How negative out-party affect influenced public attitudes about the coronavirus crisis in Norway

Introduction

There is no denying that political developments during the last decades have sparked interest in political polarization throughout the Western world. However, rather than increased ideological differences between political parties, the form of polarization that attracts the most debate relates more to partisan antagonism toward out-parties (e.g., Gidron et al., 2019, p. 30; Ward and Tavits, 2019, p. 1). This form of negative affect toward out-parties is commonly referred to as negative partisanship (NPID) or affective polarization. However, despite the interest in the subject, there are still relatively few studies that examine political consequences through experiments or with panel data (Iyengar et al., 2019; Druckman et al., 2021a). Relatedly, given that excessive levels of negative out-party affect can have harmful consequences for a country, it remains an important objective to explore how its effects can be mitigated (Iyengar et al., 2019; Ridge, 2020; Druckman et al., 2021a).

From the very start of the coronavirus pandemic, negative affect toward out-parties was highlighted as something that could lead partisans to form different opinions about the crisis and thus undermine an effective collective response. The concern has been that partisans select news sources that focus on criticizing disliked out-parties and interpret information in ways that would be disadvantageous to them (Grossman et al., 2020; Van Bavel et al., 2020, p. 464; Druckman et al., 2021a,b). Since then, several studies have illustrated differences along partisan lines in how people thought and acted in response to the crisis (Grossman et al., 2020; Gadarian et al., 2021; Kerr et al., 2021; Bolsen and Palm, 2022). In addition to these concerns, the pandemic was also seen as a rare opportunity for political elites to present a unified front against the pandemic that could lead people to rally around the flag and thus potentially reduce polarization toward disliked out-groups (Van Bavel et al., 2020).

While there has been much research that connects the pandemic to partisanship and political polarization, they tend to focus on Anglo-Saxon countries or do not directly examine how out-party dislike influences attitudes about the crisis response. Consequently, the first aim of this paper is to explore how negative out-party affect shaped attitudes about the political response to the crisis in Norway and legislation meant to mitigate its consequences. The second aim is to examine whether these hypothesized effects of out-party dislike are reduced when Norwegians think about the country's response instead of partisan actors. As such, the article is based on similar studies in the US, specifically the article by Druckman et al. (2021a), which examines how affective polarization shaped attitudes about the political response in the US and how this effect was altered when people evaluated the superordinate category of the United States instead of President Trump. However, Norway presents a contrasting case to the US: while the response in the US was polarizing and closely associated with the president himself (Gadarian et al., 2021; Kerr et al., 2021), Norwegian politicians managed to successfully portray an image of political consensus that was echoed in the media narrative (Christensen and Lægreid, 2020; Fonn and Hyde-Clarke, 2021; Moss and Sandbakken, 2021). Hence, studying the effects of negative out-party affect in Norway provides a test for how well results travel across countries and an opportunity to explore differences between countries that could depend on contextual factors.

To answer the research questions, the paper utilizes two survey experiments. The first experiment mimics Druckman's et al. (2021a) study. However, instead of focusing on the US case, the article uses representative panel data from the Norwegian Citizen Panel (NCP) to examine the effect of negative out-party affect prior to the coronavirus outbreak (in January 2020) on Norwegian voters' perception about how Norway and the conservative government handled the crisis (in November 2020). Specifically, because of the unified political response in Norway, I expect that the effect of negative out-party affect will be significantly reduced when people evaluate the response of Norway as opposed to the conservative government. This would then contrast with the results of Druckman et al. (2021a), who found that partisans actually evaluated the Trump and US response more similarly as they became more affectively polarized.

The second survey experiment manipulates the party cues ascribed to a policy proposal about whether the Norwegian government should provide generous financial assistance to companies with losses in turnover during the coronavirus crisis. Such economic rescue packages are vital for an effective crisis response but could be hampered if negative out-party affect undercuts elite cooperation and public support. Moreover, while numerous studies have shown that positive partisanship moderates the effect of party cues, there is less research on how out-party affect impacts people's reaction to cues from out-parties. Accordingly, the second survey experiment explores whether out-party cues polarized opinions on the issue of providing financial assistance to companies and whether this effect increased when voters disliked the parties more.

First, the results show that out-party dislike significantly influenced attitudes about the conservative government's management of the crisis, meaning that it was associated with positive attitudes for supporters of the three governing parties and negative attitudes for opposition party supporters. However, in contrast to similar research on the US case (Druckman et al., 2021a), this effect either disappeared (government supporters) or was weak and not robust (opposition supporters) for those who evaluated Norway's response instead. Second, assignment to cues from the two out-parties—Labour and Progress Party—decreased support for the proposal about providing financial assistance to companies with losses in turnover due to the pandemic. This effect, in turn, tended to be stronger for those who disliked the Progress Party more while there was no consistent moderating effect for dislike toward the Labour Party.

As such, the results contribute to the research field on negative out-party affect by focusing on its political consequences outside of the US context. While neither of the two experiments prove causality by manipulating out-party affect (Druckman et al., 2021a), they utilize observational panel data and manipulation of party cues to show how out-party affect impacted attitudes. Perhaps more importantly, the results indicate how this effect could depend on different circumstances, such as whether evaluations are framed in more unifying or partisan terms, the parties for which cues are assigned to or contextual factors at the country level. Furthermore, the Norwegian case, not least in relation to the coronavirus response, serves as somewhat of a least likely case to find such consequences, and the study therefore complements research that has focused on countries with more polarizing elite cues. Last, although not surprising, the article makes a contribution to research on crisis management by exploring how support for important proposals, and the crisis response more broadly can be strengthened by framing the proposal/response in more unifying terms, i.e., as the country's response or as having been supported by parliament. Specifically, the results indicate that this could even reduce partisan differences among those who dislike out-parties, although this may depend on contextual factors in the country.

