Introduction

This review discusses why parties and candidates choose to criticize their competitors rather than emphasizing their own policy proposals or highlighting candidate attributes. Moreover, it deals with literature on how this strategy affects voters and general perceptions of democracy.

The origins of negative campaigning probably concur with the emergence of political competition and electoral campaigns. Sources go back to 64 BC, when Quintus Tullius Cicero, probably among the first spin-doctors in the world, drafted a letter of advice to his brother, Marcus Tullius Cicero, then running for the consulate. He insisted on including ‘negative campaigning’ in the campaign, to remind the people ‘(…) of what scoundrels your opponents are and to smear these men at every opportunity with the crimes, sexual scandals, and corruption they have brought on themselves’ (Cicero 2012).

Several centuries later, negative campaigning ‘took off’ in early US campaigns. A prominent example is the 1800 presidential race opposing John Adams and Thomas Jefferson. Notably, this was also the first and only example in US history, when a president was running against his former vice president. In the campaign, both camps launched strong, sometimes anonymous, personal attacks in newspapers or secretly funded pamphlets. Americans were warned that ‘murder, robbery, rape, adultery and incest, will openly be taught and practiced, the air will be rent with the cries and distress, the soil soaked with blood, and the nation black with crimes’Footnote 1 if Jefferson were to be elected. Similarly, John Adams was characterized as ‘hideous hermaphroditical character, which has neither the force and firmness of a man nor the gentleness and sensibility of a woman’ (Callender 1800) by his opponents.

Today, parties and candidates around the world ‘go negative,’ and they may choose among a palette of tools. In 2016, Donald Trump’s campaign team used footage of Hillary Clinton’s collapse at a campaign event for an ad suggesting a lack of ‘stamina’ to face the challenges of presidency.Footnote 2 In the 2010 British election, the Tories produced a series of posters attacking Prime Minister Gordon Brown, who was held responsible for, ‘taking billons from pensions,’ having ‘doubled the national debt’ or having ‘let 80,000 criminals out early.’ During the French 2017 presidential election, rumors, suggesting that the later-elected president, Emmanuel Macron, ‘was part of a secret cabal,’ ‘worked for the Rothschilds’ or ‘was gay,’ were spread.Footnote 3 With Marine Le Pen and Jean-Luc Mélenchon, these elections also witnessed successful negative campaigns denouncing the ‘established’ political class and fueling public Euroscepticism or nativism. Together with other populist extremist candidates, they obtained almost half of all votes in the first round. Their success mirrors that of similar parties across Europe and abroad and is typically associated with negative political communication.

These examples illustrate that negative campaigning takes a variety of forms. It may cover substantive criticism, such as disagreement between two parties or candidates over a specific policy, character assassinations, pejorative language or insinuate rumors about a politician’s very private life.

In light of its worldwide proliferation and increasing public debates about its (potentially negative) effects on democracy, research on negative campaigning has been flourishing in the past decades. This manuscript reviews the existing body of literature with a particular focus on multi-party competition. The review tackles three major questions related to negative campaigning research: (1) What is negative campaigning and how should we measure it? (2) Why do political actors engage in it? and (3) What are its consequences?

Understanding why political actors use negative campaigning and how it affects voters is important for at least three reasons. First, parties and candidates excessively use negative campaigning because they believe that it helps them to win votes. Second, criticizing government failure, blaming corruption or public mismanagement and providing voters with electoral alternatives through contrasting policy proposals are key functions of modern democracies and may help voters to make more informed voting decisions. Third, public concern about potential detrimental consequences of negative campaigning and studies pointing to a decline in turnout or an increased disaffection with democratic politics highlight the importance of understanding the causes and consequences of negative campaigning. Even more so in the context of successful populist radical and extremist parties, whose fundamental opposition to the status quo and ‘established parties’ may deteriorate political discourse and lead to polarization and party system fragmentation.

The road thus far: research on negative campaigning

Despite its rich tradition, the wealth of its means and increasing public attention, empirical research into negative campaigning only got off slowly in the 1990s. Based on a keyword search in the ProQuestFootnote 4 database, Fig. 1 illustrates that the number of annually published articles was at the low binary level at the start of the decade. From the mid-1990s on, the number of yearly articles grew steadily. Research on negative campaigning finally took off in mid-2000 with more than 200 research articles dealing with the topic in each year and about 100 from political science. (Other disciplines are psychology, economics, or marketing researches.) Since then, negative campaigning has become a prominent topic in political science research with about 100 research articles per year. Many studies have investigated why parties and politicians attack their opponents and explored its consequences on vote choice, turnout and perceptions of democracy (see Nai and Walter 2015; Lau and Brown Rovner 2009 for reviews).

