Rally Effect in the Covid-19 Pandemic: The Role of Affectedness, Fear, and Partisanship

Abstract

When the Covid-19 pandemic hit internationally in March 2020, governments and political incumbents received exceptionally high approval ratings. Such a sudden spike of public support in times of crisis is often explained as the ‘rally ‘round the flag’ effect. This paper has three goals: first, to examine whether a rally effect indeed occurred; second, to analyse whether and how much it is related to (i) affectedness, i.e. the occurrence of infections on individual and aggregate level, and (ii) fear of Covid-19; and third, to examine an assumed moderating effect of partisanship. We merged individual survey data from an online survey conducted in September 2020 as part of the German Longitudinal Election Study (GLES) with infection rates on the state level (Bundesländer) published by the Robert Koch Institute. We detect a striking rally effect in all partisan camps. Furthermore, we identify fear of Covid-19 as the driving mechanism while there is no evidence that affectedness is a major force behind the rally effect. Furthermore, we show that partisanship takes on a moderating role for fear of Covid-19 regarding satisfaction with government.

Introduction

The world has been holding its breath since the infectious disease Covid-19 began spreading. After the initial outbreak of the virus in March 2020, many countries implemented a comprehensive shutdown of social life and the economy, while closing borders to neighbouring countries and imposing entry bans. At the same time, governing parties and heads of government received exceptionally high approval rates (The Economist 2020; Jennings 2020). Such an erratic increase in support in times of crisis is often explained by the ‘rally ‘round the flag’ effect (Mueller 1970, 1973; Hetherington and Nelson 2003; Lai and Reiter 2005). Although previous studies have primarily investigated the effect of war operations and terrorist attacks, the pandemic can also be identified as a ‘rally point’ due to its immediate severe and negative impact. A straightforward assumption is that the more the virus spreads, the more menacing the situation appears and thus, turning to the government and its leaders for protection and guidance seems to be a most likely case. But what exactly caused the steep increase in national governments’ approval ratings during the Covid-19 pandemic? Is affectedness decisive for the approval? And if so, is it personal or collective affectedness? Or is there a socio-psychological aspect to the phenomenon, and fear of Covid-19 is a major contributor? Is the rally effect feeding from all political camps? Unfolding the puzzle of rapid increase in approval rates for governments amid the pandemic could give insight into the underlying mechanisms behind and enhance our understanding of how dramatic international events affect political attitudes that are fundamental to a functioning democracy (Easton 1965, 1975). Furthermore, rally effects can have far-reaching consequences. Usually, election days are fixed long in advance. External threats, by contrast, come surprisingly and unscheduled and therefore might lead to significantly different election results compared to times without an essential crisis. Moreover, the vote shares might significantly differ after an acute menace, when the public is able to evaluate the governments past performance vis-a-vis the threat.

We shed light on the relationship between the spread of the virus and the surfaced rally effect during the pandemic. In this analysis, Germany serves as a case study, since the government’s approval rating experienced a sharp increase and additionally, the data available allows a more nuanced distinction of the occurrence of infection and asks specifically about the extent of fear of the pandemic at the individual level. From these considerations, we adopt a multilevel perspective by looking at personal affectedness by and fear of the Covid-19 virus on an individual level and the rates of infections within states on an aggregate level. For this, individual survey data from an online survey conducted in September 2020 as part of the German Longitudinal Election Study (GLES) are merged with infection rates on the state level published by the Robert Koch Institute (RKI).

This paper proceeds as follows: First, we present a discussion of the literature on the ‘rally ‘round the flag’ effect and carry out the argument as to why the Covid-19 pandemic identifies as a rally point and thus allows for a rally effect. Next, we describe what we believe to be the main indicators – personal and collective affectedness and fear – that can explain the strong and sudden increase in approval ratings and elaborate which role partisanship plays. In the analysis section, we first establish an empirical basis by showing that a rally effect did indeed occur in Germany during the Covid-19 pandemic. We then examine how affectedness and fear contribute to the explanation of the pandemic rally effect and show that partisanship functions as a moderator. Finally, we discuss the implications of our findings.

The Covid-19 Pandemic and Approval of the Government

Citizens constantly observe and evaluate their governments, especially in times of crisis. On 11 March 2020, the World Health Organization (WHO) declared the Covid-19 infectious disease outbreak a global public health emergency (WHO 2020). Shortly thereafter, most countries responded with a comprehensive lockdown and strict measurements that essentially aimed to contain the further spread of the virus. At the same time, the rapid growth in the number of infections in many countries was accompanied by greater support for the governing parties and heads of government (The Economist 2020; Jennings 2020).