The article proceeds in the following way. First, I present the hypotheses and theoretical motivation underlying them. After providing a brief overview of the Norwegian case and the coronavirus response, the next section then details variable operationalization and the methodology used. Then, the main results from the survey experiments are presented, and the paper ends with a brief discussion of the results and the most important implications deriving from them.

Theoretical framework

The theory section summarizes previous research on how negative out-party affect should influence attitude formation, independent of positive in-party affect. I also discuss why the effect, as it relates to attitudes toward the coronavirus response, might be different in Norway than in the US.

Negative partisanship and attitude formation

Numerous studies based on observational panel data (e.g., Lenz, 2012; Achen and Bartels, 2016) and survey experiments (e.g., Slothuus and de Vreese, 2010; Brader et al., 2013) have suggested a causal connection between partisanship and subsequent attitude formation. Scholars have pointed to several social-psychological mechanisms that drive this behavior. One such mechanism is people's tendency to partake in motivated reasoning, whereby they process information to fit with their partisan identification (Rudolph, 2006; Taber and Lodge, 2006; Druckman et al., 2013). Alternatively, people who are not motivated to actively seek out and process political information often rely on party cues as an informational shortcut (Leeper and Slothuus, 2014).

Although Norway handled the coronavirus crisis well according to objective indicators (Gordon et al., 2021), the question of how the country responded to the crisis is still sufficiently broad for those who strongly dislike the government/opposition parties to find certain aspects of the government's actions—for instance, the closing of schools or not mandating face masks—that they can use to distinguish their assessment from others. Moreover, the rather uniform response to the crisis among Norwegian parties (Christensen and Lægreid, 2020; Moss and Sandbakken, 2021) does not necessarily rule out this type of motivated reasoning, since negative partisanship can lead people to politicize even apolitical issues (Druckman et al., 2021a, p. 223; Iyengar et al., 2019). Relatedly, another potential source of partisan bias is people's tendency to seek out information that is consistent with prior beliefs and group identities, i.e., selective exposure. The rise of cable news and social media has made it particularly easy for already polarized people to find an abundance of news sources expressing negative stories about political out-groups and positive depictions of the in-party(s) (Lelkes et al., 2017).

In contrast to positive partisanship, there is limited research on the political consequences of negative out-party affect and the adjacent concept of affective polarization (Iyengar et al., 2019; Druckman et al., 2021a). That being said, research on NPID has found several instances in which NPID correlates with political behavior—for instance, electoral turnout, satisfaction with democracy, political participation, evaluations of political leaders, and vote choice—independent of positive partisanship (Medeiros and Noël, 2014; Soroka, 2014; Mayer, 2017; Ridge, 2020). Similarly, other studies have found that affective polarization impacts political trust and partisan attitudes toward democratic norms in the US (Hetherington and Rudolph, 2015; Kingzette et al., 2021). However, the existing research on the political consequences of NPID, and in particular its effect on attitude formation, tends to be either cross-sectional and/or focused on the rather exceptional US case. Consequently, we have limited knowledge about how results travel to other countries where the political culture, institutions and leadership differ markedly from those of the US.

From the perspective of social identity theory, Zhong et al. (2008) outline how people can develop negational group identifications and thus define themselves as primarily against some groups; that is, they define themselves as who they are not, rather than who they are (Zhong et al., 2008, p. 793). This then leads to favoritism toward fellow “non-outgroup” members and discrimination toward members of the out-group, even when the out-groups are designed to lack substantive meaning (Zhong et al., 2008). In line with scholarship on social identity, negative out-party affect should also motivate people to differentiate themselves from out-groups by adopting attitudes that they think are disadvantageous or contrary to that of the party(s). In support of this, at least a few studies have examined the effects of out-party cues on attitude formation and found that these cues led to polarizing shifts away from the party's position (e.g., Druckman et al., 2013; Bankert, 2020; Skytte, 2020; Bäck et al., 2021). Yet, these analyses often do not explore how negative out-party affect toward the parties moderates this tendency.

Based on these theoretical considerations, the first experiment mimics the study of Druckman et al. (2021a). However, while they focus on the US case, I do so for Norway, whose coronavirus response differs from that of the United States. First, as noted earlier, the Norwegian political establishment responded to the crisis in a more unified manner. Instead of attacking the other side's actions and rhetoric, parties across the aisle generally worked together and supported important legislation while also toning down differences (Christensen and Lægreid, 2020; Moss and Sandbakken, 2021). Second, studies have shown that out-party dislike is low in Norway (Reiljan, 2020; Wagner, 2021), which suggests that there was less intense partisan dislike/distrust to start with when the crisis hit the country, although it is of course possible that this does not influence the relative effect of disliking out-parties. Last, the fact that Norway had nine parties in parliament, a parliamentary system and coalition government could make it more difficult for negative out-party affect to translate to partisan interpretations of how the crisis was managed, especially in comparison with the US system, where it is easier to associate the political response to one governing party and/or leader.

Accordingly, the first hypothesis tests whether supporters of the opposition parties who strongly disliked government parties prior to the COVID-19 crisis also expressed greater skepticism about how the country and the government had handled the crisis. Conversely, I expect the opposite to be true for supporters of government parties, whereby dislike of opposition parties is associated with more positive opinions.

