Who rallies around the flag? Evidence from panel data during the Covid-19 pandemic

ABSTRACT

Recent studies on political trust during the Covid-19 pandemic diagnosed a rally-around-the-flag effect leading to exceptionally high levels of trust in politics. While this finding has been established over various country contexts, our understanding of the precise dynamics behind the rally effect remains limited. In this paper, we argue that socio-demographic characteristics, in particular age differences, as well as pre-existing trust levels moderate the extent of the rally effect. Using individual-level panel data from the Netherlands, covering the time before and during the first Covid-19 wave, we show that the rally effect is particularly pronounced among older individuals, while it is absent among the young. Furthermore, we find a catch-up effect among the more distrusting parts of the population, such as populist supporters and low-income earners, who seem to largely drive the rally effect during the initial stage of the pandemic. This shows that the extent of the rally effect is conditional on socio-demographic characteristics, pointing towards the role of group risks and pre-crisis trust differentials in shaping people’s response to an existential threat.

Introduction

Many policy measures to contain the Covid-19 pandemic impose strong constrains on individual freedoms and rights. Consent and compliance with these pandemic policies critically hinge on citizens’ confidence in political institutions (see Marien and Hooghe 2011; Marien and Werner 2019). Therefore, a basic level of political trust is a fundamental prerequisite for a successful policy response against Covid-19 (Charron, Lapuente, and Rodríguez-Pose 2022). In this paper, we study public opinion dynamics at the start of the pandemic to provide insights into the immediate societal response to a collective crisis event. This analytical interest is distinct from research that tries to understand the evolution of political support over the course of the pandemic, as the relevance of different theoretical mechanisms most likely varies over time. While public opinion at the beginning of the pandemic appears more emotionally driven (De Vries et al. 2021; Jørgensen, Bor, and Petersen 2021; Schraff 2021), long-term dynamics might relate more to rational evaluations of government performance (Bechtel and Hainmueller 2011; Colaresi 2007; Johansson, Hopmann, and Shehata 2021). In this paper, we contribute to the first strand of literature, while the latter strand would require a different set of theoretical and empirical tools.

At the outset of the pandemic, public support for government action has been exceptionally high across many democratic societies. Various studies suggest a sharp increase in citizens’ diffuse political support, diagnosing a rally-around-the-flag effect with surging levels of political trust during the first wave of the pandemic (Devine et al. 2021). While research over manifold country contexts confirms this finding (Baekgaard et al. 2020; Bol et al. 2021; Esaiasson et al., 2021), we still have a limited understanding of how this rally effect actually works. For instance, cross-sectional evidence from Germany suggests that personal circumstances, such as family situation, condition trust at the start of the pandemic (Rump and Zwiener-Collins 2021). Furthermore, a comparative study by Galasso et al. (2020) shows large gender differences in Covid-19 health beliefs and behavior. These findings already point towards potentially important individual-level differences in public opinion dynamics during the arrival of the pandemic.

We extend this research by providing more systematic evidence on what types of individuals rally around the flag using high quality panel data. This allows for important insights into how political trust is polarized across society during a pandemic. Moreover, studying effect heterogeneity in the rally effect allows us to understand how extensive the mobilization around political institutions was and which social groups might have remained more skeptical. In short, in this paper, we therefore develop a more comprehensive understanding of who rallies around the flag in times of a global health crisis.

We address this question by analyzing individual-level panel data among a sample of 1,832 respondents covering the time before and during the first Covid-19 wave in the Netherlands. These data allow us to track individual-level changes in political trust when the first Covid-19 wave hit and thereby make possible to identify individual-level differences in people’s responses to the pandemic. Our data havethe advantage that they comprise a special wave within the Longitudinal Internet studies for the Social Sciences (LISS) panel, which we fielded during the arrival of the pandemic in the Netherlands and combined with data from pre-pandemic waves.

Theoretically, we argue that individual reactions to the pandemic vary because of age differences. We suggest that this might be due to variation in the health and economic risks induced by the crisis. Moreover, we propose that the rally effect differs by pre-existing trust levels, as people start from very different baselines of trust when entering the pandemic. Therefore, socio-economic groups with different pre-crisis trust levels will vary in their response to the arrival of the pandemic.