Fig. 1
figure 1

Academic research on negative campaigning. Note: ProQuest results including ‘negative campaigning’. n = 2335 (all) and 1549 (Political Science)

Full size image

Before reviewing the existing literature on negative campaigning, it is important to stress the issue of properly defining and operationalizing it. The question of what constitutes negative campaigning and how to measure it has been widely debated in the literature (see Nai and Walter 2015 for an overview). The conceptualization of negative campaigning may determine both, findings on party strategy and its consequences.

Defining negative campaigning

The majority of studies on negative campaigning define it as ‘attacking’ an opponent (Geer 2006). Geer (2006: 23) argues that his definition of negative campaigning is, ‘(…) simple and straightforward: negativity is any criticism levelled by one candidate against another during a campaign.Footnote 5 Under this definition, there is no gray area. (…). Any type of criticism counts as negativity.’ Accordingly, all, remaining campaign communication is thus ‘positive’ campaigning, which includes the emphasis of own policy positions or a party’s record in government (Benoit 1999).Footnote 6 Figure 2 provides a simplified graphic illustration of such a binary definition of negative campaigning.

Fig. 2
figure 2

Binary definition of negative campaigning

Full size image

This common definition puts the focus on ‘attacks.’ However, the negative category on the right side of Fig. 2 still covers a vast array of messages including comparative advertising, dirty attacks or mudslinging.

This conceptualization of negative campaigning has been criticized for being too broad and unspecific. Jamieson et al. (2000) advise against conflating and obscuring legitimate and illegitimate attacks and suggest to differentiate between ‘contrast or comparative’ advertisements and ‘attacks.’ Sigelman and Kugler (2003) reveal substantive inconsistencies in perceptions of negative campaigning in the scientific literature and among voters. The standardized, dichotomous definition and measurement of negative campaigning does not reflect how voters perceive negative party communication, because most voters only disapprove negative messages that are unsubstantial, focus on apolitical candidate attributes or use extreme language (e.g., Lipsitz and Geer 2017; Mattes and Redlawsk 2014; Mutz and Reeves 2005). Hence, Lipsitz and Geer (2017) urge researchers to collect data that are consistent with the public’s understanding of the concept if they want to understand its effects on voters.

The inconsistency between a dichotomous conceptualization and a more nuanced perception of voters could account for the ambiguity of findings on the effects of negative campaigning as reported in meta-analyses (Lau et al. 1999, 2007; also see below). As these studies do not account for differences within negative messages, they may both overstate or understate its effects depending on the distribution of unsubstantial, personal attacks or pejorative language. Hence, ‘[e]mbracing the variance in the content and tone of messages may help explain whether negative messages enhance or depress turnout’ (Fridkin and Kenney 2011: 323). Scholars examining differences between weak expressions of criticism and strongly worded attacks or uncivil messages find that exposure to the latter may produce negative feelings about politicians and democracy (Fridkin and Kenney 2008, 2011; Brooks and Geer 2007; Mutz and Reeves 2005).

Typically, these operationalizations use a twofold (civil versus uncivil) category scheme. Yet, there are concerns about how such a measurement travels to large-scale content analyses of party communication, such as advertisements or press releases. The question, when a political message ‘crosses the line,’ is challenging for empirical research: ‘The more subjective and complicated a coding scheme, the more difficult (and thus expensive) it is to use, and typically the less reliable are its results’ (Lau and Brown Rovner 2009: 292).

Recent studies apply a graded conceptualization of negative messages (Haselmayer and Jenny 2017, 2018; Rudkowsky et al. 2018). In the context of multi-party competition and coalition governance, this shows that coalition parties refrain from strongly worded attacks against their coalition partners even if they criticize each other frequently (Haselmayer and Jenny 2018).

These studies rely on sentiment analysis, which aims at detecting and assessing expressions people use to evaluate persons, entities or events (Liu 2012). It identifies the polarity (positive, neutral, negative) of texts and their strength or intensity (how positive/negative). Thus, it may provide a graded measure of negative campaigning.