H1: Supporters of opposition parties who express greater dislike of government parties will be more critical of the coronavirus response. This effect will be the opposite for supporters of government parties.

Following this, the second hypothesis focuses on how partisans react when exposed to explicit party cues on an otherwise rather uncontroversial (albeit important) policy proposal meant to mitigate the economic consequences of the pandemic. According to the theories of motivated reasoning and information heuristics described above, people should be more positive toward a policy proposal when it is ascribed to a party they support, while they become skeptical when they receive information that a competing party supports it (Rudolph, 2006; Taber and Lodge, 2006; Druckman et al., 2013). Moreover, this effect should increase among those who dislike the out-parties more (Bankert, 2020). This leads to the following hypothesis:

H2: People will evaluate the policy proposal less favorably when they receive information that a political opponent supports it. Moreover, this effect increases among those who dislike the out-party more.

Elite cooperation and the priming of national identity

In the aforementioned study by Druckman et al. (2021a), the authors also explore whether affective polarization had a different effect on how respondents evaluated the performance of the United States and President Trump. Their results show that polarization actually had a larger effect on how partisans evaluated the country's response, so the most polarized voters did not distinguish between the United States and President Trump on the issue. This led the authors to conclude that “superordinate appeals to the nation are ineffective for those who are most polarized, and hence policymakers need to craft strategies to appeal directly to them and work on depolarizing strategies rather than appeals to a shared identity” (Druckman et al., 2021a, p. 232).

However, we might expect different effects for the Norwegian case, where the actions of political elites signaled a much more unified and consensual response to the pandemic (Christensen and Lægreid, 2020; Moss and Sandbakken, 2021), thus increasing the likelihood of a rally round-the-flag effect around the issue. There is evidence for a rallying effect in Norway, as levels of political trust and support for the government increased after the start of the crisis in both Norway and neighboring Sweden (Christensen and Lægreid, 2020; Ares et al., 2021; Esaiasson et al., 2021; Sætrevik, 2021).

More generally, researchers have tended to highlight the priming of shared identities as something that could mitigate affective polarization. This idea builds on research in social psychology that has described it as a means to reduce the salience of conflictual social identities while also highlighting commonalities and shared identities (e.g., Gaertner et al., 1996; Levendusky, 2018). Levendusky (2018), for instance, finds that AP was lower among those who had their national identity primed, both through experimental survey manipulation and during events such as the Summer Olympics and Fourth of July celebrations in the United States. Similarly, in relation to the coronavirus crisis, Van Bavel et al. (2020) describe how political elites could “highlight an overarching identity” and thus reduce polarization over the issue. Therefore, I expect that the more consensual/cooperative Norwegian response, coupled with a lower baseline level of out-party dislike, will mean that negative out-party affect has a weaker effect when partisans evaluate the country's response. This leads to the following final hypothesis:

H3: Dislike of out-parties will have a smaller effect on attitudes toward the coronavirus response when respondents evaluate the country's response as opposed to that of the conservative government.

The Norwegian case

Comparative indices of affective polarization have indicated that negative out-party affect is modest in Norway (Boxell et al., 2020; Gidron et al., 2020; Reiljan, 2020; Wagner, 2021). As an example, Gidron et al. (2019) calculate party thermometer ratings between the largest left- and right-wing parties in 20 democracies and find that Norway is the least polarized country. Other studies that include all parties in their indices of affective party polarization also point in the same direction (Gidron et al., 2020; Reiljan, 2020; Wagner, 2021).

That being said, the Norwegian party system has become more fragmented over the last couple of elections, with the Green and Red parties gaining parliamentary seats in 2013 and 2017, respectively (Aardal and Bergh, 2018). Likewise, the populist-right Progress Party elicits negative feelings from supporters of left and center parties. Since a left-of-center coalition, consisting of the Labour Party (DNA), Center Party (Sp), and Socialist Left (SV), formed a government coalition in 2005, party competition has followed a bloc dynamic whereby these three parties have competed against a right-of-center alternative that includes the Conservatives (H), Christian Democrats (KD), Liberals (V), and Progress Party (FrP). However, the blocs have not been fully consolidated, both in terms of elite cooperation and voters' affective evaluations toward the parties. In particular, there is a rift between the populist-right Progress Party and the more centrist parties in the right-of-center bloc. Likewise, the left-of-center opposition became more fragmented when the Red and Green parties gained parliamentary representation (Sitter, 2006; Allern and Karlsen, 2014; Aardal and Bergh, 2018; Knudsen, 2021).

Clearly, the media's reporting about a crisis such as the coronavirus pandemic plays an important role in influencing public attitudes. In particular, the prominent role of public broadcasting in Norway probably made the media narrative less partisan, and Fonn and Hyde-Clarke (2021, p. 1) conclude that there was a “conspicuous lack of journalism that questioned the policies” of Norwegian authorities during the second wave of the coronavirus pandemic (Hujanen et al., 2013, p. 17–50). Moreover, Scandinavian countries are often referred to as consensual democracies in terms of their parliamentarianism, cooperation between minority governments and the legislative opposition, and neo-corporatist interest group systems (Arter, 2008; Lijphart, 2012, p. 165). Together with other well-known features, such as an extensive welfare state and high trust in political institutions, this constitutes relevant contextual factors where Norway distinguishes itself from, for example, the United States.

