Negative Campaigning and Vote Choice in Europe: How Do Different Partisan Groups React to Campaign Attacks?

Abstract

Parties spend parts of their campaigns criticizing other parties’ performance and characteristics, such as honesty, integrity, and unity. These attacks aim to negatively affect the target parties’ electoral performance. But do they work? While attacks are informative, we argue that how voters react to negative campaigning depends on their partisanship. While the target’s copartisans are more likely to get mobilized in favor of their party, the attacker’s copartisans are expected to punish the target due to their respective partisan motivations. We expect null effects for attacks for partisans of third parties as well as nonpartisans. Combining a new dataset on campaign rhetoric with survey data from eight European countries, we show support for most but not all of our expectations. These results have important implications for the electoral campaigns literature.

Keywords

partisanship party rhetoric negative campaigning vote choice

Introduction

Leading up to the 2015 election in the United Kingdom, David Cameron, the Conservative Party leader, and the party’s old guard were hard at work warning voters about a potential Labour government. The Telegraph reported the following on its front page on May 6, the day before the election:

The British people must come together and unite against the nightmare prospect of a Labour-SNP government which will “tear our nation apart,” Sir John Major [Conservative Party] says today …Meanwhile, David Cameron, the Prime Minister, described Mr. Miliband [the leader of the Labour Party] as a “very dangerous person” who is using a “con trick” to get into Downing Street (Dominiczak, 2015).

With these statements, the Conservatives attacked Labour for a potential divisive performance if the latter party was elected (“will tear our nation apart”) and criticized the character of the Labour leader (“very dangerous,” “using a con trick”). In this paper, we examine how negative campaign rhetoric like this affects the electoral performance of the target party (the Labour Party in the example above). We follow Geer (2006) and define negative campaigning as any negative reference toward rivals by highlighting their failings, misdoings, and negative character traits. These discussions can be about the leader or the party without any specific reference to any policy (“the party is incompetent”) or may be related to a specific policy (“the party is dishonest in its tax policy”).

The evidence in the literature for the electoral consequences of campaign attacks is either mixed (especially in the case of the United States, see Lau & Rovner, 2009 for a summary) or, in the comparative case, mostly missing (but see, Haddock & Zanna, 1997; Jung & Tavits, 2021; Pattie et al., 2011; Roy & Alcantara, 2016). We argue that two theories direct us to two competing expectations for how negative campaigning works for or against the target party. On the one hand, according to the information theory, by highlighting rivals’ weaknesses and negative characteristics, negative campaigning educates voters about their opponents’ flaws and should work against the target party. Yet, these attacks may also backfire and improve the target party’s electoral performance by motivating voters to rally behind it (see Banda & Windett, 2016 for a similar argument in the United States). In particular, voters who reject the negative messages and find these opportunistic or in bad taste may sympathize with the target party and turn out in favor of it in greater numbers.

Do all voters respond to negative campaigning similarly? According to the partisan-motivated reasoning theory (see, e.g., Taber & Lodge, 2006), individuals seek information consistent with their partisan beliefs and ignore inconsistent information. Hence, we expect different partisan groups to react differently to these informative campaign messages. Copartisans of the target party (i.e., those who identify with the party under attack) should be more likely to reject attacks targeting their party and be more motivated to turn out to vote in favor of their party. In contrast, copartisans of the attacking party should be more easily convinced by their party’s messages against the rival and be more likely to punish the target party. Although we also expect negative campaigning to inform the partisans of third parties and nonpartisans/independent voters, we largely expect null effects of negative campaigning on these voters’ behavior toward the target party.

We combine a new dataset on party campaign discussions (the Comparative Campaign Dynamics dataset) with survey data to test the effects of negative campaigning on vote choice across eight European democracies. Our results show support for the expectations about the target party’s copartisans (positive effects), the attacking party’s copartisans (negative effects), and the nonpartisans expectations (null effects). However, our findings go against the “third parties’ copartisans” expectation, suggesting that these voters are more likely to reward the target party. We attempt to explain these results in the results section.

Our findings have important implications, first, for the growing negative campaigning literature, which so far mainly focused on the American case (see, e.g., Lau & Rovner, 2009; Lau et al., 2007). As we detail below, there have been mixed findings regarding how attacks affect party performance in the United States. Our results suggest that being attacked may not move nonpartisans, that is, those who do not identify with a party. These are the respondents that parties aim to mobilize and convince to vote for them. Hence, it appears that being attacked does not help (nor does it hurt) the party among nonpartisans. At the same time, the results suggest that the target’s copartisans are even more likely to support their party in elections when said party is attacked. More importantly, we find that the likelihood of voting for the target party also increases among third (or other) parties’ copartisans (partisans of parties other than the target and the attacker). These results suggest that negative campaigning, on average, may benefit the target party and not work as the attacking party would hope.

Second, by unpacking the negative campaigning effects across different partisan groups, we also contribute to the growing literature examining how different personality traits condition individuals’ evaluation of negative campaigning. Nai and Maier (2021) and Nai and Otto (2021) already explore how personality traits mediate the effectiveness of negative campaigning. Our results suggest that partisanship is another important factor we should consider.

Finally, our results contribute to the partisan-motivated reasoning literature. Consistent with this literature’s arguments, the target’s copartisans ignore attacks against their own party and even become more mobilized to vote for their party when under attack. Similarly, we see the attacker’s copartisans appear to respond to their own party’s messages and become even less likely to support the target party. Yet, we also find that third parties’ partisans react to attacks by increasing their support for the target party. This latter finding suggests that partisans may not always be blind supporters of their own party. Rather, they might also be influenced by the rhetoric of other parties and change their voting behavior.

Campaign Attacks and Target Party Performance

Political parties have different goals (Strøm & Müller, 1999). Some may have office aspirations or survival concerns; others may seek policy influence by becoming the prime ministerial party, a coalition partner, or simply a pivotal party in the parliament forcing the hand of the government. Whatever the party’s ultimate goal, in the end, all of these goals require votes. Over the past decades, a sizeable spatial competition literature has examined how parties’ various strategies affect their electoral performance. The main focus in this literature has been parties’ ideological position-taking (see, e.g., Adams, 2012 for an overview of this literature). Yet, taking specific issue positions is only one strategy. Political parties can also distort each other’s issue positions (Somer-Topcu & Tavits, 2020) to convince voters that they are ideologically the best alternative. More importantly, for our purposes, parties can attack their counterparts by discussing other parties’ and their leaders’ lack of competence, integrity, honesty, unity, etc. Less is known about how these campaign attacks affect voter behavior.

To be clear, in the last two or three decades, a burgeoning literature about the U.S. case has showcased how negative campaign advertisements affect voter behavior (see Lau and Rovner, 2009 for a summary of this literature, and Banda and Windett, 2016 for a recent examination of the consequences of negative advertisements in the United States). Still, this literature has so far focused almost solely on campaigns in the United States. In addition, the findings of observational and experimental studies on the consequences of negative campaigns are mixed. As Lau et al. (2007) report, out of 43 studies examining the effects of negative campaigning on actual or intended vote choice, 12 find that attacks decrease the target party’s vote shares, with only four of them having statistically significant effects.