Determining sentiment (strength) is easier than to establish, if a message is (un)civil or (un)substantial and comes closer to general perceptions of negativity. For example, Brooks and Geer (2007: 5) define incivility as ‘claims that are inflammatory and superfluous.’ The difference, however, is ‘two strong, pointed words’—such as cowardly or utterly—that transform a ‘civil negative’ message into an ‘uncivil negative’ one (Brooks and Geer 2007: 5; examples in Appendix A). Thus, implicitly, these authors rely on negative sentiment strength to determine the ‘(un)civility’ of a campaign message, but use a dichotomous measurement and rather unspecific labeling.

Figure 3 provides a simplified representation of such a conceptualization based on a sample of hypothetical campaign messages.

Fig. 3
figure 3

Graded conceptualization of negative campaigning

Full size image

Thus far, there have been doubts on how to reliably gather a more nuanced measure of negative campaigning in quantitative studies (Lau and Brown Rovner 2009). However, crowdcoding, that is the large-scale online coding of texts using lay coders (Benoit et al. 2016), has produced promising results for large-scale text analysis of negative campaigning (Haselmayer and Jenny 2018, Haselmayer et al. 2017). Besides directly scaling the sentiment strength of texts, crowdcoding enables researchers to build sentiment dictionaries or to train machine learning algorithms for large-scale applications to political communication (Rudkowsky et al. 2018; Haselmayer and Jenny 2017).

Data sources and research designs

Despite the issue of conceptualizing negative campaigning, emphasizing the impact of data selection is of similar importance. Different sources for measuring negative campaigning may affect the results obtained (Haselmayer et al. forthcoming; Lau and Brown Rovner 2009). Researchers may choose among a palette of sources of party communication, such as advertisements, press releases, social media or media reports. The most important decision is probably between direct and indirect communication channels.

Biases in the representation of actors and campaign tone suggest that research into (negative) campaigning using indirect channels, such as newspaper articles, is likely to produce different results than studies based on direct party communication, such as advertisements or social media (Haselmayer et al. forthcoming). Accordingly, scholars should select communication channels that fit best with their research goals (Bodlos 2015; Elmelund-Præstekær and Molgaard Svenson 2014; Elmelund-Præstekær 2010; Walter and Vliegenthart 2010; Ridout and Franz 2008). Those under direct party control are more suitable for studying party strategies (e.g., choice of target, issue-based negative campaigning). On the other hand, research on voter perceptions of negative campaigning should look at communication channels that reach voters directly, such as media reports or social media.

Party incentives to ‘go negative’

As an electoral strategy, negative campaigning aims at persuading risk-averse voters ‘not to vote’ for a party or candidate and to mobilize own supporters (Riker 1996, 1991; Ansolabehere et al. 1994; Lau 1985). Hence either by capturing or appealing to voters or by deterring them from casting a vote, negative campaigning should help attacking parties to maximize votes—either in absolute or relative terms.

Rational parties will ‘go negative’ if the presumed benefits outweigh its potential costs. They will attack if they expect the damage done to the target to be greater than the risk of alienating (potential) voters. The risk stems from potential backlash or boomerang effects (Garramone 1984). If potential voters or partisans dislike negative campaigning, they might withdraw their support if messages exceed their individual levels of acceptance for this campaign strategy.

A more general explanation for the use of negative campaigning comes from cognitive psychology and the ‘negativity bias.’ Accordingly, individuals pay more attention to and give more weight to negative information, compared to positive one (Soroka 2014; Baumeister et al. 2001; Rozin and Royzman 2001). Hence, negative campaigning is a promising strategy to raise awareness and gain publicity (Haselmayer et al. forthcoming). Communication research attests that the presence of negativity or conflict increases the ‘newsworthiness’ of stories and events with journalists reporting more on negative news (Galtung and Holmboe Ruge 1965).

Based on these broad ideas, researchers have derived a set of expectations for the behavior of political actors. Basically, the literature expects three main factors to determine which candidates or parties will ‘go negative’ during an election campaign: (1) candidate or party attributes, (2) ideology and (3) campaign context.