The results of our analysis lend credence to these arguments. First, we show that the rally effect is strongly pronounced among older respondents, while it is absent among the young. We suggest that this might be explained by the asymmetric distribution of costs and benefits from pandemic policies across age groups. Second, we find that the rally effect is strongest among low-income individuals and those who are the least satisfied with the current state of the economy. We theorize that this might be due to an anxiety driven rally effect crowding out conventional explanations of political trust (Schraff 2021; Van der Meer and Hakhverdian 2017). Third, the rally effect appears to be strongest among those individuals who support populist parties and enter the pandemic with lower levels of pre-crisis trust. Building again on the mechanism of anxious arousal, we propose that increasing anxiety during the dawn of the pandemic leads low-trust individuals to abandon their skeptical priors and catch up in their trust levels.

The remainder of this paper proceeds in the following way. After reviewing the existing literature identifying a rally effect during the pandemic, we illustrate how this dynamic might differ at the individual-level. We then present our panel data, method and results. We conclude by outlining some broader implications of our findings, suggesting that the convergence of trust we observe points to the malleability of public opinion during a crisis event. However, we also highlight the difficult situation of the young and underline that our evidence presents first age-based cracks in the societal response to the pandemic, providing a preview of the problems complicating a successful response to the Covid-19 crisis at later stages.

Political trust during the pandemic: a rally effect

From a classical perspective, rally-around-the-flag dynamics have been studied with regard to national political leaders in times of international conflict. Here, international events trigger a short-term increase in support for incumbent executives providing them with a basis to take immediate action. Rally effects are thus theorized to arise when events are international in nature, directly involve national leaders, and are “specific, dramatic, and sharply focused” (Mueller 1970, 21). Explanations for this dynamic may be summarized along three main schools (Hetherington and Nelson 2003). Firstly, the patriotism school postulates that in times of an outside threat, citizens rally around national symbols such as incumbent leaders (Feinstein 2018; Kobayashi and Katagiri 2018; Lai and Reiter 2005; Lee 1977). Secondly, an alternative explanation relates to anger, suggesting that the rally effect is fueled by a desire for retaliation (Lambert et al. 2010). Thirdly, a different set of explanations follows the opinion leader school (Brody 1991; Brody and Shapiro 1989; Edwards and Swenson 1997; Groeling and Baum 2008). This literature departs from the observation that during a crisis event executives essentially possess an information monopoly (Brody and Shapiro 1989), which leads the media to report uncritically and opposition leaders to remain silent or even supporting the government (Baker and Oneal 2001).

We follow the classical literature on the rally effect by investigating the rise of political support in response to an immediate collective threat. However, our definition of the rally effect is broader, as we do not focus on the evaluation of the executive leader, but on a more general measure of systemic trust (Easton 1975).¹ Comparative evidence on the rally effect during the pandemic shows that specific and diffuse political support moved in parallel (Bol et al. 2021).

Research on the impact of the Covid-19 crisison political support has documented rising levels of trust in political institutions at the onset of the pandemic (e.g. Baekgaard et al. 2020; Bol et al. 2021; Schraff 2021). This rally effect was sizable during the first wave, but has also been shown to be short lived, disappearing as the pandemic became normalized (Johansson, Hopmann, and Shehata 2021; Kritzinger et al. 2021). Existing studies on the Covid-19 rally effect differ in important ways from the more classical works. To start, while pandemics are international events, they are in many ways different from wars, which are the main triggers of rally effects in the classical literature (see Mueller 1970, 1973). Importantly, earlier work has demonstrated that non-war crises often seem not to induce rally effects (Lai and Reiter 2005). Moreover, whereas classical works focus on an increase in specific political support for incumbent leaders, many studies during the pandemic look at trust in political institutions as an indicator of diffuse political support. In relation to this, although many of these works particularly focus on trust in government, the rally effect has also been shown to expand to regime institutions that are not directly involved in crisis management, such as parliaments and courts (Baekgaard et al. 2020).

Apart from this, explanations of the Covid-19 rally-around-the-flag dynamic differ from the classical patriotism and opinion leadership schools. Fundamentally, it is often argued that the rally effect during the pandemic is based on rational evaluations of the political system and its performance. For instance, Bol et al. (2021) argue that lockdown measures across Europe have found approval among voters, rewarding political institutions with increased trust. In this sense, this literature follows a more general argument on how citizens rationally reward political institutions for swift crisis relief (Bechtel and Hainmueller 2011). This mechanism is further substantiated by evidence from panel data and survey experiments conducted in Spain, suggesting that people tend to switch to a strong preference for unified and technocratic governance in reaction to the pandemic (Amat et al. 2020). Therefore, these studies maintain that the rally effect is rooted in a rational response of the public to support policy measures issued by governments to fight the crisis.