Compared to this U.S. literature, only a handful of studies examine the consequences of negative campaigning in multi-party systems, and most of these studies are focused on single countries. Analyzing the 2007 Scottish elections, Pattie et al. (2011) show that attacks can backfire and reduce the attacker’s vote share in Scotland. In Canada, Roy and Alcantara (2016) find that going positive is more beneficial than going negative, although negativity certainly drives attention (but does not help the attacker in electoral terms). In an experiment with college students in Canada, Haddock and Zanna (1997) show that the leader of the target party was seen more sympathetically after an attack. Most relevant for our study, and in the only comparative cross-national work, Jung and Tavits (2021) show that only voters with a left-leaning ideological position punish their own parties when their parties get attacked by other parties.¹ As Haselmayer (2019) says, “we lack research on the effects of negative campaigning in multi-party systems,” and there is a lot that we do not know about how negative campaigning in Europe affects voter behavior.

In this paper, we focus on the performance of the target party and ask whether we should expect negative campaigning to hurt their electoral performance.² Two competing expectations lead us in two different directions on this question. On the one hand, yes, we expect negative campaigns to work, that is, harm the target party. Negative campaigning provides voters with information about the target party that they would not otherwise know and educates them about the shortcomings and flaws of the party. This argument builds on the informative effects of the election campaigns and the social psychology literatures, showing that negativity attracts attention and increases knowledge. We know that voters have limited political knowledge and minimal interest in politics (Delli Carpini & Keeter, 1996; Converse, 1964). However, to make informed decisions, citizens must acquire a significant amount of information (Zaller, 1992). Election campaigns provide this information: they help voters learn about party positions and then vote according to this newly acquired information (e.g., Fernandez-Vazquez, 2014; Iyengar & Simon, 2000; Somer-Topcu et al., 2020).

As part of modern election campaigns, negative campaigning provides voters with a significant amount of information about a party. We cannot expect parties to discuss their own failures and shortcomings (Geer, 2006); thus, attacks help voters learn about political parties’ less desirable traits and issue performances. Citizens then use this negative information to inform their choices. Negativity also breeds more attention, which is required to get informed. Brians and Wattenberg (1996) show that consumption of negative advertisement in the United States is associated with greater issue knowledge, which builds on the social-psychological effect of negativity bias: voters pay more attention to negative information (Brader, 2006; Ito, Larsen, Smith and Cacioppo, 1998) and negative news receive more coverage from the media, generating more interest (Kalb, 1998). Hence, negative campaigning is more likely to expose voters to parties’ and their leaders’ negative traits (regardless of whether these depictions of parties are accurate or not), more likely to capture the attention of potential voters (Banda & Windett, 2016), and more likely to allow voters to incorporate this information in their voting decisions. In turn, we expect voters to use this negative information and vote against the target party.

Despite these informational effects, political parties do not spend their whole electoral campaign discussing other parties’ negative performance or traits. This is because the strategy of negativity may also backfire with a so-called backlash or boomerang effect (Garramone, 1984). Hence, political parties must strategically weigh the costs and benefits of a potential campaign attack against their rivals (Haselmayer, 2019). There is empirical evidence that voters react negatively to attackers (see, e.g., Ansolabehere et al. (1994) and Banda and Windett (2016) in the U.S. literature, and Haddock and Zanna (1997), Pattie et al. (2011), and Walter and Van der Eijk (2019) in the Canadian, Scottish, and U.K. contexts, respectively). How does the target party fare in a potential backlash situation, which is our focus in this paper?

In a two-party system like the United States, this backlash effect suggests that the target will likely benefit, as the attacker is punished. What about in the multi-party systems of Europe? Anecdotal evidence suggests that target parties may benefit from negative campaigns against them. One such example comes from the recent 2017 German federal election campaign. The German Social Democratic Party had an initial surge in the polls in early 2017 when they elected Martin Schulz as their new leader. However, this surge did not last long. As the party began losing support in the polls, it had to decide whether to attack the Christian Democrats and their leader, Angela Merkel. Feldenkirchen (2018) accompanied Schulz’s campaign during this time and summarized the events of those days. He reports that Schulz’s advisors warned him explicitly that if he attacked Merkel, many people who were torn between the Social Democrats and the Christian Democrats would vote for the latter (p. 98). This is because voters dislike negativity (Banda & Windett, 2016) and would likely evaluate such an attack as opportunistic, delivered by an opponent aiming to defeat the other party (Roy & Alcantara, 2016). Pinkleton (1997) showed with an experiment that the attacker was evaluated as mean-spirited. Similarly, Haddock and Zanna (1997) provide experimental evidence from the Canadian 1993 election campaign that there was more anger and disgust toward the attacking party leader, while the target party’s leader was evaluated more sympathetically. The latter finding that voters may sympathize with the target party and its leader and vote for it to punish the attacker whom they find mean-spirited and opportunistic is why we argue that negative campaigns may result in a boost for the target party.³

The Conditioning Effects of Partisanship

We argue that these competing expectations (informational value of negative campaigns vs. potential backlash against attacks) are likely due to how different partisan groups react to negative campaigning. After all, “campaign messages work their influence in concert with voters’ prevailing dispositions and sentiments” (Iyengar & Simon, 2000, p. 158). Therefore, the informational theory should work especially for the copartisans of the attacking party, who otherwise would not know much about the target. Meanwhile, negative messages might still be informative to copartisans of the target party. Still, we expect these individuals to be more likely to reject the messages and get further motivated to turn out and vote for their party. Let us elaborate on these different group dynamics.

We know from the political psychology literature that party supporters are more likely to discount political messages inconsistent with their beliefs and project their ideals onto their own party (Heider, 1958; Taber & Lodge, 2006). According to the partisan-motivated reasoning theory, voters pay more attention to messages that are consistent with their partisan identity (Bolsen et al., 2014). In addition, messages from parties the voter does not identify with are more likely to be dismissed and discounted (Aaroe, 2012; Lavine et al., 2012; Nicholson, 2012). Therefore, copartisans of the target are likely to discount or flat out reject campaign attacks against their own party.⁴ Of course, this does not mean that the informational effects of negative campaigns are not at play. These negative messages still reach the target’s copartisans; they are just more likely to be dismissed by them. Besides ignoring the negative campaign against their own party, it is also expected that these copartisans will mobilize to vote for their party under attack. Martin (2004) argues that emotions that result from campaign attacks, such as anxiety, motivate participation. When opponents paint a party in a negative light, the target’s copartisans may worry about their own party’s electoral performance. All this would strengthen their ties to their party and increase their likelihood to vote. The attackers’ copartisans, on the other hand, should be more likely to believe in these messages against the target. Therefore, we expect them to get motivated to turn out and vote for their own party in greater numbers.