At the level of candidates (or parties), incumbents should rely less on negative campaigning than the opposition. Incumbents benefit from their ability to emphasize their record in government and should therefore run rather positive campaigns (Benoit 1999). Also, those with fewer resources should rely on negative campaigning to increase visibility of their campaign. This relates to the surplus of information to negative messages by both, media and the public (Druckman et al. 2009; Petersen and Djupe 2005; Lau and Pomper 2004; Haynes and Rhine 1998). The literature on gender differences in negative campaigning has produced mixed results on the gender of the attacker and more consistent evidence that men are more often targets of negative campaigning than women (Lau and Pomper 2004; Kahn and Kenney 2000; Kahn 1993).

Turning to the ideological factors determining negative campaigning, parties and candidates may use it to shift their opponents’ ideological position, thereby reducing their appeal to undecided voters (Harrington and Hess 1996). Political actors may also select particular issues for maximizing the impact of negative campaigning. Following issue-ownership theory (Petrocik 1996; Budge and Farlie 1983), they may attack on topics where they have a competence advantage in the voters’ eyes to increase the credibility of attacks (Damore 2002; Riker 1996) or attack their opponents’ weak spots (Geer 1998).

Context includes the position of parties in the polls or the closeness of the race. Parties or candidates leading in the polls should predominantly rely on positive messages, because they can secure their lead by appealing to their supporters through these kinds of messages, whereas challengers have to resort to negative campaigning to catch up with them (Skaperdas and Grofman 1995).Footnote 7

Challengers predominantly use negative campaigning to raise attention of their campaign (Druckman et al. 2009; Haynes and Rhine 1998). If the outcome of an election is uncertain, the level of negative campaigning by frontrunners and challengers should increase (Druckman et al. 2009; Lau and Pomper 2004; Skaperdas and Grofman 1995).

A related argument states that the use of negative campaigning will spread in the course of a campaign as parties reiterate attacks with counterattacks (Lau and Pomper 2004; Damore 2002).

These accounts for explaining the behavior of parties and candidates come from the context of the US two-party system and assume that negative campaigning is a zero sum game, where vote gains by party A translate to votes lost by party B. How well do these general expectations and observations travel to European multi-party systems, where different institutional rules and traditions of governance are likely to shape negative campaigning?

Negative campaigning and multi-party competition

Many studies have occasionally touched upon the issue of negative campaigning (see Nai and Walter 2015 for an overview). For the French case, Yanoshevsky (2009) finds that video blogs in the 2002 French presidential election contained more negative than positive content, whereas Dolez and Laurent (2007) describe negative campaigning in the Socialist nomination campaign.

However, only a few studies have empirically studied negative campaigning in a handful of Western European countries (e.g., Haselmayer and Jenny 2018; Maier and Jansen 2018; Dolezal et al. 2015; De Nooy and Kleinnijenhuis 2013; Nai 2013; Walter 2012; Curini and Martelli 2010; Elmelund-Præstekær 2008; Hansen and Pedersen 2008). Thus, despite a growing interest in negative campaigning in Western European countries, we still know relatively little about how negative campaigning and its effects differ in these party systems as most research transferred hypotheses from the USA.

Hence, similar to incumbents in US elections, there is strong evidence that opposition parties are more likely to ‘go negative’ than government parties (Hansen and Pedersen 2008; Elmelund-Præstekær 2010; Walter and Van der Brug 2013), which are in turn more likely to be targets of negative campaigning (De Nooy and Kleinnijenhuis 2015; Dolezal et al. 2015; Walter 2014a).

Many researchers emphasize that multi-party competition obscures the electoral risks and benefits of negative campaigning (Walter 2012; Elmelund-Præstekær 2010, 2008; Hansen and Pedersen 2008). If two parties attack each other, voters may still decide to cast the ballot for a third party. For example, an undecided voter may follow corruption allegations from party A against party B and disregard to vote for party B. Yet, she may also dislike negative campaigning and exclude party A from her consideration. In a two-party system, this voter might stay away on Election Day, which would benefit neither of the parties. In a multi-party system, they could simply vote for party C, which runs an entirely positive campaign.