In contrast, another set of recent works agrees with the empirical diagnosis of a rally effect, but does more strongly rely on psychological arguments to explain it. This line of reasoning has received increasing attention as anxiety appears to play an important role in people’s response to the pandemic (Elmer, Mepham, and Stadtfeld 2020; Salari et al. 2020; Tabri, Hollingshead, and Wohl 2020). Moreover, some accounts suggest systematic differences in Covid-19 risk perceptions across countries and individuals (Dryhurst et al. 2020). Following that, some rally effect studies build on psychological mechanisms to argue that the extent of anxious arousal evoked by the pandemic shapes the way people rally around political institutions (Esaiasson et al., 2021). For example, looking at the relationship between negative emotions and public opinion dynamics in Switzerland, Erhardt et al. (2021) find that the rally-around-the flag effect during the pandemic seems to be driven by fear, while heightened feelings of anger appear diminish citizens’ trust in political institutions. In a similar vein, using data from Germany, Dietz et al. (2021) point towards fear of Covid-19 as a major explanation of the rally effect. This emotional explanation is further supported by the finding that standard explanations of political trust, such as economic perceptions, lost relevance as the pandemic spread (Schraff 2021), and by evidence that government support already increased as first lockdowns were implemented abroad (De Vries et al. 2021).

Following these theoretical accounts, we argue that one should not take for granted a homogenous rally effect across society. Indeed, first evidence suggests large gender differences in health beliefs and behavior (Galasso et al. 2020), as well as conditional effects of personal circumstances on trust during the pandemic (Rump and Zwiener-Collins 2021). Furthermore, research on rally effects during military conflict provide additional support for the relevance of effect heterogeneity. Here, pre-crisis levels of political trust seem to substantially moderate the size of the rally effect itself (Chatagnier 2012). In light of this, we argue that differences in socio-demographic characteristics and pre-crisis levels of political trust shape the size of the rally effect during the pandemic.

Who rallies around the flag?

Intensity of the pandemic

Timing is a crucial component to understand citizens’ response to a sudden crisis event. The arrival of the Covid-19 pandemic emerged in a dynamic process, posing the question as to when a potential rally effect might kick in. Following previous literature, we expect that the exponential growth in Covid-19 cases provides a good indication of the intensity of the pandemic for understanding the rally effect (Schraff 2021). The pressure to rally around political institutions was weaker in the beginning as the first Covid-19 cases emerged, but increased as cases were rising exponentially. We therefore approach the rally effect as an increase in political trust due to rising Covid-19 case numbers.

Age groups

One important but so far overlooked socio-demographic moderating the rally effect during the pandemic is age. Consequences of the Covid-19 pandemic strongly vary across age groups. The young and the old experience different costs and benefits from the pandemic and the societal pandemic response. We therefore propose that older citizens have a stronger incentive to rally around political institutions as the pandemic hit.

Older people face substantially higher health risks from Covid-19, in particular due to the chronic diseases that become more frequent with higher age (Zheng et al. 2020). Therefore, older people might be more willing to bear the costs of preventive Covid-19 policies, such as restrictions to public life. Because of that, older generations see a stronger need to rally around the political institutions that provide these preventive measures. Young people, in contrast, face a much lower risk of serious illness from the virus. This is reflected in lower health risk perceptions of the young regarding Covid-19 (Nisa et al. 2021). Moreover, younger citizens might perceive higher costs from lockdown measures since they are usually more mobile and socially active. In this regard, school closures or the restriction of leisure activities constitute a major disruption for young people’s lives. Indeed, young people are frequently hit harder by lockdown measures, which has been shown to lead to increased psychological distress among this age group (Elmer, Mepham, and Stadtfeld 2020; Justo-Alonso et al. 2020).

This asymmetry in the costs and benefits of pandemic health policies should shape young and old people’s evaluations of the political system. Of course, all people face a high level of uncertainty and a potential for anxious arousal in the face of the pandemic. It is in the interest of the old and the young that the health care system does not collapse. Yet, older people have the strongest incentive to rally around political intuitions as they face a higher health risk, which requires effective collective action. Younger people, on the contrary, face much more mixed signals. Health risks are less pronounced for the young, while the perceived social and psychological costs are arguably higher. It therefore is likely that the young are less enthusiastic about Covid-19 containment measures and feel less inclined to cling on to political institutions. To be sure, we do not argue that the young do not show solidarity or that young people are actively opposed to the Covid-19 prevention policies by default. Instead, we merely suggest that the rally effect should be less pronounced among the young, as they have fewer incentives to attach themselves to political institutions as a lifebuoy.

H1: As Covid-19 cases rise, older people rally more strongly around political institutions compared to the young.