In general, international crises can provide an uplifting booster or become a ruthless castigator for leaders’ or governments’ public support. In this paper, we refer to the existing literature of the rally effect, the people’s tendency to rally around the flag during times of crisis regardless of their political predispositions. The original concept refers to a sharp spike in presidential approval in the United States across all political camps in response to a sudden and intense national threat (Mueller 1970, 1973). Through continuous extensions, the concept also became applicable outside of the US and to national political institutions. A rally effect is triggered by an exogenous event that causes a conflict or a direct threat.

To constitute a rally point an event must have certain characteristics: it must (1) be international, (2) directly involve the nation,¹ and (3) be specific, dramatic, and sharply focused (Mueller 1973, 208). The triggering moment is thus subject to certain preconditions but is not limited to a specific type of event, as empirical evidence has shown throughout the years. A ‘rally ‘round the flag’ effect for national governments and leaders could be found, for example, in war and military operations (e.g. Callaghan and Virtanen 1993; Parker 1995; Lai and Reiter 2005; Berinsky and Druckman 2007), after terrorist attacks (e.g. Hetherington and Nelson 2003; Wollebæk et al. 2012; Coupe 2017) or attacks on persons of public interest (e.g. Ostrom and Simon 1989). In contrast, even positive events such as important diplomatic developments, e.g. German reunification in 1990, can also serve as a rally point (Norpoth 1994). More important, more recent studies have already suggested that an epidemic can also trigger a rally effect (Prati, Pietrantoni, and Zani 2011; Blair, Morse, and Tsai 2017).

In theory, the Covid-19 pandemic indeed has the characteristics necessary to identify it as a rally point. It is, by definition, international, as it is rampant worldwide and therefore affects a nation directly and as a whole. In the beginning, the infectious disease spread slowly, but then steadily and exponentially; it did not just suddenly affect all areas of society but also posed a real threat to life. However, the Covid-19 pandemic exhibits three differences compared to previously studied rally points. First, the global spread of the virus is not based on a conscious decision of action. Second, it is not aimed at a specific target, such as individual actors, a specific group of people or public institutions. The pandemic does not create a clearly identifiable image of two opposing parties but an abstract threat that cannot be pinpointed nor captured conclusively.² Third, the event lacks a clearly defined timeframe. In sum, compared to classic rally points induced by war or terrorism, the major conditions for a rally effect are met, while others are either not or less present. Thus, we assume that the Covid-19 outbreak served as a rally point that induced a sudden increase in public support for national governments.

The Role of Affectedness, Emotions and Partisanship

We assume that the relationship between the emergence of the pandemic and the increase in approval ratings is due to a threat situation arising. In this context, affectedness by and fear of the Covid-19 pandemic serve as indicators for a threat appraisal and risk perception, which, in turn, are used as a reference point for generating evaluative attitudes towards the national government. Both constructs make the pandemic tangible and palpable. Since partisanship is generally the most influential orientation heuristic through which citizens perceive and construe the political world, we also address the role of party identification.

Affectedness by the pandemic can indicate the extent of exposure to the threat. This construct enables it to be conceptualised on an individual and aggregate level. First, the Covid-19 pandemic can be seen as a collective affectedness, where the occurrence of infection on an aggregate level, such as geographical region, is used as a reference point for assessing the danger of the situation. We assume that the higher the number of confirmed cases of infection, the greater an individual appraises the threat to be. This would lead more people to seek protection and rally around central actors such as the national government. Several studies suggest a connection between number of cases on a national level and government approval. Yam et al. (2020) confirm a positive relationship between new daily confirmed and total confirmed cases and approval ratings of prime ministers and presidents in their overall sample. Similarly, Schraff (2020) shows a strong increase in trust in the national parliament related to the cumulative number of daily reported Covid-19 infections in the Netherlands. Based on one region in Germany, Bavaria, Leininger and Schaub (2020) showed that local incumbents of the governing party, the CSU, experienced up to 4 per cent of gains in vote share in local elections in March 2020. Conversely, Bol et al. (2020) find a weak but significant negative effect of number of deaths on trust in national government in Western European countries, thus refuting a rally effect. These conflicting results thus leave open the question of whether Covid-19 cases trigger a rally effect. But looking at affectedness on a national level through confirmed cases and fatalities might obscure underlying mechanisms. Therefore, it would make sense to lower the level of analysis to examine a possible rally effect thoroughly. For example, people may be more concerned with the occurrence of infection in their closer geographical surrounding such as the state they are living in than the whole nation. Individuals may therefore use corresponding case rates to assess the threat, which then represents a reference value for generating political attitudes. Derived from the above considerations, we expect a positive relationship between collective affectedness on a regional level and the approval of government.