A tradition of coalition governance introduces further nuances: During elections coalition partners need to sharpen their profiles to distinguish themselves from each other (Sagarzazu and Klüver 2017). On the other hand, even opposition parties aspiring government membership may have to restrain from heavy attacks against potential coalition partners to preserve their office-seeking aspirations (De Nooy and Kleinnijenhuis 2015). The complexity of these strategic incentives results in ambiguous findings on the amount of negative campaigning among government parties: Whereas some find coalitions parties to restrain from criticizing each other (Walter 2012; Elmelund-Præstekær 2008, 2010; Hansen and Pedersen 2008), they attack each other frequently during some Austrian and Dutch elections (De Nooy and Kleinnijenhuis 2015; Dolezal et al. 2015).

Some authors further account for ‘coalition potential’—that is, the likelihood of political parties and candidates to enter future coalition governments. Thus far, these studies use a varying set of single indicators, such as party size, government experience or left–right positions to account for a party’s likelihood to participate in a future government coalition (Walter et al. 2014; Walter and van der Brug 2013; Elmelund-Præstekær 2008, 2010; Hansen and Pedersen 2008). Results from these studies show mixed results regarding the characteristics of senders, targets and their relationship across countries and elections (Walter 2014a).

Turning to the content of campaign messages, Elmelund-Præstekær (2011) finds Danish parties to go negative on issues that they do not own. He argues that parties ‘go negative’ when an issue is too salient to avoid—for example because the media highlights a particular topic, such as immigration. A party with low credibility on that issue will find it hard to make a positive statement or claim a record on it. Instead, it will resort to negative campaigning to damage their opponent’s issue reputation ‘and perhaps in the longer run try to conquer the issue ownership in question’ (Elmelund-Præstekær 2011: 212).

At the level of candidate characteristics, recent evidence suggests that the social context matters for gender differences in negative campaigning: Women adhere more to a stereotypical behavior (and thus less negative campaigning) when their parties have an uneven gender balance (Ennser-Jedenastik et al. 2017). Maier and Renner (2018) suggest that men adopt their behavior when confronting a female candidate in TV debates: They find that attacks are less frequent in mixed-gender debates.

Recent research further attests that parties retaliate attacks from their competitors, which mirrors expectations from the USA (Dolezal et al. 2016). Moreover, there is a division of labor within parties: Cabinet members or party leaders refrain from negative campaigning, whereas party floor leaders and general secretaries will attack rival parties and candidates (Dolezal et al. 2017). This is quite similar to evidence from the USA where parties outsource negative campaigning to PACs and SuperPACs (Brooks and Murov 2012; Painter 2014).

Overall, and despite the growing number of studies on negative campaigning in the context of multi-party or multi-candidate competition, there is no pertaining evidence for structural differences stemming from the different pattern of party competition. However, the majority of this research has not developed and tested original theoretical expectations and uses the ‘standard’ conceptualization of negative campaigning (Geer 2006).

Accounting for the sentiment strength of negative messages, Haselmayer and Jenny (2018) show that negative campaigning among coalition partners is less ‘virulent’ than that of parties that cross the government-opposition divide. This finding could indicate that a dichotomous conceptualization of negative campaigning does not perfectly fit the complex campaign environment of European countries. Moreover, it underscores that research into party strategies would benefit from theories accounting for the peculiarities of multi-party competition and coalition governance.

Negative campaigning and the voter

Turning to the consequences of negative campaigning, political practitioners typically assume that negative campaigning ‘works,’ which provides an important explanation for the proliferation of this campaign strategy (e.g., Kamber 1997).

However, two meta-analytical reviews of the literature on negative campaigning effects on voting (intentions) conclude that negative campaigning is not a particularly effective campaign strategy (Lau et al. 1999, 2007). Comparing studies on evaluations of sponsors and targets of an attack, Lau et al. (2007) find a majority of studies reporting that negative campaigning lowers the evaluations of a target (e.g., Kahn and Kenny 2004). Yet, attacking an opponent is a risky strategy. Thus, there is also some evidence showing that voter evaluations of the attacking party suffer considerably (also see: Kahn and Kenny 2004). Overall, this backlash effect is slightly stronger and more consistent than the effect on target evaluations in the meta-analysis (Lau et al. 2007). On balance, there is no evidence supporting common wisdom about negative campaigning representing an effective strategy for maximizing votes.

Any test of causal relationships between campaign strategies and its effects face several issues stemming from the anticipation of election results or multicollinearity among some of the independent variables. This is particularly relevant in single election and/or country studies, which constitute the vast majority of research on negative campaigning (see Lau and Brown Rovner 2009). Thus, methodological issues account for the ambiguity of findings to some extent.