Economic groups

Covid-19 has led to severe negative economic consequences. Therefore, the distribution of economic risks across the population may influence support for the political system as crisis hits. We see two competing scenarios on how economic risks might shape the rally effect. On the one hand, the concentration of economic costs within certain groups might lead these people to withdraw support for emergency policies. The rally effect, therefore, should be absent in economically hard-hit sectors or occupations. Of course, the immediate individual-level economic costs of the pandemic might not be that pronounced as it may take a while until negative effects reach individual households. However, it was rather clear from the beginning that the measures to fight the pandemic would entail high economic costs (Baekgaard et al. 2020). Moreover, research during the early stages of the crisis has documented a considerable increase in economic anxiety (Fetzer et al. 2020). Thus, individuals who have reason to expect financial hardship might be less enthusiastic about the emergency policies.

On the other hand, economic costs could still be too diffuse at the beginning of the pandemic. This might explain a rally effect across all economic groups. Indeed, recent evidence suggests that economic performance perceptions have lost explanatory power for political trust as the pandemic spread (Schraff 2021). This is an interesting insight, as economic position and perceptions of economic performance are central determinants of political trust (Van der Meer and Hakhverdian 2017). Therefore, if an anxiety driven rally affect crowds out the usual cognitive evaluations, we should expect a strong increase in trust, especially among individuals with lower income and negative evaluations of economic performance, which are often shown to exhibit lower levels of trust (Gallego 2016; Hetherington 1998; Schraff 2019). Thus, if economic evaluations indeed lose relevance during the pandemic, the less trusting economic groups might show the strongest rally effect, catching up to the better off economic groups.

H2a: As Covid-19 cases rise, people in more adverse economic conditions rally less around political institutions.

H2b: As Covid-19 cases rise, people in more adverse economic conditions rally more around political institutions.

Pre-existing trust levels

Finally, the level of trust with which individuals enter the crisis might play an important role in itself (Chatagnier 2012). If the rally effect indeed works through general human reactions, such as anxious arousal, we should expect those who enter the pandemic with a more skeptical baseline to exhibit the strongest rally effect. In other words, those with low levels of pre-pandemic trust should show a catch-up effect as the pandemic unfolds. The extent of this moderation effect might in turn be context dependent. In a highly polarized context, such as the United States, a rally effect might be suppressed due to deeply entrenched distrust, while in a less polarized context people might be able to overcome some of their distrust as the pandemic hits (Altiparmakis et al. 2021). In this regard, the Dutch case is more favorable to the latter mechanism, as moderate polarization could favor a catch-up effect among the more distrusting parts of the population (see Wagner 2021).

H3: As Covid-19 cases rise, people with lower pre-existing political trust rally more strongly around political institutions.

Our argument on the moderating role of pre-existing trust levels reverberates in the literature about partisan responses to the Covid-19 pandemic. On the one hand, part of the rally effect literature suggests that the pandemic has unified political camps (Anderson and Hobolt 2020; Merkley et al. 2020). If at all, heterogeneity in the rally effect on political trust has been found to be small (Esaiasson et al., 2021). On the other hand, behavioral evidence suggest that partisanship does shape individual compliance with public health measures. Studies from the United States show that support for and compliance with Covid-19 policies varies across Republicans and Democrats (Allcott et al. 2020; Bhanot and Hopkins 2020; Gadarian, Goodman, and Pepinsky 2021; Grossman et al. 2020). It is therefore still unclear how partisanship shapes the rally effect. In light of this, our argument suggests that these mixed findings can be addressed by considering the overall level of political polarization and the moderating role of people’s pre-crisis institutional trust.

Data and method

To test our expectations, we rely on nationally representative individual-level panel data, collectedamong 1,832 respondents before and during the first Covid-19 wave in the Netherlands. In terms of generalizability, we believe that the Netherlands present a good setting for our study as comparative research on the rally effect has shown that the Dutch case exhibits similar public opinion dynamics to the rest of Europe (Bol et al. 2021). Furthermore, policy measures implemented by the Dutch government were comparable to those of other countries in Northwestern Europe, making a similar public response more likely (Hale et al. 2021).