The actual political response to the crisis in Norway also differs vividly from that of the US and other countries where it has been more polarizing (Gadarian et al., 2021; Kerr et al., 2021). As previously described, research on the coronavirus response in Norway has emphasized the success with which political authorities presented a unifying message. Likewise, parties from across the aisle collaborated closely in drafting legislation to mitigate the crisis, which meant that there was broad political support behind the actions taken (Christensen and Lægreid, 2020; Moss and Sandbakken, 2021). Norway had also been comparatively successful at handling the crisis when the survey was fielded in terms of infection rates and deaths (Gordon et al., 2021). In short, in regard to examining how partisanship and out-party animus shape attitudes toward the pandemic response, Norway presents a contrasting case to the US case that has received the bulk of scholarly attention. In fact, it could even be considered a least likely case to find an effect of out-party dislike when considering the unified elite response, multiparty system and seemingly modest levels of partisan antagonism.

Materials and methods

This section provides an overview of the experimental design and data sources. In addition, operationalizations for the central independent variable of negative out-party affect are detailed alongside other control variables included in the analyses.

Experimental design

I rely on two survey experiments that were part of wave 19 of the Norwegian Citizen Panel (NCP)—fielded between the 2nd and the 27th of November 2020—to test the hypotheses (Ivarsflaten et al., 2021). The NCP is a nationwide randomly sampled internet survey of Norwegian residents, with a response rate for panelists recruited to wave 19 of 76.8 percent. The two survey experiments were embedded in a larger survey, and 1 990 respondents answered at least one of the two questions (Study 1 = 1 987, Study 2 = 1 984, answered both = 1 981). However, since most of the necessary independent variables were not included during wave 19, they had to be added from the latest available survey wave, resulting in a smaller sample for the actual analyses. Moreover, for the analyses pertaining to Study 1, the central independent variable—like-dislike party evaluations—comes from the 17th wave of the NCP, fielded between January 15 and 27. In line with Druckman et al. (2021a), this is to ensure that out-party dislike is measured prior to the outbreak of the virus in Norway. A methodology report concerning sample characteristics and the data collection procedures for the NCP can be found on the DIGSSCORE webpage, which details that there is an overrepresentation of older and better educated respondents (Skjervheim et al., 2020). To account for this, all statistical analyses include controls for age, education, gender and income.

Starting with the first experiment, the question asks the respondent if they “agree or disagree that [the conservative government/Norway] has done a good job at managing the coronavirus pandemic.” When doing so, 953 respondents were randomly assigned to the conservative government question wording, while the remaining 1,037 were asked about how good a job Norway has done.¹ The response options range from one (strongly agree) to seven (strongly disagree), which I first recode so that higher values denote positive attitudes toward how the crisis has been managed. As Figure 1 illustrates, the distribution of responses is very negatively skewed, meaning that people had a positive attitude about the political response.

Figure 1. Histogram for the question about whether respondents agree/disagree that Norway/the conservative government has done a good job. Higher values denote more positive attitudes.

Because of this skewness toward positive scores, I collapsed values 1–5 for the analysis of government party supporters and values one and two for opposition supporters so that the scales range from 1–3 and 1–6, respectively. Moreover, for this reason, ordered logistic regression is performed instead of linear regression.

The question of Norway's performance does not exclusively target political elites in the way the other question does, and people might base their considerations on, for example, people in their vicinity, health care providers or the country more broadly. However, this seems less of a concern when the experiment is also thought of as a test of how out-party dislike polarizes attitudes about the performance of a non-partisan target (the country) compared to an explicitly partisan one (Druckman et al., 2021a, p. 225) on a highly salient issue where political elites have presented a unified front. That is, the extent to which it polarizes attitudes (or not) can be interesting, even if a non-partisan interpretation leaves room for different consideration. It should also be noted that the measure of crisis response used by Druckman et al. (2021a) combines three survey items that capture confidence/evaluations of past performances/preparedness and preparedness for future outbreaks. Hence, the dependent variable is different from theirs, even though they also examine attitudes about the political response to the pandemic.

The second question asks about a component of the Norwegian economic rescue package, whereby companies with a loss in turnover of 30 percent were entitled to government support for up to 80–90 percent of their fixed costs. Parliament enacted the proposal in April with support from the main government and opposition parties, and it was initially estimated to cost more than 20 billion (ca. €1.95 billion) in Norwegian currency per month (Sørenes and Ask, 2020). Moreover, the fact that the proposal had bipartisan support meant that the party cues could be experimentally manipulated without misleading the respondents. The respondents received answering options ranging from one (strongly agree) to seven (strongly disagree), which I once again recoded so that higher values indicated more positive responses. Approximately 1/3 of the respondents (N = 695) received the text stating that the Labour Party had supported the proposal, another ca. 1/3 (N = 629) that the Progress Party had done so, while the remaining ca. 1/3 (N = 666) received the control question stating that the Norwegian Parliament had supported it.

The choice to include party cues from the Progress and Labour Party stems from their position as antagonists in the Norwegian party system and that they have been part of the right- and left-leaning government alternatives. Previous research has also indicated that populist-right parties are subject to intense loathing from supporters of mainstream parties, and it is therefore useful to include cues from parties associated with both mainstream and populist parties (Harteveld et al., 2021; Meléndez and Kaltwasser, 2021). Likewise, studies on out-party cues indicate that they could have a stronger influence when assigned to parties characterized as populist-right (Bolin et al., 2021). The distribution of responses to the question is illustrated in Figure 2, which are also broken down for left- and right-leaning respondents separately and for each treatment condition.