Beyond effects on party or candidate evaluations, the so-called ‘demobilization hypothesis’ (Ansolabehere and Iyengar 1995) has been central to research into the effects of negative campaigning. It suggests that negative campaigning provokes voter disaffection and lowers turnout (also see: West 2014; Ansolabehere et al. 1994; Jamieson 1992). However, the findings (Ansolabehere and Iyengar 1995; Ansolabehere et al. 1994) have been challenged by follow-up studies (e.g., Finkel and Geer 1998), which show no support for such an effect. More recent studies even argue that negative campaigning may provide voters with important information on the weak spots of political actors and should therefore enable them to make more informed political decisions (Mattes and Redlawsk 2014; Geer 2006; Lau and Pomper 2004).

The most recent meta-analysis on the effects of negative campaigning does not find any demobilizing effect (Lau et al. 2007). However, the authors present evidence that negative campaigning has the ‘potential to do damage to the political system as it tends to reduce feelings of political efficacy, trust in government, and perhaps even satisfaction with government itself’ (Lau et al. 2007: 1184).

What determines if voters are able to learn from negative messages or if they are disgusted with democratic politics? Lau and Brown Rover (2009: 300) suggest that differences in data sources, research strategies and operationalization may be the root for (some of the) inconsistencies in results of research into negative campaigning effects.

Variation in the intensity of campaign communication and the content of attacks may influence voters’ perception of negative campaign messages (Mattes and Redlawsk 2014; Fridkin and Kenney 2011; Brooks and Geer 2007; Kahn and Kenney 1999). Accordingly, positive and negative effects may depend on how strongly parties attack and whether they focus on substantive topics or political candidate attributes such as competence rather than employing pejorative language and attack apolitical candidate attributes, such as their private life. If voters react differently to negative messages of varying strength or content, this is likely to affect how they perceive and evaluate sponsors and targets of negative campaigning.

Hence, differences in the strength of negative messages may have different effects on senders, targets and perceptions of campaign communication. As political actors may deliberately determine the targets and strength of their negative messages, moving the study of negative campaigning beyond the dichotomous level described in Fig. 2 could improve our understanding of this campaign strategy and its effects.

For what is more, we lack research on the effects of negative campaigning in multi-party systems. The literature on party strategies in these countries assumes that voters dispose of varying options when parties employ negative campaigning (see above). However, these expectations have never been tested, which underscores that a lot remains to do to establish negative campaigning as comparative research field.

Conclusion

This review argues that it is important to understand why political actors choose to attack their competitors and how this affects voters. Beyond its worldwide proliferation and the fact that parties and their advisors believe that it helps them to win votes (Kamber 1997; Nai and Walter 2015), negative campaigning may help or hurt democracy. Thus, by informing voters about failures and misdoings or by providing them with electoral alternatives through contrasting policy proposals it may enable more informed electoral decisions. However, it may also distort political discourse, lead to polarization and party system fragmentation or demobilize voters and promote political disaffection.

Beyond reviewing research on negative campaigning, this manuscript has discussed its conceptualization. Several authors suggest that the established definition and operationalization lack conceptual clarity (Fridkin and Kenney 2008; Richardson 2002; Jamieson et al. 2000; Kahn and Kenney 1999; Kamber 1997; Mayer 1996; Jamieson 1992) and are far off from how ‘common people’ perceive it (Lipsitz and Geer 2017; Sigelman and Kugler 2003). Accounting for differences between various types of campaign messages shows that voters react differently to (un)civil campaign rhetoric or varying types of attacks (Mattes and Redlawsk 2014; Brooks and Geer 2007; Mutz and Reeves 2005). Moreover, a graded conceptualization of negative campaigning may account for the strategic dilemma of parties competing under the shadow of post-electoral collaboration (Haselmayer and Jenny 2018).

Thus far, we know little on how conceptual differences affect our understanding of party behavior and its consequences on voters. Therefore, investigating these differences more closely seems a worthwhile endeavor. Such research could replicate existing studies using a new conceptualization and should theorize more strongly about the expected differences and communalities of dichotomous and graded understandings of negativity.