Our survey data for the pre-pandemic wave come from December 2019 and January 2020, establishing a clear pre-pandemic baseline. The data for the pandemic wave come from March 2020, the month in which the first Covid-19 wave hit the Netherlands.² The pandemic wave was fielded as a short special survey among a representative subset of LISS panel participants. This data structure allows us to analyze individual-level changes in political trust from the pre-pandemic to the pandemic wave, establishing a strong benchmark to assess the impact of the pandemic. Compared to existing survey experiments on the political consequences of Covid-19, our panel data provide a more realistic setting by capturing individual-level changes in the real world, instead of an artificial experimental context. Moreover, the panel structure improves the quality of inferences compared to existing studies that track single survey responses over the fieldwork period.³

We focus on respondents’ political trust in the national parliament using an 11-point scale, where low values indicate low trust, while high values indicate high trust.⁴ In this regard, we diverge from the classical works on the rally effect by looking at a more diffuse object of political support than incumbent office-holders (Norris 2017). However, existing studies have found a rally effect in both, more diffuse and more specific political support (e.g. Bol et al. 2021), and have shown that the rally effect also expands to regime institutions not directly involved in managing the pandemic (Baekgaard et al. 2020). Besides this, institutional political trust tends to be a unidimensional constructwith people expressing a rather general feeling of trust which seems not to differentiate much between different types of institutions (Marien 2011).

We capture the impact of the pandemic by merging oursurvey data with Covid-19 statistics reporting the number of daily-diagnosed cases.⁵ As we merge Covid-19 incident numbers by day, they are, by design, always zero in the first wave. In the second wave, fieldwork stretches over the whole month of March 2020, and we therefore match the Covid-19 numbers with the reported day of the interview. In all models specified below, we use the cumulative number of Covid-19 infections.

Besides this, the data come with standard demographic variables including age, gender, education, income, and migration background. We classify age groups by younger (<=30), middle (>30 and <55), and older (>=55).⁶ For our investigation of economic groups, we rely on a measure of net household income, defining low-income earners as respondents falling within the first income quartile. We also use a question on respondents’ satisfaction with the state of the economy to approximate economic risk perceptions.⁷ We capture respondents’ pre-pandemic trust in the national parliament with the political trust variable in our pre-pandemic wave. Finally, as an alternative operationalization for pre-pandemic trust, we also facilitate people’s vote choice in the last general election. We focus on supporters for populist parties who can be expected to be more skeptical (Hooghe, Marien, and Pauwels 2011). We code individuals as populist supporters if they indicate a vote choice for the Forum for Democracy, the Party for Freedom, or the Socialist Party following the coding of the PopuList (Rooduijn et al. 2019).

Our results are based on linear regression models with random effects at the individual-level.⁸ We prefer random effects models, as our moderator variables are largely stable over the two waves. Moreover, random effects models are recommended for datasets with few observations per unit (Clark and Linzer 2015). Random effects models also allow us to include a set of temporally stable socio-demographic controls, such as education or migrant background. However, we provide two-way fixed effects models as a robustness check in the Appendix. The substantive findings are the same.⁹

To detect possible effect heterogeneity in the rally effect, our models use interactions between the standardized cumulative number of Covid-19 infections and indicators for respondents’ age (H1), income and economic satisfaction (H2a and H2b), as well as pre-pandemic trust levels and support for populist parties (H3). For H2a and H2b, we restrict the estimation sample to respondents with direct risk exposure, omitting respondents who derive their income outside the labor market, such as pensioners, pupils, or the unemployed.¹⁰ This leaves us with a sample of respondents who actually are at risk of losing out materially from the pandemic.

Results

Figure 1 presents the marginal effects of cumulative Covid-19 cases on political trust conditional on three different operationalizations of age shown in panel A, B and C.¹¹ Full regression output is provided in Table A13 in the Appendix. All interactions are statistically significant at the one percent level. In line with our arguments, the rally effect seems to be absent among younger and strongly pronounced among older respondents. With growing numbers of Covid-19 cases, older respondents substantially increase their trust in the national parliament, while political trust for younger respondents remains stable. Based on this, there is strong evidence in favor of H1. Older people strongly rally around political institutions once infection numbers surge. In contrast, younger individuals change little in their political trust. These findings also hold for alternative age operationalizations and under a fixed effects specification, presented in Table A7 and A9 in the Appendix. Overall, our panel data show that the rally effect substantially varies across age. Young respondents retain a moderately high level of trust but also do not increase their trust further as the pandemic hits. The rally effect among the middle aged is positive and significant, yet only oldest respondents’ political trust increases sharply in response to the Covid-19 crisis.

Figure 1. Marginal effects of cumulative Covid-19 cases on political trust conditional on different operationalizations of age with 95% confidence intervals.

Note: For full model results see Table A13 in the Appendix.