Figure 2. Histogram for the question about whether generous financial assistance should be given to companies with losses in turnover. Higher values denote more positive attitudes. Respondents with ideological self-placement above or below the midpoint of five. The upper panels are divided based on the assigned party cues.

Last, the information given to the control group, namely, that parliament supported the proposal, could be interpreted differently by respondents. For example, while some might interpret it in non-partisan terms, others could associate parliamentary support with the governing coalition or larger parties with more influence in parliament. Even so, it stands to reason that the Norwegian parliament should be viewed as less partisan than specific parties. Furthermore, the wording refers to a more collective political entity without being misleading to respondents since parliament supported the proposal.

Negative out-party affect

My main independent variable is negative out-party affect. I operationalize it with the standard party feeling thermometer question, asking respondents to rate the parties on a scale from one (intensely dislike) to seven (intensely like). However, for the analyses related to the first survey experiment—that is, management of the coronavirus crisis—I use like-dislike ratings during wave 17 of the NCP, which is before the coronavirus outbreak in Norway.

The same three parties—the Conservative, Liberal and Christian Democratic Party—governed Norway from January 24 in 2020 until the end of the last survey wave in November the same year. As a result, I measure out-party affect among supporters of the remaining six parliamentary (opposition) parties as the average unweighted like-dislike rating of these three parties, recoded so that higher values denote greater dislike. For supporters of the three governing parties, on the other hand, out-party affect is the average unweighted dislike of the parties to the left and center-left: The Center Party, Labour Party, Socialist Left, Greens, and the Red Party. The reasoning behind excluding the right-wing Progress Party is that they governed together with the Conservative Party from 2013 to 2020, and with the Conservatives and Liberals during the end of this period (Aardal and Bergh, 2022, p. 2). Hence, it is plausible that supporters of the Conservative Party especially perceived them as part of a right-wing or center-right political alternative.

Accordingly, the variable for average out-party dislike ranges from 1 to 7 in theory, with higher values denoting greater dislike. However, there are, of course, very few respondents who strongly like all out-parties, meaning that the actual scale is more condensed (see Figure A1 in Appendix). For the second experiment concerning the proposal for financial assistance, I simply use the like-dislike ratings (scale 1–7) from the latest available survey wave toward the two parties for which respondents received party cues. The measure is then recoded so that higher values indicate greater dislike.

Last, when evaluating the results, it is important to keep in mind that studies of affective polarization often use survey items that measure traits, feelings and social distance toward voters of the parties (Iyengar et al., 2012; Druckman and Levendusky, 2019; Knudsen, 2021). Specifically, the study by Druckman et al. (2021a) employs an aggregated index consisting of party feeling thermometer evaluations as well as three items that capture trust, traits and social distance toward the out-party's voters. These measures are closer to the operationalization used in psychological research of how group identities spill over into intergroup conflict.

Control variables

To operationalize out-party dislike and breakdown the sample along partisan lines, a variable is needed to determine the respondents' party support. For this, I use a variable during the 17th (Study 1) and 19th (Study 2) waves that asks which party the respondent would vote for if there was an election tomorrow.

My measure of out-party affect described above focuses on the negative component of partisanship. However, most of the research on social identity theory and group conflicts views in-group identification as the central component that leads to out-group derogation and processes of group differentiation (Stets and Burke, 2000). Therefore, I also include a dummy variable for the respondents' highest like-dislike evaluation of a party, in-party affect. I coded those who rated a party at the highest mark of seven (ca. 24 percent) as one, while all other scores were coded as zero.²

It is plausible that ideology could affect both dislike of the competing parties as well as people's attitudes toward the two survey experiments. In addition, it is also valuable to add this control as a way to separate the effect of the more instrumental concept of ideological self-positioning from the more expressive concept of party affect. Because of this, I include a variable for the respondents' left-right (ideological) placement (scale 0–10), with higher values indicating a more right-leaning position.

Clearly, some groups are more vulnerable to the coronavirus pandemic, especially older people, those with prior health conditions, and groups with lower socioeconomic status (Van Bavel et al., 2020). Accordingly, the analyses include control variables for age (seven age cohorts), income (scaled 1–8), and the respondents' subjective health (scaled 1–5). I code the variables so that higher values denote younger age cohorts, higher income, and better subjective health.

In addition, it is advisable to separate the effect of general dislike toward political institutions and political parties—i.e., general skepticism and political alienation—from out-party affect targeting political competitors. To do so, I include a categorical variable for whether the respondents reported having low, medium, or high confidence in Norwegian politicians.³ Last, the analyses include control variables for education, where the highest level (university/college) is coded as one and other responses as zero, as well as for gender.

Results

The results section describes the results from the two studies separately.

Study 1

Table 1 presents the results from ordinal logistic regression analyses where evaluations of the government's/Norway's performance are regressed on negative out-party affect. Note that the dependent variable is coded into three categories for government party supporters and six for opposition supporters because of the negatively skewed distribution discussed in the Methods section.

Table 1. Ordered logistic regression analysis of the determinants of attitudes about how the crisis has been handled.

