pandemic using data from an original cross-national survey fielded in seven European countries in June 2020. In doing so, our contribution expands on previous empirical studies on the link between output legitimacy and democratic regime support by applying it in multiple and novel domains of policy performance under a unified framework. We believe that our results provide important hints to the future of democratic stability in Europe and more generally, to the prospects of democratic resilience in undemocratic times.

The findings suggest that some of the democratic sacrifices during the pandemic notwithstanding, people with positive evaluations on the government’s response are more likely to maintain support for the democratic system compared to citizens with more critical evaluations. In other words, the temporary sacrifice in basic civil liberties does not reduce citizens’ support for democratic regimes if they see the measures as effective and adequate responses to the pandemic. Also, our results show that people weighed responses to the health domain more heavily than that in the economic domain, at least at the end of the first wave of the pandemic when our survey was fielded. This suggests that the link between output legitimacy and democratic support varies across domain-specific evaluations and the source of this variation might be the salience of the respective domains.

When it comes to other individual-level factors, we show that it is mostly subjective threat perception that moves pro-democratic sentiment, especially in the economic domain. As to whether these concerns are egotropic or sociotropic, we uncovered two opposing patterns: while citizens with higher levels of egotropic considerations are less likely to support democratic regimes, sociotropic concerns have a positive effect on citizens’ support for democracy. Additionally, our results suggest that ideological losers are less likely to support democracy, while political loser-winner status does not alter democratic ideals.

While our results provide important insights into the drivers of democratic support in times of crisis and its link to output legitimacy, they suffer from a number of limitations that offer fruitful avenues for future research agendas. One such limitation is related to the limited country sample: seven major European countries. Because of this, we resorted to controlling for unaccounted country-level factors through fixed effects while verifying the robustness of our results in the individual countries. While the results on the relationship between output legitimacy, socioeconomic status, political orientations, and democratic regime support presented here are highly robust in individual countries (as shown in Appendix Table A4), they also show some significant country differences in democratic regime support, even after introducing all these various independent variables in our models. Specifically, the fixed effects (see Appendix Table A2 and Figure A2) show that French and Italian respondents support democracy significantly less than Spanish respondents, who appear the most supportive after controls. One tentative explanation for this difference could lie in Spain’s relatively late democratization process, which entails that the foundations of the democratic regime’s institutions never quite disappeared from the political agenda. Due to this, Spain may not have yet started to experience an erosion pattern in pro-democratic attitudes like the one indicated by Claassen (2020b). Generally, as pointed out in the theoretical section of the paper, this further underscores the importance of contextual and institutional factors in explaining democratic regime support on top of the individual-level factors we have focused on here. Nevertheless, due to the limited sample size, we leave this result as a matter of speculation outside the scope of our paper, as we could not explore the contextual characteristics that might drive these differences in a more systematic fashion. Other studies with a wider geographical coverage could focus on the ways in which such contextual characteristics such as institutional structures, democratic legacies, civic society development, etc., interact with the individual-level predictors that we used in this study. Also, the results can only speak for Europe, given that we do not have data on other established democracies in the democratic world. We hope specialists working on other countries can collect and analyze similar data and more systemically test the cross-national validity of our hypotheses.

Secondly, our study only deals with the first wave of the pandemic, so we could only provide an insight into a specific time-point of the crisis. Therefore, we could not consider any temporal change in people’s attitudes. Meanwhile, Europe has already endured the hit by the second and third waves of the pandemic. It is very well possible that the effect of output performance on democratic support changes in size or even in the direction as different dimensions of performance increase or decrease in salience.

Finally, we need to note the possible endogenous relationships between support for democracy and policy evaluations, threat perception, and ideological orientations. It might well be the case that not only do these factors contribute to the level of democratic support, but that democratic support meanwhile also drives policy evaluations as well as threat perceptions and ideological positions. Considering the well-established link in the literature between output legitimacy and democratic support, we considered the direction of causality as a given here. However, further research could focus more on disentangling causality in this relationship through longitudinal or experimental designs. All in all, while by no means causally identified, we do believe that our study does highlight the important role of individual-level factors in the democratic support-output legitimacy nexus.