Finally, we expect essentially null effects for other parties’ copartisans and nonpartisans/independents, for three reasons. First, we expect these respondents to be less attentive to messages aired by a party they do not identify with. Hence, the informational value of campaign attacks (and their negative consequences for the target party) is lower for these respondents. Second, even if these individuals hear such messages, they are less likely to be convinced to support or vote against the target party due to a lack of strong partisan motivations in favor of the target or the attacker. Third, as Ansolabehere et al. (1994) present, voters dislike negativity. This dislike may either alienate these respondents from turning out (as Ansolabehere et al., 1994 find) or encourage them to turn out and vote for the target of the attack out of sympathy. The alienation argument is especially valid for nonpartisans who are already mostly alienated from politics to begin with (Dassonneville & Hooghe, 2018). Hence, we expect these negative, null, and positive effects to cancel each other out and result essentially in null effects for other partisans and nonpartisans.

In sum, we argue that different partisan groups will respond to negative campaign messages differently. We expect (1) copartisans of the target party to become more motivated to turn out to vote for their (target) party, (2) copartisans of the attacking party to be more likely to punish the target party in response to negative information, given their partisan motivation in favor of the attacker, and (3) null effects for other parties’ copartisans and nonpartisans. To clarify, we are not the first to argue that negative campaigning should have different effects on different groups of partisans. Using three surveys from the U.S. presidential elections, Stevens et al. (2015) show strong partisan biases in how respondents evaluate negative and positive campaigning. Haselmayer, Hirsch and Jenny’s (2020) survey experiment in Germany similarly presents evidence that partisan respondents perceive negative messages about their party as less damaging than nonpartisans. Still, no work so far has explored how negative campaigning affects different partisan groups’ voting behavior, and to our knowledge, none of the previous work on negative campaigning has differentiated between copartisans of the target party, attacking party, third parties, or nonpartisans.

Research Design

To test our expectations, we need data on parties’ negative campaigning, as well as survey data for our dependent variable, vote choice, and the conditioning variable, copartisanship. The negative campaigning data come from a new dataset on political parties’ campaign rhetoric, the Comparative Campaign Dynamics (CCD) project (Debus et al., 2018). Debus et al. collected newspaper articles from 10 European countries and 21 elections between 2005 and 2015 for the 1-month period before each election. We use eight of these countries and 15 of these elections in this paper due to the availability and accessibility of the survey and polling data.⁵

The newspaper articles come from the two highest-circulating newspapers in each country, chosen to represent one center-left-leaning and one center-right-leaning newspaper to eliminate any potential ideological biases by a given newspaper. Country experts gathered all election-related front-page articles and a random sample of 5–10% of the rest of the election-related articles. Three coders, who are native speakers of the respective language, coded each article using an extensive online questionnaire. Table 1 shows the elections and newspaper coverage for each country in our data.

Table 1.

Campaign Dynamics Data Coverage and Newspapers Used.

Country	Years	Left-Leaning Daily	Right-Leaning Daily
Czech Rep.	2010, 2013	Právo	Mladá fronta Dnes
Denmark	2007, 2011	Politiken	Jyllands-Posten
Germany	2009, 2013	Süddeutsche Zeitung	Frankfurter Allgemeine
Poland	2007, 2011	Gazeta Wyborcza	Rzeczpospolita
Portugal	2009, 2011	Público	Jornal de Notícias
Spain	2008	El País	El Mundo
Sweden	2014	Dagens Nyheter	Aftonbladet
United Kingdom	2005, 2010, 2015	The Guardian	The Daily Telegraph

The CCD project provides data on (1) how parties discuss their own and other parties’ issue positions and (2) how parties discuss their own and other parties’ valence characteristics, both in relation to and independent of their issue positions. The CCD dataset uses the term valence to refer to any party or party leader characteristics that are broadly desirable (Stokes, 1992; Clark, 2009). In coding these valence characteristics, coders also identify whether these characteristics are framed as positive or negative. We use these negative references to code negative campaigning.⁶

More specifically, negative campaigning is coded as follows. Each coder first identifies the subject party, that is, the party that makes a statement via the newspaper article. After answering several questions about whether and how they discuss their own issue positions and valence characteristics, the coder is asked whether the subject party discusses another party’s issue positions and valence characteristics. If the coder identifies an issue position discussion about another party, they are then asked to indicate any valence characteristics associated with that issue. Finally, they are also asked whether the subject discusses the valence characteristics of this other party independent of an issue. The coder is expected to evaluate the issue-related and non-issue-related valence discussions in the following categories: (1) party’s honesty/integrity, (2) party’s competence/performance, (3) party’s unity, (4) party leader’s honesty/integrity/character, (5) party leader’s competence/performance, and (6) party leader’s charisma.⁷ Finally, the coder evaluates whether the specific valence characteristic is presented as positive or negative.

As an example, The Daily Telegraph reported the following in the 2015 election campaign:

In a highly personal attack, Michael Fallon, the Defence Secretary, wrote in The Times that Mr. Miliband has a “lust for power” and would betray Britain’s defenses just as he “stabbed his own brother in the back to become Labour leader” (Holehouse et al., 2015).

The CCD codes this statement twice: first, “Mr. Miliband has a ‘lust for power’” and ‘just as he “stabbed his own brother in the back to become Labour leader’” are coded as the Conservative Party (the subject party) discussing the Labour Party’s (the target party) leader’s (Mr. Miliband) honesty/integrity/character in a negative direction without any issue connection. Second, “would betray Britain’s defences” is coded as an attack on the Labour leader’s honesty/integrity/character related to the defense policy. We sum together these issue-related and non-issue-related valence attacks to finalize our negative campaigning variable. Supplemental Appendix 1 shows three snapshots of the online survey used to code the issue-related and non-issue-related valence discussions.⁸

To test our expectations regarding the copartisans of the target party and the nonpartisans, we need information about the total number of attacks by all parties on the target party. The Received Attacks variable is calculated by summing all negative party and leader valence discussions that are made about the target party (ω_other) and dividing this number by the total number of positive (α) and negative (ω) valence statements made about the target party in that election by either itself or any other party. Note that each term in the equation includes both issue-related and non-issue-related valence statements summed together.⁹

R e c e i v e d A t t a c k s = \frac{ω_{o t h e r}}{α_{s e l f} + α_{o t h e r} + ω_{s e l f} + ω_{o t h e r}}

(1)

To go over an example, in the month ahead of the 2007 Danish parliamentary elections, the Liberal Alliance party (L.A.) received a total of 25 attacks from other parties (ω_other = 25).¹⁰ The party only made two negative self-valence statements (ω_self = 2) but 15 positive self-valence statements (α_self = 15). There were two positive valence statements made by other parties about the L.A. (α_other = 2). Therefore, the independent variable, Received Attacks, is coded as 0.568 for the L.A. in 2007. In our dataset, attacks received by target parties (ω_other) constitute an average of 53% (SD 25%) of all valence campaign discussions (α_self + α_other + ω_self + ω_other) during election campaigns. To normalize the attacks received by a given party relative to all campaign discussions about this party, we calculate it as a share. Unlike raw counts, this weighing approach allows us to adjust for varying campaign message totals of different parties.¹¹ Adding self-statements to the denominator also more accurately reflects the complex real-world electoral information environment in which parties and voters make their decisions.