First, models one and three show that average dislike toward opposition/government parties has a significant negative effect for opposition supporters and a non-significant positive effect for government supporters. However, when the interaction effect with treatment condition is included for government party supporters, the results show that the effect is significant for those rating the government's performance (b = 0.41, p = 0.045). Subsequently, models two and four feature an interaction effect between average out-party dislike and a dichotomous variable for whether a person evaluated the response of Norway or the conservative government (“Norway”). In line with hypothesis three, the effect is significantly weaker for those rating the country's performance; that is, out-party dislike has a less negative (positive) effect for opposition (government) supporters. While the interaction effect does not reach statistical significance at the p ≤ 0.05 level for government supporters, it comes close to doing so (p = 0.058). Consequently, the results are opposite of what Druckman et al. (2021a) found in their study of American voters, where the effect of affective polarization was significantly larger for those who evaluated the performance of the “United States” as opposed to “President Trump.”

To facilitate interpretation of the results, Figures 3, 4 report the predicted probabilities of choosing different response options across levels of out-party dislike while holding all other variables at their mean values. Starting with opposition supporters' evaluations of the government's performance in Figure 3, the two upper panels to the left show that the probability of choosing the more negative response options increases notably as dislike increases. However, it is equally interesting to note that although it becomes higher when dislike increases, the probability of choosing a negative or neutral response remains rather low, even for those who intensely disliked the governing parties. This is despite the explicitly partisan target being evaluated in terms of “the conversative government.” Next, the bottom two panels to the right show a sizable drop in the probability of choosing the most favorable options for those with higher dislike. Going from dislike scores of three to seven is associated with a ca. 35 and 19 point drop in the probability of answering agree or strongly agree, respectively.

Figure 3. Predicted probability of choosing different response options for opposition supporters at different levels of dislike toward government parties.

Figure 4. Predicted probability of choosing different response options for government supporters at different levels of dislike toward opposition parties.

The effect is, however, considerably weaker for ratings of the country's performance, but in the same direction. This is mostly visible in a declining probability of choosing the most positive evaluation as out-party dislike increases and a reverse pattern in regard to the less positive “somewhat agree” option.

Figure 4, in turn, shows the predicted probabilities for government supporters across levels of out-party dislike. First, in regard to evaluations of the government's performance, the probability of choosing the most positive evaluation becomes notably higher as dislike increases, while the effects are opposite for the less positive response options. Going from dislike scores of three to seven increases the probability of choosing “strongly agree” by ca. 38 points. Conversely, the same change is associated with a drop of ca. 11 points of choosing a less positive response of “somewhat agree” or lower. However, unlike opposition supporters, there is no noticeable tendency toward a significant effect for those rating the country's performance.

In terms of the other variables, the variable capturing the extent to which respondents had confidence in politicians is particularly interesting, both in and of itself and as a control variable. One reason for this is that populist, or anti-elitist, attitudes could be associated with a greater willingness to disregard elite consensus over the crisis response, perhaps even making political elites seem like a more cohesive outgroup than before. Moreover, there are examples of studies that show an association between populist attitudes and skepticism about public interventions to mitigate coronavirus infections (Juen et al., 2021; Bolsen and Palm, 2022). In short, the results show that going from the category that trusted politicians the least to the most is associated with a significant change in the ordered log odds of being in a higher level of ca. 1.23 and 1.17 (models one and three). In addition, the results show that younger people tended to express more negative attitudes and a significant effect of the variable for treatment condition (“Norway”).

The Appendix includes linear analyses (for opposition supporters) and bivariate analyses without control variables. When doing so, the effects of out-party dislike are somewhat larger: for instance, the effect is significant for both treatment groups (Norway and the government) for opposition party supporters, although still significantly stronger for those evaluating the government. Second, the Appendix explores the effects of variables for average out-party dislike toward all nine parties, the row minimum like-dislike evaluation toward out-parties and a version of the average dislike score where the parties are weighted by vote share.

Taken together, the results in Study 1 support the assumptions of the first and third hypotheses: out-party dislike prior to the coronavirus crisis shaped attitudes toward the response of the government, even in a less (or least) likely case such as Norway, where attitudes were overwhelmingly positive. However, the effects are weaker (opposition supporters) or non-existent (government supporters) for respondents who evaluate the superordinate category of the country's response.

Study 2

The second analysis presented in Table 2 examines whether out-party cues impacted people's attitudes about providing financial assistance to companies that had been hit with losses in turnover as a result of the pandemic and whether negative out-party affect moderated the effect. Models 1–3 display the results for supporters of all nine parties except the Progress (FRP) and Conservative Party (H), since they all, on average, viewed the FRP as a disliked out-party. Models 4–6, in turn, restrict the analyses to supporters of the center-right parties, who have competed for power against a government coalition centered on the Labour Party (DNA).

Table 2. Linear regression analysis of the determinants of attitudes toward the proposal about providing financial assistance.

Starting with Model 1, respondents who learned that the FrP had supported the proposal evaluated it ca. 0.66 scale steps lower. When considering that the standard deviation is 1.24 on a scale from 1 to 7, then the effect can be considered relatively impactful and significant at the highest level. Similarly, Model 4 shows that supporters of the right-of-center parties expressed significantly greater skepticism when they received the Labour Party (DNA) treatment (b = −0.47) compared to the “parliament” control group. Moreover, the effect is also significantly negative for the FrP treatment, although their supporters belonged to the group of right-of-center supporters.