Figure 2 plots the marginal effects of cumulative Covid-19 infections on citizens’ political trust conditional on respondents’ household income (panel A) and economic satisfaction (panel B), as to investigate H2a and H2b. Full model results are provided in Table A14 in the Appendix. Both interaction terms are statistically significant at the one percent level. Looking at panel A of Figure 2, we see that political trust increases with rising numbers of Covid-19 infections across all income groups. However, while richer as well as poorer individuals seem to rally around the flag, they appear to do so at different rates. Individuals whose income falls within the first quartile exhibit a significantly stronger leap in political trust. Furthermore, as shown in panel B of Figure 2, the rally effect is more pronounced among respondents who are the least satisfied with the current state of the economy. These results are also robust to a fixed effects specification shown in Table A10 in the Appendix. This lends support for H2b, suggesting that the pandemic’s arrival unifies people’s trust in politics as economic groups with generally lower levels of political trust catch up as Covid-19 spreads. In this regard, these results illustrate that the pandemic led to converging levels of trust, rather than further polarizing existing economic divides.

Figure 2. Marginal effects of cumulative Covid-19 cases on political trust conditional on household income quartiles (panel A) and economic satisfaction (panel B) with 95% confidence intervals.

Note: For full model results see Table A14 in the Appendix.

We also tested for effect heterogeneity across alternative operationalizations of economic groups. H2a suggests that occupations or economic sectors hard hit by the pandemic might not rally around the costly Covid-19 policies. To give this argument a fair shot, we tested for interactions between the cumulative Covid-19 numbers and occupational/sectoral groups. The results are presented in Tables A15 and A16 in the Appendix. Overall, we do not find systematic evidence that the rally effect varies across occupational and sectoral groups. We do see that the rally effect is significantly less pronounced in the business services sector (including real estate) and is more pronounced among people employed in industrial production. Yet, we do not consider this to be strong evidence for systematic polarization across occupational or sectoral groups.

Finally, Figure 3 engages with H3 plotting the marginal effects of cumulative Covid-19 infections on citizens’ political trust conditional on their pre-pandemic trust levels (panel A) and support for populist parties (panel B). Full model results are presented in Table A17 and Table A18 in the Appendix. Looking at panel A, the marginal effects show that people with the lowest levels of pre-pandemic trust rally the strongest. This effect heterogeneity is sizeable. On average, low-trust respondents increase their trust in politics by around 1 point on an 11-point scale. In contrast, the rally effect ceases among respondents with a trust level of over 6.25 points. This finding squares well with the catch-up effect discovered among different economic groups above.

Figure 3. Marginal effects of cumulative Covid-19 cases on political trust conditional on pre-pandemic trust levels (panel A) and support for populist parties (panel B) with 95% confidence intervals.

Note: For full model results see Table A17 and Table A18 in the Appendix.

Furthermore, the catch-up dynamic seems not to be driven by a ceiling effect, where low-trust respondents only catch up as high-trust respondents have already maxed out on political trust in the pre-pandemic period. Panel A in Figure 3 shows that the effect already ceases long before the maximum value of 10 is reached. Also note, that we do not want to over-interpret the negative predicted values at the upper end of the x-axis, as the number of observations gets very low towards the end of the scale.

We also discover this catch-up effect when we look at support for populist parties as an alternative operationalization for pre-pandemic trust as shown in panel B of Figure 3. While political trust levels for both populist and non-populist voters rise with the Coronavirus spreading, the rate at which populist voters rally around the flag is significantly higher than that of non-populist voters, narrowing down the trust gap between both groups substantially. Table A11 in the Appendix shows that this finding also holds under a fixed-effects specification. Although the interaction coefficient just misses the five percent significance level, the marginal effects suggest a very similar effect as depicted in panel B of Figure 3. Overall, these results speak to the proposition of H3, implying that people with lower pre-existing trust levels exhibit the strongest rally effect.

In summary, our analysis shows that the rally effect varies substantially across individuals. Whereas younger respondents’ political trust remains unaffected by the crisis, older individuals strongly increase their trust in parliament as the pandemic hits. This could be explained by the asymmetric distribution of costs and benefits from the pandemic response among age groups. On the one hand, older respondents, who are confronted with higher health risks from Covid-19, might rally around political institutions to back preventive emergency policies passed to contain the virus. On the other hand, the rally effect could be absent among younger respondents because their risk of a severe Covid-19 infection is lower, while their perceived costs under pandemic policies are arguably higher. In addition to that, we find that individuals who earn less and are dissatisfied with the current state of the economy catch up with the more trusting parts of the population.