To test how the attacking party’s copartisans as well as third-party copartisans react to negative campaigning, we calculate two additional attack share variables based on the previous formula (Equation 1). These variables are constructed in a similar fashion to the Received Attacks variable but differ in whose statements are included in the calculation. First, Party’s Attacks focuses only on the attacking party, calculating the share of this party’s attacks on the target party. Though we use the same formula, the focus is not on the overall share of attacks a party receives from all other parties but on the share of attacks it receives from each other party competing in an election. To calculate this variable, we restructure our data into a dyadic data format, where each party is paired with all other parties in the election. The Party’s Attacks variable measures the share of valence attacks in each dyadic pair. The data are still stacked, but now each respondent enters the data as many times as there are dyadic pairs of parties in an election. We expect this variable to have a strong negative effect on the copartisans of the attacking party.

Second, Others’ Attacks is also dyadic and measures the attacks by all other parties (other than the parties in the dyad) on the target party. This variable allows us to test how respondents react to other parties’ attacks on the target party. To calculate this share, we subtract all attack statements from the party a given respondent is copartisan of from the total attack statements that a target party received during an election. Hence, the variable is calculated as the share of the valence attacks by all other parties in proportion to the valence statements by all other parties about themselves and the target party. We expect null effects for this variable.

Our dependent variable, y_ijk, is Vote Choice. We rely on the post-election surveys from each country and use the question “which party did you vote for in the most recent election?” to identify voting behavior. Most post-election surveys come from waves 2–4 of the Comparative Study of Electoral Systems (CSES) dataset. Survey data from the countries and elections that are not available via CSES were added separately. These were the British National Election Study 2010, the Danish National Election Study 2011, and the Portuguese National Election Study 2011. As stated before, our data are stacked. The dependent variable, Vote Choice, then is coded 1 for the party the respondent voted for in the most recent election and 0 for all other party-respondent pairs.

To test our partisanship conditioning hypotheses, we rely on the same survey data. CSES and the additional national election surveys use the closeness to a party question to measure partisanship. For each survey, we measure partisans as those who indicate that they are close (or closer) to a party. Copartisan is a dummy variable coded 1 if the respondent’s party identification matches the attacked party in the party-election observation. We also generate separate dummy variables for the Copartisan of Attacker, Other Copartisan, and Nonpartisan for separate analyses of our hypotheses.¹²

We also control for several variables. First, we include two measures of party performance: how the party is polling in the opinion polls at the beginning of the campaign period (i.e., 1 month before the election) (Party Performance) and how much the party’s polling performance at the beginning of the campaign period was different compared to the party’s last election result (ΔParty Performance). The election result data come from the Comparative Manifesto Project dataset (Volkens et al., 2019), and our polling data are from Jennings and Wlezien (2016) and Pereira (2019). It is crucial that we control for these factors, given that the extent of attacks a party receives is possibly a result of their performance. Larger parties and especially those gaining in the polls might be more likely to receive attacks.¹³ We also include the variable Government_(t−1), coded 1 for parties in the government before the election, as we expect that governing parties might be punished ahead of elections. We use ParlGov data (Döring & Manow, 2019) to code this variable.¹⁴

At the respondent level, we control for political knowledge, gender, and age, which should affect vote choice, and which have been commonly controlled for in the vote choice models to account for socioeconomic cleavages (see, e.g., Duch & Stevenson., 2005).¹⁵ The knowledge variable (Political Knowledge) is based on the knowledge batteries in the CSES and NES data.¹⁶ The number of knowledge items varies between different election studies and ranges from 3 to 8 items. To obtain a comparable measure across our cases, we construct our political knowledge variable such that it is the number of correct responses to three political knowledge questions. For NES with more than three knowledge items, an algorithm randomly selected three items for each individual.¹⁷ The Age and Male variables are from the CSES and NES, and the latter is coded such that 1 refers to male and 0 refers to female. Table 2 shows the descriptive statistics for our variables.¹⁸

Table 2.

Descriptive Statistics.

	Min	Max	Mean	SD
Party-level variables
Received attacks	0	0.81	0.52	0.18
Party’s attacks	0	0.73	0.16	0.19
Others’ attacks	0	0.87	0.47	0.20
Party performance	1.60	46.31	19.50	11.86
ΔParty performance	−15.28	7.36	−0.74	5.16
Government_t−1	0	1	0.43	0.49
Individual-level variables
Vote choice (DV)	0	1	0.20	0.40
Political knowledge	0	3	1.54	0.95
Age	18	106	51.21	17.34
Male	0	1	0.48	0.50

Dataset includes 16,467 individuals, 38 parties, eight countries, and 15 country-elections.

Our main independent variable is at the party-level, while our dependent variable is at the individual level. In addition, our dependent variable is a dummy variable, coded 1 if the person voted for the focal party and 0 otherwise. For these reasons, we estimate multi-level logit models with random effects at the party-election level to test our hypotheses. The model specification is identical to a conditional logit and allows us to take the different levels into account. We add country fixed effects to our models to control for any unobserved country-specific effects. In the model where we test the effect of attacks by the attacking party and by all other parties, the dataset is structured in a dyadic manner. Hence, we use party-dyad-election intercepts along with country fixed effects. Our modeling approach also allows us to address the issue of clustering in our data by specifically modeling it (with party-election and party-dyad-election intercepts). Below are the models estimated in Table 3.

Table 3.

The Effect of Negative Campaigning on Vote Choice—Cross-National Analyses.

	All	Copartisans		Nonpartisans
		Target	Attacker
	Model 1	Model 2	Model 3	Model 4
Received attacks	0.46 (0.43)	1.50* (0.55)		0.51 (0.50)
Party’s attacks			−1.27* (0.38)
Others’ attacks			1.94* (0.43)
ΔParty performance	0.01 (0.02)	0.03 (0.02)	0.00 (0.02)	−0.01 (0.02)
Party performance	0.05* (0.01)	0.02* (0.01)	0.04* (0.01)	0.06* (0.01)
Copartisan	4.54* (0.04)
Government_t−1	−0.04 (0.18)	−0.08 (0.20)	−0.03 (0.20)	−0.06 (0.21)
Political knowledge	−0.10* (0.02)	−0.03 (0.04)	−0.01 (0.04)	−0.02 (0.02)
Age	−0.00* (0.00)	0.00* (0.00)	−0.00 (0.00)	0.00* (0.00)
Male	−0.09* (0.03)	−0.10 (0.06)	0.01 (0.06)	−0.04 (0.04)
Intercept	−3.59* (0.27)	0.59 (0.33)	−5.16* (0.32)	−3.22* (0.32)
Log likelihood	−18,360.46	−3,520.09	−4,546.18	−8,084.77
Observations	69, 559	10, 698	39, 662	19, 126
Groups	63	63	328	63
Var: Groups (intercept)	0.21	0.18	0.77	0.24

Dependent variable is vote choice for the target party. Standard errors in parentheses. Models include party-election random intercepts and country fixed effects. Reported are the standard log odds coefficients. * p < .05.