H_1a: The higher the Covid-19 affectedness on the state level the higher the satisfaction with the government.

Second, people can be personally affected by the Covid-19 disease. By declaring the Covid-19 outbreak a pandemic, the WHO defined an objective threat to people. Yet research shows that risk or threat perception has a subjective component and depends on individual factors (Slovic 2010). We, therefore, assume that incidents on an individual level play a significant role. Thus, personal affectedness can occur when persons or their immediate social environment come into contact with the virus, whether as a confirmed or suspected infection or even death. Drawing on the victimisation hypothesis that focuses on fear of crime, we can formulate assumptions about the relationship between personal affectedness and risk perception: persons who became victims, are acquainted with someone victimised or just heard of victimised persons through gossip, for example, show higher levels of fear³ (Hale 1996). Dryhurst et al. (2020) ascertain that people who tested positive for Covid-19 or suspected they had the disease indeed expressed a higher perception of risk. The same may apply to people’s personal surrounding as Parlapani et al. (2020) find evidence of an association between fear of Covid-19 and a significant other’s Covid-19 illness. We, therefore, assume that being personally affected and incidents in the personal surrounding raise the awareness of the danger making the situation more tangible and closer to the self. However, even without coming into contact with the virus, the anticipated probability of becoming infected and the expected severity of the course of the disease can be used to assess the danger of the situation. A perceived high risk of infection and people expecting to become seriously ill may lead to greater anxiety. To sum up, increased concern due to personal affectedness by the pandemic can trigger a rally effect and leads us to the following assumptions:

H_1b: Personal affectedness by Covid-19 infection cases (regarding oneself or one’s social surrounding) increases satisfaction with the government.

H_1c: The higher the perceived risk of infection and becoming seriously ill the higher satisfaction with the government.

Emotions may gain great momentum in affecting attitudes and behaviour, especially in times of crisis and great uncertainty. A more recent strand of research based on various social-psychological models deals with the influence of emotions – predominantly anger and fear as affective reactions to a perceived threat from out-groups – in connection with policy preferences after or during a rally effect (for an overview see Lambert, Schott, and Scherer 2011). In terms of terrorism and war, fear increases support for precautionary antiterrorism measures but decreases support for risky military responses to terrorist attacks, while anger promotes support for military invention (Lerner et al. 2003; Huddy et al. 2005; Schoen 2006a, 2006b; Lambert, Schott, and Scherer 2011). We assume that fear plays a more significant role than anger in the Covid-19 pandemic, however. Previous findings show fear is a widespread reaction during the pandemic (Tabri, Hollingshead, and Wohl 2020; Knowles and Olatunji 2021). Moreover, from the perspective of cognitive psychology, anxiety is a response to an external threat over which the threatened person has little control (e.g. Eysenck 1992). Thus, fear arises from a position of insecurity that one’s own actions cannot (sufficiently) reduce. In a global health crisis, the scope of action is simultaneously extremely complex and limited. Therefore, turning to a central actor, such as the national government, is very likely in order to reduce feelings of insecurity and associated fear. Albertson and Gadarian (2015) find that anxiety arising from crisis-afflicted situations such as migration, terrorist events or public health that do not originate from government actions increases citizens’ trust in a variety of public actors, including national governments and political leaders. In such situations, anxiety and fear can even overcome the power of partisanship (e.g. Porat et al. 2019). We therefore expect:

H_1d: The greater the fear of Covid-19 the higher the satisfaction with the government.

Satisfaction with the government is generally strongly influenced by party affiliation (Evans and Chzhen 2016). In times of severe crisis, we expect a cross-party rally effect due to either increased identification with the threatened nation (e.g. Mueller 1970, 1973; Kam and Ramos 2008) or because the opposition is muted (e.g. Brody 1991; Gelpi, Feaver, and Reifler 2009). As recent studies have shown, political camps were unified and political conflict was attenuated at least in the beginning of the pandemic (e.g. Merkley et al. 2020). Thus, we assume that the rally effect feeds from all political camps. Simultaneously, party identification can lead to ideologically distinct government approval rates in highly polarised systems (e.g. Allcott et al. 2020; Cornelson and Miloucheva 2020; Goldstein and Wiedemann 2020). Therefore, it is plausible to expect that the sudden enthusiasm for the government is not equally distributed across all partisan camps. Partisanship is a powerful social identity in the political domain and the lens through which citizens perceive politics. Thus, exclusive partisan identity with its in- and out-party perceptual and attitudinal biases, the selective attention to specific political actors and the partisan view on the pandemic might result in a moderating effect of partisanship. Based on these considerations we formulate a corresponding hypothesis:

H₂: The influence of the explanatory factors of the rally effect varies by party identification.