This refines our understanding of the rally effect during the pandemic in important ways. The rally effect is largely driven by people who are more distrusting of political institutions under normal times, giving up their skeptical stance in face of a collective societal threat. We show that pre-crisis levels of political trust strongly moderate the rally effect, as the low trusting parts of the population rallied around political institutions at large. This pattern is reflected in a range of effect heterogeneities uncovered, such as larger rally effects among populist voters, low-income earners, and economically dissatisfied people. This squares well with emotional explanations of the rally effect, which temporarily crowd out more conventional drivers of political trust. This, however, also implies that the rally effect during the Covid-19 pandemic was not a society-wide effect. Only specific groups have decided to take a leap of faith. Some, such as the young, have not moved in their institutional evaluations, running the risk of losing them as the pandemic drags on. Others have swung to a more trusting stance, such as low-income earners, creating the potential for a backlash as the pandemic becomes a polarizing issue again. Understanding these early dynamics in public opinion thus appears insightful to make sense of the unfolding polarization of public opinion we witness at later stages of the pandemic.

Conclusion

A successful response to Covid-19 fundamentally depends on citizens’ political trust, without which consent and compliance with policy measures containing the virus are put in jeopardy. Against this backdrop, research has documented a rally-around-the-flag effect, with exceptionally high levels of political trust at the beginning of the pandemic (Devine et al. 2021). This contrasts with earlier research on rally effects showing that they tend to be largely absent for non-war crises (Lai and Reiter 2005). We add to this literature by providing evidence for a heterogeneous Covid-19 rally effect, shedding light on what kind of people drive this dynamic.

In this paper, we argue that the rally effect varies over socio-demographic groups. First, we suggest that age groups have different incentives to rally around political institutions as the pandemic hit. Older respondents have reason to worry about a severe Covid-19 infection and therefore rally around political institutions for shelter. Younger respondents face lower health risks and higher costs from Covid-19 policies, and therefore do not rally around political institutions. Second, we argue that generally more distrusting people catching up to the rest of the population might drive the rally effect. We suggest that such a dynamic could be visible among low-income earners and individuals who are the least satisfied with how the economy is going. Alternatively, we propose that potential economic repercussions from the pandemic might attenuate the rally effect. Third, we argue that the trust baseline with which individuals enter the pandemic might play an important role in moderating the rally effect by leading to a rally effect particularly pronounced among low-trust individuals.

We test our arguments by using individual-level panel data on a nationally representative sample of 1,832 respondents covering the time before and during the first Covid-19 wave in the Netherlands. First, we show that the rally effect is strongly pronounced among older respondents, while it is absent among the young. This might be explained by the asymmetric distribution of costs and benefits from pandemic policies across age groups. Second, we find that the rally effect is strongest among individuals from low-income groups and those who are the least satisfied with how the economy is going. We theorize that this might be due to an anxiety driven rally effect crowding out cognitive explanations of political trust. Third, in line with the previous point, we find that the rally effect is strongest among individuals who support populist parties and enter the pandemic with a low-trust baseline. We propose that increasing anxiety during the dawn of the pandemic leads less trusting individuals to abandon their skeptical priors and catch up in their trust levels.

These findings are important in two primary ways. First, our results shed light on the dynamics of the rally effect, illustrating that people surrender their skepticism towards political institutions in view of a collective social threat. This provides for a silver lining during a period of uncertainty and upheaval. After all, if people from opposite ends of society manage to come together to address an urgent problem by means of collective action, a determined and ultimately successful response to handle the crisis becomes more likely. A catch-up effect among high-risk groups and low-trusting individuals, and not a general leveling up of trust among the whole population, however, more adequately describes the rally effect.

Second, on a less optimistic note, we find that the young have not increased their political trust at the start of the pandemic. This suggests that political trust among the young remains as polarized as in normal times, as our results provide no indication of convergence along existing societal divides. This underlines the difficult situation of the younger generation and hints at first cracks of the societal consensus. Unfortunately, our sample is not powerful enough to provide detailed analysis of the rally effect among young respondents. However, our findings suggest that this would be a highly relevant endeavor for future research.

Finally, there are some important limitations of our study. Our panel data is currently not suitable to study more long-term public opinion effects of the pandemic. However, the rally effect, which is the focus of this study, is most likely a short-term phenomenon that helps us to understand a society’s immediate response to a collective threat. Our study is well suited to investigate this, while other studies are needed to track the evolution of public opinion over the course of the pandemic as a whole.