To test our hypotheses for all respondents, copartisans of the target party, and nonpartisans, we rely on the respondent-party-level stacked dataset and use the following statistical model:

\begin{array}{l} y_{i j k} = & β_{0} + β_{1} {Receive Attacks}_{ij} + β_{2} Δ {Party Performance}_{ij} + β_{3} {Party Performance}_{ij} \\ + β_{4} {Copartisan}_{i j k} + β_{5} {Government}_{t - 1 i j} + β_{6} Pol . {Knowledge}_{jk} + β_{7} {Age}_{jk} \\ + β_{8} {Gender}_{j k} + u_{i j} + e_{i j k}, \end{array}

(2)

where each party i in election j is matched with every respondent k in the corresponding election study,

e_{i j k} \sim N (0, σ_{e}^{2})

is the level 1 variation, and

u_{i j} \sim N (0, σ_{u}^{2})

is the level 2 variation. To test the target’s copartisans and nonpartisans hypotheses, we run this model separately (Columns 2 and 4) for these groups and do not include the copartisan variable.

To test our hypotheses for the attacking party’s copartisans and all other parties’ copartisans (Column 3), we rely on the respondent-party dyad stacked dataset and estimate the following model:

\begin{array}{l} y_{i d j k} = & β_{0} + β_{1} {Party}^{'} {s Attacks}_{idj} + β_{2} {Other Parties}^{'} {Attacks}_{idjk} + β_{3} Δ {Party Performance}_{ij} \\ + β_{4} {Party Performance}_{ij} + β_{5} {Government}_{t - 1 ij} + β_{6} Pol . {Knowledge}_{jk} \\ + β_{7} {Age}_{j k} + β_{8} {Gender}_{j k} + u_{i d j} + e_{i d j k}, \end{array}

(3)

where each party i in dyad d in election j is matched with every respondent k in the corresponding election study,

e_{i d j k} \sim N (0, σ_{e}^{2})

is the level 1 variation, and

u_{i d j} \sim N (0, σ_{u}^{2})

is the level 2 variation.

Results

Table 3 presents the results. Column 1 shows the results for all respondents, while Columns 2–4 show the effects of attacks for the target’s copartisans (2), the attacker’s copartisans and other parties’ copartisans (3), and nonpartisans (4), respectively. To recap, we expect the copartisans of the target party to rally behind their own party when other parties attack it, partisans of the attacking party to punish the target party, and other parties’ copartisans and nonpartisans, on average, to have null effects. In Table 3, we report the log odds coefficients from the multi-level logit models. A positive (negative) coefficient indicates an increased (decreased) likelihood of voting for a party.

Our key independent variable in all models (other than Column 3) is the share of campaign attacks received (Received Attacks). Interpreting a coefficient for this variable (the effect of a one-unit increase) is a change from no attacks received (when the share is 0) to a campaign where all campaign discussions were attacks on that party by other parties (when the share is 1). Since 1 is not a realized negative campaigning share in our data, we provide, for all relevant models, the predicted probabilities over the range of actually observed negative campaigning shares in Figure 1. In the first column of Table 3, we test the effect of negative campaigning overall on all voters ahead of the election. As we can see, the share of negative campaigning a party receives during an election does not affect vote choice. While the coefficient is positive, it is not statistically significant. Still, these results are not surprising, as (1) we expect with competing theories that voters should punish or reward the target party under attack, which may cancel each other out, and (2) Model 1 pools all voters and, in the theory section, we already formulated distinct expectations for the copartisans of target and attacker as well as other partisans and nonpartisans.

Figure 1.

Predicted Values of Negative Campaigning for Different Partisan Groups. Note: Panel 1 is based on model 2, panels 2 and 3 are based on model 3, and panel 4 is based on model 4 of Table 3. Shown are predicted probabilities with 95% confidence intervals for vote choice for the observed values of Received Attacks for each country in the dataset. The values of continuous covariates are set to their mean, those of categorical covariates to their reference level (Lüdecke, 2021).

We test the different partisan dynamics by sub-setting the dataset on partisanship in Columns 2–4. As hypothesized, copartisans of the target party rally behind their own party. The statistically significant coefficient of 1.50 in Column 2 suggests that copartisans of the target party respond strongly to the negative campaigning against their party and are more likely to vote for it. In Panel 1 of Figure 1, we calculate the predicted probabilities of copartisans of the target party voting for their party as the share of negative campaigning increases. Unsurprisingly, copartisans—across all countries in the dataset—have a high baseline likelihood of voting for their party in general, and this likelihood increases across all countries as the party is subject to higher shares of negative campaigning.¹⁹ We hence find support for our first hypothesis: copartisans of the target party appear to respond to attacks on their party by turning out to vote for it.

Our expectation for the copartisans of the attacking party is that they should be motivated to turn out and vote for their party. Hence, we should see a negative coefficient for the effects of the Party’s Attacks variable on vote choice for the target party. Recall that this variable is the share of attacks on the target party from the party a given respondent is a copartisan of. The negative and statistically significant coefficient of −1.27 supports our expectation. How much does this likelihood decrease? In Panel 2 of Figure 1, we show the predicted probabilities for attacking party’s copartisans to vote for the target party as the share of negative campaigning increases. Across all countries in our analysis, copartisans of the attacking party start with a low baseline likelihood of voting for the target party. This likelihood further decreases in all countries as the share of attacks by the party the survey respondent identifies with increases. The magnitude of the effect of negative campaigning varies by country. The United Kingdom experiences the most significant drop by about two percentage points, while there are no discernible effects in countries like the Czech Republic or Denmark. Most other countries see a change of about one percentage point.

Our final hypothesis formulates expectations about the behavior of the partisans of other parties and nonpartisans. In both cases, we expect null findings. The effect for partisans of other parties is tested with the Others’ Attacks variable in Column 3. Here, we examine how much the share of negative campaigning from other parties affects the respondent’s vote choice while controlling for the share of attacks from the party a respondent identifies with (Party’s Attacks). Counter to our hypothesis, the coefficient is positive, substantively large, and statistically significant. Panel 3 of Figure 1 provides more insight. As expected, the copartisans of third parties start out with a very low baseline likelihood of voting for another party. However, as the share of attacks from other parties on the target party increases, so does the likelihood of voting for this party. While there is variation in the size of the effect, from almost non-existent in the Czech Republic to a rise from 1.5% to over 5% in the United Kingdom, the general trend suggests that copartisans of third parties appear more likely to vote for the target party when this party is under attack.

What explains these findings? While it is not possible to exactly know why we see strong positive effects of other parties’ attacks on other partisans’ voting behavior toward the target party given the data limitations, we believe that our backlash theory may explain the finding. As discussed in the theory section, while negative campaigning aims to educate respondents about the target party’s negative traits/performance (and hence encourage them to not vote for the target party), the backlash theory suggests that people may turn against the attacker. This may result in a boost for the target party if voters sympathize with the victim/target. Haddock and Zanna’s (1997) experimental work from Canada shows that the target party leader is evaluated more sympathetically, and we know that voters in general dislike negativity (Banda & Windett, 2016). While the attacking party’s copartisans are likely going to believe in their party’s negative rhetoric and punish the target, it is possible that other parties’ partisans sympathize with the target party and decide to vote for it. Future work is needed to further unpack these interesting findings.