Data, Variables, and Methodology

To empirically address the question of what exactly drives the increase in approval rates, we take individual-level survey data from the German Longitudinal Election Study (GLES) that was carried out in September 2020 (GLES 2020d, 2020e). The web-based survey captures central aspects of electoral and political attitudes and includes comprehensive information about the occurrence of infection on an individual level and various attitudes specific to Covid-19 infection, which were not previously available. The sample consists of German citizens eligible to vote in national elections at the time of the survey. After eliminating missing values and invalid cases, we yield a sample size of 1,385 respondents from 360 counties in 16 states.

Our dependent variable is satisfaction with government, where respondents were asked to report their contentment with the performance of the federal government in general on an 11-point scale ranging from ‘completely dissatisfied’ to ‘completely satisfied’. The main independent variables were by (i) (personal and collective) affectedness and (ii) fear of Covid-19. On the individual level, we used four different variables to measure personal affectedness. The first indicator asked whether the respondents are or have been infected with the Covid-19 virus (Infection, ego). Additionally, respondents could indicate whether suspected or confirmed cases occurred, whether there are recovered or deceased persons or whether they were not aware of any cases in their personal surrounding (Infection, surrounding). In addition to reporting the actual incidence of infection, two subjective indicators of risk perception were included: Respondents estimated their probability of becoming infected with Covid-19 (Infection, probability) and, in the case of an infection, the anticipated severity of the course of the disease (Assessed disease progression). As we also argue that a general fear of Covid-19 is important for government approval, we include a measure that captures the extent of anxiety. Respondents were asked to indicate how afraid they were of the Covid-19 crisis on a 7-point scale ranging from ‘not afraid at all’ to ‘very afraid’.

Since we argue that affectedness can additionally be modelled on an aggregate level, we use the occurrence of infection on the state level as an indicator of collective affectedness. To include region-specific rates of Covid-19 infection, we merged context information based on official Covid-19 case numbers provided by the central public health institute in Germany (RKI 2021) with our dataset. Since it would be instructive to examine different indicators that reflect the threat of Covid-19 pandemic, we captured the cumulative number of cases per 100,000 inhabitants as well as deaths per 100,000 inhabitants.⁴ To generate the corresponding variables, we aggregated the case numbers from the beginning of the pandemic to the day before the survey started (for descriptive statistics see Figure A1 in the online appendix). We used the postcode to assign the respondent’s place of residence to county in order to merge the Covid-19 context variables with the survey data. Cases were dropped from the dataset if the postcode resulted in an ambiguous assignment to a county or state.

Furthermore, we include dummy variables for party identification. Party affiliation is a strong predictor of satisfaction with the national government. According to theoretical approaches, it loses some of the explanatory power during a rally event but may continue to moderate the central mechanisms of the rally effect.

On the individual level, we control for several additional indicators. To separate the rally effect from a performance evaluation, we include a metric for satisfaction with the government’s handling of the Covid-19 pandemic, as studies show that implementing a lockdown has a positive effect on approval ratings (Bol et al. 2020; Sibley et al. 2020; Baekgaard et al. 2020). People faced with insecurity in various domains, e.g. finances, health, politics and social relations, are prone to high levels of fear (e.g. Caliso, Francisco, and Garcia 2020) suggesting that fears in different domains reinforce each other or have common roots (e.g. Lippold et al. 2020). Thus, we included other domain-specific anxieties that are available in the data set. By conducting an exploratory factor analysis, two indices are generated that depict fears regarding migration, economy, global warming, terrorism, EU, globalisation, crime and populist parties (see Table A1 for question wording as well as response scales and Table A2 for the result of the factor analysis in the online appendix). We also control for socio-economic status by including indicators for education and occupation as well as assessing the individual and national economic situation that have proven to be strong indicators for explaining support for governments (e.g. Foster and Frieden 2017). People that trust fellow human beings in general might also trust the government and therefore show higher values of satisfaction (Zmerli and Newton 2008). Lastly, we control for gender and age.