Following this, the remaining regression models examine the extent to which like-dislike evaluations toward the two parties moderate the treatment effects. As outlined in the theory section (H2), we would expect that stronger dislike of out-parties exacerbates partisans' motivation to differentiate themselves by adopting contrary policy stances. In line with this reasoning, the results from Model 2 indicate that respondents who dislike FrP more (scale 1–6)⁴ reacted to the FrP treatment by becoming more negative to the proposal. Conversely, dislike of the FrP was associated with more positive attitudes for those receiving the DNA treatment. Although the interaction effect with the FrP treatment is not significant at the p ≤ 0.05 level, it is close to being significant (p = 0.086), and the plot of the interaction effect in Figure 5 shows that the confidence intervals for those in the FrP and parliament group do not overlap at higher levels of dislike. Moreover, dislike toward the FrP has a significant effect when the two groups who received party cues are compared against each other, as indicated by the coefficient equality statistic.

Figure 5. Predicted response by like-dislike evaluations for the FrP and parliament conditions (with 95-percent confidence intervals).

Next, Model 3 tests the interaction effects for the same group of voters, but this time focuses on like-dislike evaluations toward the DNA (scale 1–7). As the results show, respondents who disliked the DNA did not react differently when they learned that the DNA had supported the proposal. However, for the FrP treatment, the interaction effect was significant: a one-scale-step increase in dislike of the DNA was associated with a ca. 0.23 scale-step more favorable answer among those who received the FrP cue.

Models five and six restrict the analyses to supporters of the right-of-center parties and examine the same interaction effects. Since the DNA has been the strongest member of the left party alliance competing against these parties in the last elections, I start with those who received the DNA (“out-party”) treatment. For this group, the results in models five and six are insignificant and very small. For those who received the FrP treatment, on the other hand, the effects were significant and in the expected direction, as dislike toward the DNA was associated with a more positive reaction to the FrP party cue in Model 5, while the opposite was true for dislike of FrP in Model 6 (scale 1–7). The Table A6 in Appendix replicates the analyses with those who stated that they would not vote for any of the nine parliamentary parties included. In short, the most noteworthy difference is that the interaction effects of dislike toward the FrP in Model 2 are slightly stronger and statistically significant at the p ≤ 0.05 level for both the FrP and DNA party cues.

To summarize, with the partial exception of Model 2, there are consistent interaction effects in the expected direction for those who received information that the FrP had supported the proposal, meaning that higher levels of negative affect of the party decreased the respondents' support for the proposal, while dislike toward their main antagonist—the DNA—was associated with more positive answers. Somewhat surprisingly, negative party affect did not have the same effect for those who received the DNA party cue, with the exception of Model 2, where there is an effect in the expected direction. Consequently, the results show that support for the proposal was notably lower when people learned that an out-party had supported it as opposed to the parliament. Moreover, at least in some instances, this effect is moderated by negative out-party affect.

Last, since the effect was consistently found only for the FrP treatment, Figure 2 plots the predicted linear response for this treatment group and the parliament condition at different levels of party affect. The top left-hand side and bottom right-hand side panels show that the confidence intervals for the two groups overlap among those with low levels of dislike toward the FRP and it then starts to differ at higher levels. This pattern is opposite for the other two panels that examine the effect of dislike toward the DNA, as the confidence intervals overlap at higher but not lower levels of dislike.

Discussion

Although the research literature on NPID and affective polarization has grown substantially, it has been recognized by several scholars that there is still a lack of studies examining their political consequences (Iyengar et al., 2019). Against this backdrop, the first hypothesis builds on previous research by Druckman et al. (2021a) and explores how negative out-party affect prior to the coronavirus outbreak influenced subsequent attitudes about the political response in Norway. The results are partly in line with those found for the US case, as prior levels of out-party dislike were associated with more negative (positive) attitudes about how the government had managed the crisis for supporters of opposition (government) parties. Out-party dislike thus had the expected partisan effect, even though moderate levels of out-party antagonism, coupled with a consensual and successful response to the crisis, might make such an effect less likely in Norway than in other countries where the response has been characterized as more polarized (Christensen and Lægreid, 2020; Fonn and Hyde-Clarke, 2021; Gadarian et al., 2021; Kerr et al., 2021; Moss and Sandbakken, 2021).

Following this, the third hypothesis examines whether this effect was reduced when people evaluated the country's performance instead of the conservative government. In this regard, the results are contrasted with previous research that has viewed the priming of national identity as something that can reduce affective polarization (Levendusky, 2018; Iyengar et al., 2019) and, more specifically, the study by Druckman et al. (2021a) that focuses on the coronavirus response in the US. As expected, the effects of negative out-party affect are either weaker (opposition voters) or non-existent (government voters) when people rate the superordinate category of Norway's performance. Although Druckman et al. (2021a) focus on affective polarization and different questions for rating the response of Trump/the Unites States, the results differ distinctively from the ones they found. The authors conclude that asking people to evaluate the superordinate category of the US response did not reduce the effect of affective polarization compared to those who rated the performance of President Trump. This difference between the countries likely stems from the way in which political elites handled the crisis: while the response in the US was polarizing and closely associated with the president himself, Norwegian politicians managed to successfully portray an image of political consensus that was echoed by the media narrative (Christensen and Lægreid, 2020; Fonn and Hyde-Clarke, 2021; Moss and Sandbakken, 2021). Consequently, even those who strongly disliked government or opposition parties in Norway tended to have similarly positive views of how the country managed the initial phase of the pandemic.