We also expected null findings for the nonpartisans. The results in Column 4 confirm these expectations, showing no statistically significant effect of negative campaigning on nonpartisans’ voting behavior. We expected to see null effects for two reasons. First, some nonpartisans might learn about the target party’s deficiencies (correct or not) and may turn out to vote for a different party to punish the target, and some might sympathize with the target and turn out to vote in favor of the target party. We expected these effects largely to cancel each other out, resulting in null effects. Second, we expected these mostly politically alienated respondents to be further alienated by negative campaigning (Ansolabehere et al., 1994), which overall means that they would not vote in favor of the target or the attacker, resulting in null effects. The null results support these predictions, even if we cannot answer which of these mechanisms is at play. Panel 4 of Figure 1 plots the predicted probabilities for nonpartisans of every country and while there is an upward slope for each line, the confidence intervals cover significant areas, and the null finding is confirmed.

One may argue that these effects are not large or meaningful. However, these effect sizes are important in the context of Europe’s multi-party systems, where parties are fighting to increase their vote shares by even 2–3% to become the largest party, join the government coalition, or even just pass the electoral threshold and make it into parliament. Central to the relevance of our findings is the prevalence of co-, out-, and nonpartisans in the data. For example, how many of the respondents in the survey data identify with one party but then vote for another? Table 4 shows the distribution of the various cases in our data. Overall, as one might expect from reasonable intuition or our predicted probabilities, most respondents vote for the party they identify with. Yet, our results suggest that target parties gain votes from the copartisans of other parties. Table 4 also shows a significant number of people (7% of our sample) who identified with one but then voted for a different party (outpartisans who voted for a different party than their own). Therefore, the increasing likelihood of voting for the target by other partisans from about 1.5% to almost 5.5% for the United Kingdom for Germany (two very different political and electoral systems) might mean adding a significant number of voters to a party’s tally that can tip the scales toward one majority or another, or from clearing the electoral threshold or not. In addition, winning votes from other parties causes twice the damage to competitors. While winning over a nonpartisan only adds one additional vote to the tally of a party, winning over a voter from another party increases the vote tally of that party at the competitor’s expense.

Table 4.

Partisanship and Voting Behavior among the Sample of 16,467 Respondents.

Groups	Percent of Our Sample, %
Respondents who voted for the party they identify with	58
Respondents who identify with and voted for different parties	7
Respondents who did not identify with any party but voted	21
Respondents who did not vote for a party in the dataset	14

Finally, one may argue that these results are not surprising, given that political parties are more likely to attack a rival party when the rival is performing well. Hence, the likelihood of voting for the party should increase even among the copartisans of other parties, not because of the attacks but because they are performing better. However, we argue that this is not the case for two reasons. First, we control for a given party’s performance and the change of this party’s performance in the models, and still find statistically significant effects of negative campaigning on vote choice. Second, if this argument were valid, then nonpartisans should also be more likely to vote for the target party, in addition to the copartisans of other parties. However, our results show that nonpartisans do not react to the attacks. At the same time, the copartisans of other parties, within the limits of their baseline likelihood, are more likely to vote for the target party.

One may also be interested in further exploring whether negative campaigning by different parties has varying effects. For instance, the ideological distance of the attacking party to the target party, the vote share of the attacking party, and the incumbency status of the attacking party may affect how voters react to different parties’ attacks. Given space limitations, we cannot discuss these interesting potential variations here, but Supplemental Appendix 6 shows some preliminary results for these conditioning effects. Overall, while partisanship is an important conditioning variable for the effect of negative campaigning on target party vote share, the attacking party’s characteristics do not appear to affect voters’ behavior.

Conclusion

Negative campaigning is on the rise across Europe. While policy discussions still dominate elections campaigns, political parties also use election campaigns to criticize their rivals’ performance and traits. Yet, we know relatively little about how negative campaigning works in Europe. Using a new dataset on negative campaigning in eight European countries over 15 elections, we showed that voters react to negative campaigning with their vote choice. While the target party’s copartisans rally behind their party when it is attacked, attacking party’s copartisans use this information to punish the target party in elections and nonpartisans, on average, do not react to attacks. Contrary to our expectations, our results also show that other parties’ partisans turn out to vote in favor of the target party. We believe that their dislike toward negativity may motivate them to do so.

These results have important implications for our understanding of European voter behavior and for the negative campaigning literature, as described in the introduction section. Our results contribute to the predominantly U.S. and other single country-focused and mainly experimental work (see, e.g., Ansolabehere et al., 1994; Haddock & Zanna, 1997; Pinkleton, 1997) to show that negative campaigns matter. In addition, while interesting recent work has started to unpack variation across respondents’ personality traits to explain how they react to negative campaigning (see, e.g., Nai & Maier, 2021; Nai & Otto, 2021), our work is original in its focus on different partisan groups and their voting behavior in response to negative campaigning. Very few works have attempted to focus on partisanship differences for negative campaigning effects. Stevens et al.’s (2015) and Haselmayer, Hirsch and Jenny’s (2020) experimental work in the United States and Germany, respectively, focused on how different partisan groups evaluate negative campaigning. However, as noted, no work explores negative campaigning effects on different partisan groups’ voting behavior and certainly not with a focus on copartisans of the attackers, targets, other parties, and nonpartisans.

Our results suggest that negative campaigning in Europe has a bad reputation and may motivate some voters to react against the attacks and side with the target party. Hence, these results have important implications for political parties as they decide on their campaign strategies. In the theory section, we argued that political parties need votes to achieve their goals. Therefore, they aim to keep their supporters attached to the party and attract independents and other parties’ supporters. Being targeted by attacks helps with the first goal. Target’s copartisans appear to be more motivated to vote for their own party under attack. At the same time, nonpartisans, whose votes the party aims to attract by attacking a rival, do not, on average, respond to these attacks. Only the attacking party’s supporters listen and react to the negative news with their votes, and some of the other parties’ partisans also rally behind the target party. These results, then, suggest that political parties should carefully assess the possibility that their attacks may backfire, leading to increased mobilization for the target party.

Other open questions require a more detailed examination in the near future. First, our results point to the conclusion that negative attacks are disliked by voters, who may switch their votes in favor of the target party. However, we do not directly test whether voters dislike negative campaigning and if so, to what extent. Examining subjective evaluations of negative attacks may require an experimental design. Some very valuable country-specific experiments examine how voters react to negativity (see, e.g., Haddock & Zanna, 1997; Roy & Alcantara, 2016), but more comparative work is needed. Second, our focus in this paper was on the target party, and we do not know whether and when attacking parties are rewarded/punished in multi-party systems. Hence, we have not answered why political parties use negative campaigning, especially given the findings of our paper that they appear to, on average, help the target party.²⁰ Third, more work is needed to unpack the content of attacks. It would be interesting to explore, for example, whether attacks on parties versus leaders have different consequences for the target party. As politics is becoming increasingly candidate-centric across Europe (Poguntke & Webb, 2005), one might argue that leader attacks receive more attention and may be more consequential (but see, e.g., Curtice & Holmberg, 2005). Similarly, one may argue that the type of attacks matters. Our definition of negative campaigning is broad and encompasses any attacks on party/leader performance and traits. However, research suggests that dirty campaigning using defamatory and disrespectful comments is disliked more than informative and civil negative campaigning (Reiter & Matthes, 2021). Do different attacks affect the target party’s performance differently? We provide some preliminary results in Supplemental Appendices 7 and 8. These results suggest that party versus leader attacks do not result in different consequences for the target party, but the target party’s copartisans are more motivated to mobilize and vote for their party when the attacks are on character and not on performance. As discussed in the Appendix, these results are preliminary and rest on some strong assumptions we had to make to organize the data.²¹ Hence, we leave these interesting questions for future research.