For the following multivariate analysis, dummy variables were created for ordinal and nominal variables and metric variables were centred at the grand mean. Summary statistics for all listed variables can be found in Table A3 of the online appendix.

Results

Descriptive Analysis

Officials in Wuhan, China reported the first human infection with the Covid-19 virus in December 2019. At that time, for the European and German public perceived the spread of the virus as an event happening elsewhere in the world that did not cause greater concern for their own affectedness. As Figure 1 shows, satisfaction with government remained fairly stable on a relatively low level until January 2020. This changed drastically between January and May 2020, i.e. after strongly increasing infection rates and the government’s decision to shut down most aspects of public life. Within this short period, satisfaction with government rose by roughly 1.5 scale points on average. Satisfaction with the government declined somewhat in September 2020 but was still much higher than before the outbreak of Covid-19. All governing parties experienced a similar increase in their approval rating. Ratings of the CDU, Chancellor Angela Merkel’s party, are slightly better than those for its Bavarian sibling, the CSU, and the Social Democrats (SPD). This pattern is not pandemic specific but applies for the entire time period we studied (Figure 1).

Figure 1. Weighted means of satisfaction with the government’s and opposition parties’ performance. Notes: Respondents were asked to report their general satisfaction with the federal government on an 11-point scale ranging from ‘–5 completely dissatisfied’ to ‘+5 completely satisfied’. Shown are means weighted in adjustment to the micro-census 2017. Sources: GLES (2020a, 2020b, 2020c, 2020d, 2020e, 2020f).

Figure 1 also demonstrates that the rally effect is indeed limited to the governing parties. Every oppositional party, i.e. the right-wing Alternative for Germany (AfD), the Green Party (B90/Grüne), the Left Party (Die Linke) and the Liberal Party (FDP) lost support. Although decreasing satisfaction with oppositional parties is less pronounced compared to the gains of the governing parties, these two figures clearly show that the rally effect is restricted to the government and the ruling coalition partners. Figure 2 shows the general satisfaction of partisanship. It is striking that the rise in satisfaction at the beginning of the pandemic is visible in every partisan camp, even in those at the ideological fringes. Thus, government approval increased during the Covid-19 pandemic across all political camps. What causes these striking gains in government satisfaction? Subsequently, we turn to multivariate analyses and focus on our two conceptual explanations: (i) (personal and collective) affectedness by and (ii) fear of Covid-19.

Figure 2. Weighted means of satisfaction with the government’s performance by partisanship. Notes: Respondents were asked to report their general satisfaction with the federal government on an 11-point scale ranging from ‘–5 completely dissatisfied’ to ‘+5 completely satisfied’. Shown are means weighted in adjustment to the micro-census 2017. Sources: GLES (2020a, 2020b, 2020c, 2020d, 2020e, 2020f).

Multivariate Analysis

Due to the hierarchical structure of the data, where individuals are nested in federal states, we apply a multi-level analysis (cf. Snijders and Bosker 2012). Table 1 shows the respective models. M1 models affectedness on different levels: ego, personal surrounding and state level. M2 adds an indicator for fear specific to Covid-19. M3 adds party identification and several control variables, including: economic perceptions, education, class, age gender and two indices comprising domain-specific fear. Finally, Model 4 adds an indicator for the populations’ satisfaction with the handling of the Covid-19 pandemic in order to see whether we observe a rally effect based on affectedness and fear or just satisfaction with the government’s crisis-time policies.

Table 1. Multilevel models for satisfaction with government’s performance (state).