The results are also relevant for research on crisis mitigation more broadly. From the very beginning of the pandemic, scholars and public commentary emphasized the importance of bipartisanship and cooperation, both as a way to ensure compliance with mitigation efforts and to foster effective legislation (Van Bavel et al., 2020, p. 462–464). Likewise, research from the UK, the US and elsewhere has shown that partisanship bifurcated attitudes and behavior during the pandemic (Druckman et al., 2021b; Gadarian et al., 2021), thus potentially undermining an effective collective response. The results from this paper add to this research by focusing on a context characterized by greater political consensus (see also Merkley et al., 2020; Jungkunz, 2021). In this regard, it is noteworthy that both supporters of the opposition and government parties were overwhelmingly positive of how the crisis had been managed (see Figures 3, 4). While negative out-party affect significantly influenced attitudes toward the response, primarily when the government was evaluated, this was mostly a matter of differences in how positive the respondents were or the small minority who expressed a negative opinion. One reason for this is that Norway performed well according to objective indicators, but it should also reflect the unifying message presented by political elites. This is supported by the reduced effect of negative out-party affect when people evaluated the response of Norway in general instead of the conservative government.

Subsequently, the last hypothesis examined whether random assignments to out-party cues affected respondents' evaluation of a policy proposal about providing financial assistance to companies that had suffered losses in turnover as a result of the pandemic. Additionally, it explored whether party like-dislike evaluations moderated this effect. I find partial support for the hypothesis, as voters evaluated the proposal less favorably when they received information that the two out-parties—the Labour (DNA) and Progress Party (FrP)—had supported it. The moderating effect of like-dislike evaluations, on the other hand, was confirmed more consistently only for those assigned to the FrP party cue and not for the DNA. Although speculative, one possible explanation for this difference could be that the FrP is often characterized as a populist-right party. Several studies across Western democracies have documented that populist-right parties are subject to particularly strong loathing, which could be connected to the anti-populist rhetoric that portray them as racists and threats to liberal democracy (Reiljan, 2020; Harteveld et al., 2021; Meléndez and Kaltwasser, 2021; Reiljan and Ryan, 2021). Conversely, dislike of the DNA was surprisingly modest from supporters of right-of-center parties and could thus be more related to instrumental concerns such as disapproval of the party's policies and ideology.

As previously mentioned, while the experiments do not directly test for causality by manipulating out-party dislike, the article contributes to the research literature on negative out-party affect outside of the US by examining its influence on public attitudes with both panel data and manipulation of party cues. Since out-party dislike and public attitudes tend to be reciprocally related (Druckman et al., 2021a), this provides two ways of partially overcoming this problem. The results also contribute to research on factors that could mitigate the consequences of out-party dislike and bolster a more unified response to a crisis. While not surprising, it does so by indicating how thinking about more unifying concepts, such as the parliament and country, can induce bipartisan support that also includes those who dislike out-parties, although this could depend on the circumstances of the Norwegian case. Last, the second experiment adds to previous research on the effects of out-party cues by examining the moderating influence of party affect. For example, in regard to the research on how out-party cues of populist-right parties could generate different effects than those from other mainstream parties (Bolin et al., 2021).

The paper points to some avenues for future research. First, the mixed effects in regard to how out-party dislike moderates out-party cues suggest that the consequences could depend on how polarization toward out-parties is measured or the specific features of the parties themselves. Next, studies from the US and elsewhere have shown how partisanship and affective polarization influenced attitudes and behavior toward other seemingly non-partisan aspects of the pandemic, such as scientific advice about intervention strategies (Calvo and Ventura, 2021; Bolsen and Palm, 2022). Areas such as science are likely better shielded from partisan contestation in Norway and, considering how actors such as health agencies and scientific experts played an important role in the crisis response, this could have made attitudes less colored by partisanship and negative out-party affect. It would be interesting to further examine how the politization of science could influence the consequences of NPID/affective polarization in different country contexts.

There are of course several important limitations to the study. First, out of necessity, the article focuses only on negative out-party affect and can therefore not be directly compared to the results from studies examining affective polarization toward the parties' voters. This is particularly relevant in relation to the article by Druckman et al. (2021a), since the results are contrasted against the ones they found. Yet, as mentioned earlier, it seems plausible that dislike of parties could be an equally good predictor of the type of attitudes studied in this article. Second, the fact that people were so overwhelmingly positive toward the political response should be considered when interpreting the smaller or non-existent effect of out-party dislike on attitudes about the country's performance. If elite cues signal in the same direction, then this of course undermines the motivation for directional partisan reasoning. Nevertheless, the non-polarized nature of the response is also what makes it an interesting case, and the saliency of the issue meant that it should have been easy to find critical opinions, especially outside of the political mainstream.

Data availability statement

Data is available for research and educational purposes through Sikt, https://search.nsd.no/en/series/ed271b1c-2595-47e4-8c97-3fcc00f02368. However, if the panel needs to be connected with other survey waves, then the request should be directed to digsscore@uib.no.

Ethics statement

Ethical review and approval were not required for this study in accordance with the local legislation and institutional requirements. The Norwegian Citizen Panel datasets, on which this study is based on, deals with human subjects and follows the EU General Data Protection Regulation (GDPR). According to this, a Data Protection Impact Assessment (DPIA) was conducted and approved by the University of Bergen. The DPIA was conducted in cooperation with the Norwegian Agency for Shared Services in Education and Research (Sikt). The DPIA number is 118868. In addition, the Scientific Committee of the Norwegian Citizen Panel reviews all questions that are to be fielded in the NCP, reviewing based on ethical issues, scientific issues, and on how the questions will impact the respondent. The participants provided their written informed consent to participate in this panel.

Author contributions

AR is responsible for all of the work with the manuscript.

Conflict of interest

The author declares that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/fpos.2022.944783/full#supplementary-material

Footnotes

References