Supplemental Material

sj-pdf-1-cps-10.1177_00104140221074283 – Supplemental Material for Negative Campaigning and Vote Choice in Europe: How Do Different Partisan Groups React to Campaign Attacks?

Supplemental Material, sj-pdf-1-cps-10.1177_00104140221074283 for Negative Campaigning and Vote Choice in Europe: How Do Different Partisan Groups React to Campaign Attacks? by Zeynep Somer-Topcu and Daniel Weitzel in Comparative Political Studies

Footnotes

Acknowledgments

We thank all the attendees to the departmental seminar series at the University of Strathclyde, the University of Vienna, and the University of Zurich, and the APSA 2018 conference panel and the anonymous reviewers for their constructive feedback. Both authors contributed equally. All remaining errors are our own.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: Funding for the Comparative Campaign Dynamics dataset was provided by the German National Science Foundation (DE 1667/4-1).

Author’s Notes

Replication materials and code can be found at https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/NAGTDV was published in Comparative Political Studies Dataverse (view at ). Previous versions of the paper were presented at the 2018 American Political Science Association Meeting and in the departmental seminar series at the University of Strathclyde, the University of Vienna, and the University of Zurich.

ORCID iDs

Zeynep Somer-Topcu

Daniel Weitzel

Supplemental Material

Supplemental material for this article is available online.

Notes

Author Biographies

Zeynep Somer-Topcu is an associate professor in the Department of Government at the University of Texas at Austin. Her research interests include party politics, elections, and political behavior in advanced democracies. In her work, Somer-Topcu focuses on understanding party campaign and organizational strategies and the consequences of these strategies for voter behavior.

Daniel Weitzel is a postdoc in the Department of Government at the University of Vienna. His research interests include party politics, political behavior, and public opinion. In his work, he focuses on the interaction of strategic party and voter behavior in elections.

References

Aaroe

(2012). When citizens go against elite directions: Partisan cues and contrast effects on citizens’ attitudes. Party Politics, 18(2), 215–233. https://doi.org/10.1177%2F1354068810380093.

Adams

(2012). Causes and electoral consequences of party policy shifts in multiparty elections: Theoretical results and empirical evidence. Annual Review of Political Science, 15, 401–419.

Ansolabehere

Iyengar

Simon

Valentino

(1994). Does attack advertising demobilize the electorate? American Political Science Review, 88(4), 829–838. https://doi.org/10.2307/2082710.

Atkeson

L. R.

(1999). “Sure, I Voted For The Winner!” Over report of the primary vote for the party nominee in the national election studies. Political Behavior, 21(3), 197–215. https://doi.org/10.1023/a:1022031432535.

Banda

K. K.

Windett

J. H.

(2016). Negative advertising and the dynamics of candidate support. Political Behavior, 38(3), 747–766. https://doi.org/10.1007/s11109-016-9336-x.

Bolsen

Druckman

J. N.

Cook

F. L.

(2014). The influence of partisan motivated reasoning on public opinion. Political Behavior, 36(2), 235–262. https://doi.org/10.1007/s11109-013-9238-0.

Brader

(2006). Campaigning for hearts and minds. University of Chicago Press.

Brians

C. L.

Wattenberg

M. P.

(1996). Campaign issue knowledge and salience: Comparing reception from TV commercials, TV news and newspapers. American Journal of Political Science, 40(1), 172–193. https://doi.org/10.2307/2111699.

Clark

(2009). Valence and electoral outcomes in Western Europe, 1976-1998. Electoral Studies, 28(1), 111–122. https://doi.org/10.1016/j.electstud.2008.07.009.

10.

Converse

P. E.

(1964). The nature of belief systems in mass publics. Critical Review, 18(1–3), 1–74.

11.

Curtice

Holmberg

(2005). Party leaders and party choice. In Thomassen

(Ed.), The European voter: A comparative study of modern democracies (pp. 235–252). Oxford University Press. https://doi.org/10.1093/0199273219.003.0010.

12.

Dassonneville

Hooghe

(2018). Indifference and alienation: diverging dimensions of electoral dealignment in Europe. Acta Politica, 53(1), 1–23. https://doi.org/10.1057/ap.2016.3.

13.

Debus

Somer-Topcu

Tavits

. 2018. Comparative campaign dynamics dataset. http://www.mzes.uni-mannheim.de/d7/en/datasets/comparative-campaign-dynamics-dataset

14.

Delli Carpini

M. X

Keeter

(1996). What Americans know about politics and why it matters. Yale University Press.

15.

Dominiczak

(2015). Nightmare on downing street; Sturgeon boasts of talks to make Miliband PM. major warns that deal could tear nation apart. The Daily Telegraph.

16.

Döring

Manow

. 2019. Parliaments and governments database (ParlGov): Information on parties, elections and cabinets in modern democracies development version. http://www.parlgov.org

17.

Duch

R. M.

Stevenson

(2005). Context and the economic vote: A multilevel analysis. Political Analysis, 13(4), 387–409. https://doi.org/10.1093/pan/mpi028.

18.

Feldenkirchen

(2018). Die Schulz-story: Ein Jahr zwischen Höhenflug und Absturz. Deutsche Verlags-Anstalt.

19.

Fernandez-Vazquez

(2014). And yet it moves: The effect of election platforms on party policy images. Comparative Political Studies, 47(14), 1919–1944. https://doi.org/10.1177/0010414013516067.

20.

Festinger

(1957). A theory of cognitive dissonance. Stanford University Press.

21.

Garramone

G. M.

(1984). Voter responses to negative political ads. Journalism Quarterly, 61(2), 250–259. https://doi.org/10.1177/107769908406100202.

22.

Geer

J. G

(2006). In defense of negativity. Attack ads in presidential campaigns. University of Chicago Press.

23.

Grabe

M. E.

Kamhawi

(2006). Hard wired for negative news? Gender differences in processing broadcast news. Communication Research, 33(5), 346–369. https://doi.org/10.1177/0093650206291479.

24.

Green

Bradley

Schickler

(2008). Partisan hearts and minds. Yale University Press.

25.

Haddock

Zanna

M. P.

(1997). Impact of negative advertising on evaluations of political candidates: The 1993 Canadian federal election. Basic and Applied Social Psychology, 19(2), 205–223. https://doi.org/10.1207/s15324834basp1902_4.

26.

Haselmayer

(2019). Negative campaigning and its consequences: A review and a look ahead. French Politics, 17(3), 355–372. https://doi.org/10.1057/s41253-019-00084-8.

27.

Haselmayer

Hirsch

Jenny

(2020). Love is blind. Partisanship and perception of negative campaign messages in a multiparty system. Political Research Exchange, 2(1), 1–22. https://doi.org/10.1080/2474736x.2020.1806002.

28.

Heider

(1958). The psychology of interpersonal relations. Lawrence Erlbaum Associates.

29.