	M1	M2	M3	M4
Infection, ego           Yes	1.92**	1.72**	0.53^+	0.46
(Ref.: No)          Yes, probably	−0.23	−0.22	−0.25	0.25
Infection, surrounding        Yes	−0.35	−0.31	−0.47**	−0.35***
(Ref.: No)         Yes, probably	0.14	−0.13	0.18	−0.03
Recovered	0.46	0.50	0.08	−0.02
Decedents	0.34	0.23	0.19	0.39^+
Infection, probability	0.54***	0.43***	0.13	−0.03
Assessed disease progression	0.04	−0.10	0.07	−0.14^+
Fear of Covid-19		0.23***	0.26***	0.14*
Party identification (PI)     CDU/CSU			2.03***	1.22***
(Ref.: No PI)            SPD			1.41***	0.95***
FDP			0.36	0.13
Grüne/B90			0.40^+	−0.01
Die Linke			−0.71*	−0.71**
AfD			−1.21***	−0.96***
Others			−0.15	−0.50
Satisfaction with Government (COVID-19)				0.52***
Controls^c			x	x
State           Cases per 100k	0.00^a	0.00^a	0.00^b	0.00^b
Deaths per 100k	−0.05	−0.05	0.00^a	−0.01
Constant	4.47***	5.11***	4.95***	6.37***
Level 2 SD (intercept)	0.05	0.04	0.00^a	0.00^a
Level 1 SD (residual)	9.22	9.11	4.60	3.33
ICC	0.01	0.01	0.00^a	0.00^a
Maddala R²	0.01	0.02	0.51	0.64
Level 2 N	16	16	16	16
Level 1 N	1427	1427	1427	1427

Note: Beta coefficients weighted in adjustment to the micro-census 2017.

***p < 0.001.

**p < 0.01.

*p < 0.05.

⁺ p < 0.1.

^a Absolute value < 0.01 with a positive sign.

^b Absolute value < 0.01 and negative sign.

^c Controls: gender, age, education, class, two fear indices that comprise fear of migration, terrorism, globalisation, economy, crime, EU, populist parties and global warming, assessment of the individual and national economic situation and trust. Beforehand, empty models were calculated to determine the proportions of total variation at the state and individual level (see Table A5 in the online appendix).

Sources: GLES (2020d, 2020e) and Robert Koch-Institut (RKI 2021).

Starting with Model M1, objective data concerning both infection levels and mortality rates on the state level are fully insignificant in explaining satisfaction with the government.⁵ This result contrasts with our hypothesis formulated in H_1a which predicted a positive effect between aggregate-level infection rates and government satisfaction. This is an unexpected finding insofar as infection and death rates varied significantly across states (see Figure A1 in the online appendix) and were prominent in media reporting. Thus, we assumed that higher rates signify that the virus is closer to one’s home. Ego affectedness, by contrast, has a very strong impact on satisfaction with government. Individuals who either were infected or believe that their infection probability is high, tend to be far more satisfied with the government than unaffected individuals, which is in line with our assumption formulated in H_1b and H_1c. Interestingly, Covid-19 experiences within an individual’s personal network seem to be less important which in turn contradicts the prediction of H_1b. Hence, model M1 suggests that the rally effect in Germany is predominantly driven by ego affectedness (real and perceived). Note, that the explained variance of the pure affectedness model is very limited.

In Model M2 we include an indicator for fear of Covid-19, which only minimally increases the overall explained variance by one percentage point. Nevertheless, as assumed in H_1d, fear of COVID-19 shows a strong and positive effect. Introducing fear of Covid-19 does not substantially change the effects of being affected personally and the perceived infection probability. They remain strong predictors of government support.

Turning to Model M3, which includes party identification and diverse control variables (fears, economic perceptions and social structure), fear of Covid-19 remains a highly significant determinant of government satisfaction. However, we also observe some changes: first, the significance level of individual affectedness decreases. Second, the positive effect of infection probability is no longer significant. Third and unexpected, we see a significant negative effect of Covid-19 infections within an individual’s personal network. Looking at partisanship, the results indicate that partisan predispositions do not entirely disappear in a major crisis. Adherents of the governing parties (CDU/CSU and SPD) are particularly satisfied with the government. Also, adherents of the Green Party are somewhat more satisfied than the reference group, i.e. non-partisans. By contrast, those who identify with the Left Party and, especially the AfD, are significantly less satisfied than non-partisans.

Model M4 introduces one additional variable that captures whether individuals are satisfied with Covid-19 policies of the government. This variable is highly significant and signals that at least part of the assumed rally effect is rather a result of citizens’ evaluation of government performance during the crisis. However, performance evaluation is only part of the story as Covid-19 fear remains a significant predictor of government support. Accordingly, we can conclusively verify hypothesis H_1d. In contrast, H_1b and H_1c are ultimately rejected because having been infected, estimated infection probability, anticipated disease severity and infections in the personal network do not exert the expected positive effects in the full model M4. To sum up, the results strongly suggest, that fear of Covid-19 is the mechanism that carries the rally effect.