Despite bridging conceptual issues, there is a need for theorizing more strongly about negative campaigning. The promotion of comparative, cross-country research covering multiple party systems and elections should provide an important step toward achieving this goal. Despite pioneering work on rather small sets of countries (Walter 2014b; Walter et al. 2014), studies of negative campaigning would strongly benefit from accounting for the role of political institutions (see Ridout and Walter 2015 for a single-country study with a changing electoral rule).

A cross-country research design, covering different party systems and a longer time frame could theorize about and directly model the effects of electoral rules, party system patterns (Laakso and Taagepera 1979; Sartori 1976), the prevalence of government coalitions and the impact of general ‘patterns of democracy’ (Lijphart 1999) on negative campaigning.

Such studies could test whether negative campaigning under proportional electoral rule is less (strongly) negative than in majoritarian systems. As the former typically produce more fragmented party systems, which require the formation of coalition governments, we could expect higher levels of inter-party cooperation and strategic restraint.

With regard to institutional features, France would provide a particularly interesting case for testing how structural patterns shape party behavior. The semi-presidential system with direct presidential and legislative elections requires parties and candidates to adapt their strategies to both types of elections. Similarly, two rounds of elections and frequent pre-electoral coalitions may provide interesting settings for testing new theories about party behavior. The same applies to studying the role of extremist and radical parties in promoting polarization, party system fragmentation or deteriorating campaign communication.

Moreover, most studies still focus on ‘traditional’ communication channels. Future research should also devote more attention to how political actors use social media for negative campaigning (Feezell 2017; Harder et al. 2017; Auter and Fine 2016; Gross and Johnson 2016; Ceron and d’Adda 2015). Party messages distributed on Twitter, Facebook or other social media channels are likely to differ in content and tone from more established tools such as party ads or press releases. As social media platforms are decentralized and offer ungated access for rank-and-file politicians, we could expect higher levels of negativity, as these actors are more prone to rely on news factors to attract a larger audience (Haselmayer et al. forthcoming).

There is also a need for research on how differences in the strength of negative communication affect voters. Prior research has revealed variation in voter reactions to negative messages focusing on personal characteristics or featuring uncivil language (Mattes and Redlawsk 2014; Fridkin and Kenney 2011; Brooks and Geer 2007; Kahn and Kenney 1999). Distinguishing between weaker and stronger negative messages, future studies could investigate which negative messages produce positive (learning, persuasion, mobilization) or negative (disaffection, demobilization, polarization) effects by testing how interactions of sentiment strength and content affect voter perceptions.

Such studies would also benefit from exploring variation in voters’ (negative) emotional reactions to party communication. Emotions may shape political attitudes and behaviors (Marcus et al. 2011; Brader 2005; Marcus and MacKuen 1993), but effects are not uniform. Anger encourages partisan evaluations, while anxiety is more likely to unhinge partisan information processing (Weeks 2015). Similarly, differences in the content or tonality of negative messages produce variation in emotional reactions (Mattes and Redlawsk 2014; Brooks and Geer 2007; Mutz and Reeves 2005). This suggests that partisanship, message characteristics and emotional reactions could all impact on the effects of negative campaigning. For example, virulent attacks could evoke anger among partisans and anxiety among voters with lower levels of political knowledge and involvement. This could explain affective polarization of partisans (Meffert et al. 2006; Redlawsk 2002) and demobilization or disaffection among unsophisticated voters (Weeks 2015). Such research would further benefit from accounting for individual-level characteristics, such as gender, age or political knowledge in shaping perceptions of negative campaigning.

Finally, many of the most promising avenues for future research identified so far require a cross-country or longitudinal research design. These studies may benefit from crowdcoding and (semi)-automated sentiment analysis. Both approaches facilitate and accelerate large-scale analyses of text data. Crowdsourcing lends itself to cross-country applications as coders are available for numerous countries and languages. In addition, recent attempts to multi-lingual sentiment analysis indicate promising results (Proksch et al. 2019). Using machine translation and/or expert translators, this could prepare cross-country and cross-language sentiment analyses in the future. Combining such approaches with automated clause analysis which automatically detects syntactic relations in texts (Van Atteveldt et al. 2017) would enable entirely automated analyses of negative campaigning or party interaction.

These innovations should enable large-scale, comparative, cross-country research and therefore contribute to resolving some of the most pertaining questions in the field. This would facilitate incorporating the role of political institutions or party systems and consolidate the theoretical foundation of negative campaigning.