Holehouse

Gosden

Riley-Smith

(2015). Election campaign turns nasty with personal attack on labour leader over defence policy and a furious riposte. The Daily Telegraph.

30.

Ito

T. A.

Larsen

J. T.

LarsenSmith

N. K.

Smith

N. K.

Cacioppo

J. T.

(1998). Negative information weighs more heavily on the brain: The negativity bias in evaluative categorizations. Journal of Personality and Social Psychology, 75(4), 887–900. https://doi.org/10.1037//0022-3514.75.4.887.

31.

Iyengar

Simon

A. F.

(2000). New perspectives and evidence on political communication and campaign effects. Annual Review of Psychology, 51(1), 149–169. https://doi.org/10.1146/annurev.psych.51.1.149.

32.

Jennings

Wlezien

(2016). The timeline of elections: A comparative perspective. American Journal of Political Science, 60(1), 219–233. https://doi.org/10.1111/ajps.12189.

33.

Jung

J.-H.

Tavits

(2021). Valence attacks harm the electoral performance of the left but not the right. The Journal of Politics, 83(1), 277–290. https://doi.org/10.1086/709299.

34.

Kalb

M. L

(1998). The rise of the new news: A case study of two root causes of the modern scandal coverage. Harvard University: Joan Shorenstein Center on the Press, Politics and Public Policy, John F. Kennedy School of Government.

35.

Lau

R. R.

Rovner

I. B.

(2009). Negative campaigning. Annual Review of Political Science, 12(1), 285–306. https://doi.org/10.1146/annurev.polisci.10.071905.101448.

36.

Lau

R. R.

Sigelman

Rovner

I. B.

(2007). The effects of negative political campaigns: A meta-analytic reassessment. The Journal of Politics, 69(4), 1176–1209. https://doi.org/10.1111/j.1468-2508.2007.00618.x.

37.

Lavine

H. G.

Johnston

C. D.

Steenbergen

M. R.

(2012). The ambivalent partisan: How critical loyalty promotes democracy. Oxford University Press.

38.

Lüdecke

(2021). sjPlot: Data visualization for statistics in social science. R package version 2.8.10. https://CRAN.R-project.org/package=sjPlot

39.

Martin

P. S.

(2004). Inside the black box of negative campaign effects: Three reasons why negative campaigns mobilize. Political Psychology, 25(4), 545–562. https://doi.org/10.1111/j.1467-9221.2004.00386.x.

40.

Nai

(2018). Going negative, worldwide: Towards a general understanding of determinants and targets of negative campaigning. Government and Opposition, 55(3), 430–455. https://doi.org/10.1017/gov.2018.32.

41.

Nai

Maier

(2021). Is negative campaigning a matter of taste? Political attacks, incivility, and the moderating role of individual differences. American Politics Research, 49(3), 269–281. https://doi.org/10.1177/1532673x20965548.

42.

Nai

Otto

L. P

(2021). When they go low, we gloat: How trait and state Schadenfreude moderate the perception and effect of negative political messages. Journal of Media Psychology: Theories, Methods, and Applications, 33(2), 156. https://doi.org/10.1027/1864-1105/a000283.

43.

Nicholson

S. P.

(2012). Polarizing cues. American Journal of Political Science, 56(1), 52–66. https://doi.org/10.1111/j.1540-5907.2011.00541.x.

44.

Pattie

Denver

Johns

Mitchell

(2011). Raising the tone? The impact of ‘positive’and ‘negative’campaigning on voting in the 2007 Scottish Parliament election. Electoral Studies, 30(2), 333–343. https://doi.org/10.1016/j.electstud.2010.10.003.

45.

Pereira

M. M.

(2019). Do parties respond strategically to opinion polls? Evidence from campaign statements. Electoral Studies, 59(4), 78–86. https://doi.org/10.1016/j.electstud.2019.02.014.

46.

Pinkleton

(1997). The effects of negative comparative political advertising on candidate evaluations and advertising evaluations: An exploration. Journal of Advertising, 26(1), 19–29. https://doi.org/10.1080/00913367.1997.10673515.

47.

Poguntke

Webb

(2005). The presidentialization of politics in democratic societies: A framework for analysis. In Poguntke

Webb

(Eds.), The presidentialization of politics: A comparative study of modern democracies (pp. 1–25). Oxford University Press Oxford. https://doi.org/10.1093/0199252017.003.0001.

48.

Reiter

Matthes

(2021). ‘The Good, the Bad, and the Ugly’. A panel study on the reciprocal effects of negative, dirty, and positive campaigning on political distrust. Mass Communication and Society.

49.

Roy

Alcantara

(2016). Fighting fire with fire: The implications of (not) going negative in a multiparty election campaign. Canadian Journal of Political Science/Revue Canadienne de Science Politique, 49(3), 473–497. https://doi.org/10.1017/s0008423916000548.

50.

Schickler

Green

D. P.

(1997). The stability of party identification in western democracies: results from eight panel surveys. Comparative Political Studies, 30(4), 450–483. https://doi.org/10.1177/0010414097030004003.

51.

Somer-Topcu

Tavits

(2020). Message distortion as a campaign strategy: Does rival party distortion of focal party position affect voters? University of Texas at Austin. [Unpublished manuscript].

52.

Somer-Topcu

Tavits

Baumann

(2020). Does party rhetoric affect voter perceptions of party positions? Electoral Studies, 65(3), 102–153. https://doi.org/10.1016/j.electstud.2020.102153.

53.

Stevens

Allen

Sullivan

Lawrence

(2015). Fair’s fair? Principles, partisanship, and perceptions of the fairness of campaign rhetoric. British Journal of Political Science, 45(1), 195–213. https://doi.org/10.1017/s0007123413000045.

54.

Stocké

(2007). Response privacy and elapsed time since election day as determinants for vote overreporting. International Journal of Public Opinion Research, 19(2), 237–246. https://doi.org/10.1093/ijpor/edl031.

55.

Stokes

(1992). Valence politics. In Kavanagh

(Ed.), Electoral politics (pp. 80–100). Clarendon University Press.

56.

Strøm

Müller

W. C.

(1999). Political parties and hard choices policy, office, or votes? In Müller

W. C.

Strøm

(Eds.), How political parties in Western Europe make hard decisions (pp. 1–35). Cambridge University Press.

57.

Taber

C. S.

Lodge

(2006). Motivated skepticism in the evaluation of political beliefs. American Journal of Political Science, 50(3), 755–769. https://doi.org/10.1111/j.1540-5907.2006.00214.x.

58.

Volkens

Krause

Lehmann

Matthießz

Merz

Regel

Weßzels

(2019). The Manifesto data collection. Manifesto project (MRG/CMP/MARPOR). Version 2019b. Manifesto Research Group.

59.

Walter

A. S.

Van der Eijk

(2019). Measures of campaign negativity: Comparing approaches and eliminating partisan bias. The International Journal of Press/Politics, 24(3), 363–382. https://doi.org/10.1177/1940161219843404.

60.

Weitzel

(2021). Valence attacks in multi-party elections. Daniel Weitzel. [Unpublished manuscript].

61.

Zaller

J. R.

(1992). The nature and origins of mass opinion. Cambridge University Press.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.56 MB