Now that we found fear to be the mechanism of the rally ‘round the flag effect, we turn to the question whether the effect of fear on government approval is moderated by partisanship. Figure 3 shows the average marginal effects for the interaction of partisanship and fear of Covid-19 (see Table A4 M8 in the online appendix). We see that fear of Covid-19 functions differently for single categories of partisanship: on average, satisfaction with the government increases with growing fear of Covid-19 for adherents of the governing party SPD. This is also the case for the Green Party and non-partisans. Interestingly, for respondents who identify with the largest governing party, the CDU/CSU, the level of fear of Covid-19 does not change government satisfaction. On average the level of fear does not affect the evaluation of the government also for respondents that identify with the Liberals, the Left, AfD or other parties. In these cases, the trajectories of the effect of fear are either not steep enough, flat or even decrease slightly, but don’t show any significance (see Figure A2 in the online appendix). Hence, identifiers with the SPD and the Green Party as well as non-partisans show the dominant mechanism of the rally effect: Confronted with an external threat that causes fear, they turn to the government and support it. These results thus confirm H₂, which posits that party identification moderates the explanatory strength of the predictor of a rally effect. However, no clear pattern emerges. Being supporter of a government or an opposition party does not play an important role. Rather, we find a distinction between party supports left and right to the ideological centre, with those on the left more satisfied with government when they fear Covid-19 and those on the right with government approval rates independent on the fear level.

Figure 3. Average marginal efffect of fear (Covid-19) by party affiliation. Sources: GLES (2020d, 2020e) and Robert Koch-Institut (RKI 2021).

However, what our result clearly indicates is, first, that there has been a rally effect in all political camps, second, that, at the time of our study fear is the dominant mechanism of the Covid-19 rally effect and, third, that this dominant mechanism is not omnipresent, but accounts for the rally effect of only some partisan groups.

Conclusion

In this paper, we examine the mechanisms of a ‘rally around the flag’ effect triggered by the Covid-19 pandemic that suddenly hits the population. Looking at Germany, we show that satisfaction with the government starkly increased between January and May 2020, when infection rates sky-rocketed and the government first mandated a lockdown. This boost in government approval is present in all partisan camps and continues in September. Based on this, we explore, whether this increase in satisfaction with government is related to personal or collective affectedness – i.e. the ego, the ego’s surrounding or official statistics of infection and mortality rates on the state level – or to a socio-psychological aspect namely fear of Covid-19. Surprisingly, we find that affectedness on an aggregate level does not matter at all. Government approval is fully independent of factual degrees of affectedness as reported by the RKI. Similarly, we do not find personal affectedness to increase satisfaction with government, neither in terms of a personal infection or in the personal network, nor with regards to the perceived infection probability or anticipated disease severity. Our findings indicate that fear of Covid-19 is the dominant predictor related to the pandemic, even when we control for satisfaction with the government’s performance in coping with Covid-19.

Moreover, the paper inquires whether the effect of fear on government approval is universal or restricted by party affiliations. Among adherents of the SPD, Green Party, and among those who don’t feel close to a party, satisfaction with government parallels Covid-19 fears. The fear effect is substantial: It increases satisfaction with government by 1.0–1.9 points on an 11-point satisfaction scale (see Figure A2 in the online appendix). These estimates are very conservative, because the model includes all control variables, including the public’s satisfaction with the Covid-19 policies. Thus, we can be relatively certain that for these partisan groups the increase in government satisfaction is a rally effect related to fear. On the other side, adherents of the governing CDU/CSU, the Liberal Party, the Left and the AfD do not react accordingly. Increasing individual Covid-19 fear within these four partisan groups does not equate to increasing government support. For these groups, there must be other mechanisms that cause their rally effect that we didn’t address in our analyses. These mechanisms as well as the explanation for the distinction between these partisan groups in the effect of fear on government approval remain beyond the scope of the paper.

To further strengthen the robustness of our findings, similar analyses in other countries that are or have been (more severely) affected by the Covid-19 pandemic should be carried out. First, this could shed light on the extent to which our results can be generalised. Second, it would allow to address some limitations of our analysis. The data we used was collected in September 2020, a time when Covid-19 cases were already declining with a widespread expectation that the pandemic was ending. Thus, the variance of the number of infections and the salience of the pandemic were thus lower than at the beginning of the outbreak. A different survey period and greater variance in number of Covid-19 cases could give information about whether objective data concerning infection and mortality rates really do not matter and fear remains the key predictor of the rally effect during the Covid-19 pandemic.