Moreover, in contrast to much of the existing literature on rally-around-the-flag effects, we focus on citizens’ trust in parliament, instead of support for incumbent executives or ruling parties (Feinstein 2018; Kobayashi and Katagiri 2018; Kuijpers 2019; Lai and Reiter 2005; Singh and Tir 2018; Tir and Singh 2013). While a measurement of trust in government would be more in line with these works, our results demonstrate that the rally effect during the first wave of the Covid-19 pandemic extends to the parliament as an institution that was more indirectly involved in crisis management (see also Baekgaard et al. 2020). This is an important finding in itself, since compliance with Covid-19 policies does not only hinge on specific, but also on more diffuse kinds of political support (see Marien and Hooghe 2011; Marien and Werner 2019).

Lastly, the cost of using high quality panel data is the limited external validity. The size of the rally effect and the relevance of moderating factors most likely vary across countries. Future research should try to identify the role of contextual factors for the nature of the rally effect. Yet, our findings demonstrate the rather general point that people’s initial levels of trust at the start of the pandemic and socio-demographics matter for the size of the rally effect. By focusing on age groups, low-income individuals and populist supporters, we have worked with social groups that travel rather well across contexts. Furthermore, comparative research on the pandemic suggests that the public opinion dynamics in the Netherlands fit well into the overall picture across Europe (Bol et al. 2021). Therefore, we are confident that our findings carry some important implications for other country contexts and invite more studies of effect heterogeneity across different societal settings.

In this regard, other sources driving effect heterogeneity might be relevant too. For example, different media diets could have an impact on the rally effect as different information environments may paint rather different narratives of how the crisis is handled. Indeed, a first study from Switzerland documents that the increase in political trust following a lockdown was moderated by the consumption of mainstream news dampening the rally-around-the-flag effect (De León et al. 2022). Therefore, investigating the relationship between media consumption and effect heterogeneity in the rally effect may provide a promising avenue for future research.

Acknowledgements

The authors would like to thank three anonymous reviewers and the editor for helpful feedback on the manuscript. The manuscript also benefitted from valuable feedback by colleagues from the European Politics research group at ETH Zurich.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by Swiss National Science Foundation [grant number 186002].

Notes

1 A lot of the classical rally effect literature investigates presidential approval in the United States. Focusing on the executive leader might be especially important in a presidential system with strong executive powers. However, the distinction between government approval and institutional trust might be less pronounced in parliamentary democratic systems.

2 Data were collected from December 12, 2019 until January 28, 2020 (pre-pandemic wave) and from March 2, 2020 until March 31, 2020 (pandemic wave). For a detailed overview and descriptive statistics of our final samples see Table A1 to A6 in the Appendix.

3 The study by Esaiasson et al. (2021) uses individual-level panel data, but their first wave lies already in the beginning of the pandemic. We believe our panel has an even clearer baseline. Moreover, while Esaiasson et al. (2021) provide evidence on Sweden, we contribute evidence from the Netherlands, with a stronger focus on effect heterogeneity.

4 The wording of the political trust question differs slightly between the waves. Pre-pandemic: “Can you indicate, on a scale from 0 to 10, how much confidence you personally have in each of the following institutions?” [Dutch parliament] (0 = “no confidence at all”–10 = “full confidence”). Pandemic: “For the following questions, please give me an answer on a scale from 0 to 10, where 0 means you have no confidence at all, and 10 means you have a lot of confidence. How much confidence do you personally have in … ?” [The Dutch parliament] (0 = “No confidence at all”–10 = “A lot of confidence”).

5 Data were retrieved from CoronaWatchNL available at https://github.com/J535D165/CoronaWatchNL (accessed May 2022).

6 For a robustness check with alternative age operationalizations see Table A7 in the Appendix. The substantive findings remain the same.

7 The wording of the economic satisfaction question differs slightly between the waves. Pre-pandemic: “And how satisfied are you with the way in which the following institutions operate in the Netherlands?” [the economy] (0 = “very dissatisfied”–10 = “very satisfied”). Pandemic: “In general, how satisfied are you with the current economic situation in the Netherlands?” (0 = “very dissatisfied”–10 = “very satisfied”).

8 We opt for a two-way fixed effects specification for the model with the pre-pandemic trust variable since the between individual variance is close to zero here. We present a random effects specification as a robustness check in Table A8 in the Appendix. Here, we lag our trust variable by t – 1. The substantive findings remain the same.

9 See Tables A9, A10 and A11 in the Appendix.

10 See Table A12 in the Appendix for a robustness check where we include unemployed respondents in the analysis. The substantive findings remain the same.

11 The large confidence intervals for the age group <=30 might point to a lack of observations for this subgroup. However, we still have N=275 in this category.