Nevertheless, our main results contribute to a research approach on the rally effect that focuses on psychological models highlighting the importance of emotions as the main driver (e.g. Lambert, Schott, and Scherer 2011). We provide evidence that emotions are crucial for the formation of political attitudes such as satisfaction with the national government in times of crisis. Thus, future research should focus more on the role of social-psychological components. As shown by Leininger and Schaub (2020), rally effects can have democratic repercussions. If elections are held close to the rally point, this can have massive implications for the voting behaviour of the electorate. Even governments and governing parties that the public rated poorly before the rally event could suddenly gain votes from a wide spectrum of voters who would have otherwise voted differently. Moreover, such external ‘attacks’ alter the public agenda. The rally issue, the Covid-19 pandemic, trumps all other issues. Even weak governments or those that failed on many issues previously dear to the public might be re-elected if elections are close to a rally point. Finally, Kam and Ramos (2008) state that there are two phases of the rally effect: ‘rally formation’ and ‘rally depression’. Thus, the heightened government approval that comes with a crisis does not last forever. The leap of faith in the government will eventually end and political support will return to well-known patterns of political predispositions such as partisanship. As the pandemic continues, citizens might get used to the ‘new normalcy’ and fear of Covid-19 might diminishes. The pandemic might remain the prevalent political issue, but the public might focus rather on the retrospective evaluation of government’s handling of the pandemic.

Supplemental Data and Research Materials

The underlying research materials for this paper, e.g. data, samples or models, can be accessed at https://osf.io/dctr7/?view_only=4ceb8fd40c6e4b1f94e85028233962ce.

Disclosure Statement

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

Additional information

Notes on contributors

Melanie Dietz

Melanie Dietz is a research associate at the Research Institute Social Cohesion, Goethe University Frankfurt and a research assistant in the German Longitudinal Election Study (GLES) since 2017. In her bachelor's degree, she studied sociology and political science at the University of Mannheim, while she completed her master's degree with a focus on political sociology and social inequality at Goethe University Frankfurt. Her research interests lie in electoral behaviour, political participation, and political attitudes with a focus on gender studies.

Sigrid Roßteutscher

Sigrid Roßteutscher is a professor in the department of social sciences, Goethe University Frankfurt. She studied history and political science at the University of Mannheim and received her PhD from the European University Institute (EUI) in Florence, Italy. She acquired her post-doctoral qualification (Habilitation) at the University of Mannheim and was Co-Director of the German Longitudinal Election Study (GLES) from 2009 to 2017. Currently, she directs an international Norface programme on The Rural Urban Divide in Europe (RUDE). Her research interests focus on political participation, electoral behaviour, turnout, political inequality, the role of religion in politics, value orientations and social capital.

Philipp Scherer

Philipp Scherer is a research associate in the department of social sciences, Goethe University Frankfurt. He studied social sciences with advanced fields of study in economics and political science at the University of Koblenz-Landau from 2002 to 2008 and received his PhD from Goethe University Frankfurt in 2017. His research interests focus on electoral behaviour, political attitudes, the socio-political relevance of the left-right dimension and content analysis.

Lars-Christopher Stövsand

Lars-Christopher Stövsand is a research associate in the department of social sciences at the Goethe University Frankfurt and a research fellow in the German Longitudinal Election Study (GLES). He studied political science and history at the Christian-Albrecht University of Kiel, the University of Salamanca and Goethe University Frankfurt. His research interests focus on electoral behaviour, political attitudes, and political polarisation.

Notes

1 Mueller (1970, 21; 1973, 208) originally writes ‘involves the United States and particularly the president directly’.

2 Anger as an emotional reaction to a threatening event is associated with agency, which grows when there is an actor clearly responsible for the threat (e.g. Carver 2004). It is debatable to what extent the COVID-19 virus can be considered an abstract actor. In the beginning of the outbreak, China was blamed as the initiator of the pandemic, which was accompanied by resentment towards Chinese people (Rafi 2020).

3 The victimisation hypothesis assumes that becoming a victim of crime can happen again. Therefore, the analogy between fear of crime and Covid-19 is limited, because confirmed reinfections of the same person are presumably extremely rare. Nevertheless, persons who survive COVID-19 could still experience higher levels of fear because they worry about their friends and family.

4 In addition, we ran models for robustness checks, where we included 7-day incidences for the number of cases and deaths as long-term indicators as well as the absolute number of cases and deaths to record short-term developments during the week before the survey was conducted. Respective models yield the same results.

5 In additional analyses we included the same indicators of the occurrence of infection on a county level and ran multilevel models with three levels (individual, county, state). But again, the results for the county indicators did not display any significance.