(European University Institute)
This essay focuses on the effects of political communication processes on the attitudes and behaviors of citizens. It reviews a set of studies conducted by political scientists, mainly based on experiments, which analyze the opinion formation process among citizens, and the factors that condition it. The main thrust of these studies is that motivated reasoning limits the maneuvering space of both political actors and the media in their attempts to persuade citizens. Context characteristics (such as elite polarization), source characteristics (e.g. credibility), issue characteristics (such as familiarity, salience, and economic performance), and characteristics of voters (e.g. motivation or capacity) moderate the general thrust. The results of these studies provide evidence in support of a «realistic theory of democracy».
Keywords: motivated reasoning, agenda-setting, priming, learning, framing, blame attribution, political context, elite configuration, polarization, type of campaign, source credibility, issue familiarity, issue salience, folk theory of democracy, realistic theory of democracy.
The public opinion formation process
Lippmann (1922) in his classic treatise on public opinion claimed that it was the leader who «manufactures consent». He asserted that a revolution was taking place, infinitely more significant than any shifting of economic power (ibidem: 158): «persuasion has become a self-conscious art and a regular organ of popular government». Two decades later, Schumpeter (1942) concurred: in one of his devastating formulas he suggested that the popular will was «largely not a genuine but a manufactured will». The first empirical studies, which were published at more-or-less the same time, however, contradicted this assertion. Lazarsfeld, Berelson and Gaudet (1948) demonstrated that the presidential campaign of 1940 essentially reinforced or activated pre-existing preferences and had only limited persuasion effects. Fifty years later, Kinder (1998: 186) came to the conclusion that «as large scale, expensive experiments in political persuasion, American presidential campaigns are a bust». He gave three reasons for this conclusion: mutual neutralization of campaign messages, resistance of those with strong predispositions to persuasive messages, and a lack of attention paid to such messages on the part of the indifferent. Kinder added that, while less glamorous than persuasion, reinforcement and activation were nevertheless important, since they allowed voters to make self-conscious choices.
Studies on agenda-setting and framing set out to change this view. While not necessarily capable of changing the beliefs of citizens, the agenda-setting approach expected political actors and the media to be able to change the relative weight citizens attached to certain beliefs as compared to others, which would influence their political attitudes and voting behavior. In a series of experiments administered in the early 1980s, Iyengar and Kinder (1987) found that concern for a particular issue increased after exposure to television news. In addition, they were also able to show a priming effect: exposure to television news could alter the criteria by which citizens evaluated public officials and thus, at least indirectly, contribute to attitude change. A number of studies have subsequently confirmed the media’s priming effects. The framing approach discovered framing effects, which occur «when (often small) changes in the presentation of an issue or an event produce (sometimes large) changes in opinion» (Chong and Druckman, 2007: 104). The discovery of priming and framing effects reinforced the notion that citizens’ attitudes were not particularly stable and that there might be more room for persuasion than suggested by earlier studies. In his ground-breaking work, (Zaller, 1992) concluded that, even if people’s opinion statements might still be authentic, people did not possess «true attitudes».
More than a decade later, Druckman (2004) argued that it was premature to abandon the concepts of preferences and attitudes. Indeed, a series of studies cast renewed doubts on the possibilities of persuasion. All of these studies pointed to the mechanism of motivated reasoning through which citizens protect their pre-established opinions and attitudes against the effect of new information. Thus, Lenz (2009) called into question the priming effects. He pointed out that the findings regarding these effects were vulnerable to alternative interpretations. Indeed, Iyengar and Kinder (1987) had already considered the possibility that the effect they found was not priming but rather a form of rationalization – a «projection» effect which they, however, dismissed based on their analysis. Relying on a suitable set of panel studies, Lenz was able to show that what had initially seemed to be a priming effect was actually the combination of a learning effect with an issue opinion change: indeed, the issues in question had become more important with regard to electoral choice only for those voters who had actually learned about the respective party positions in between two panel waves. Moreover, the people who had learned did not adjust their vote choice to their own opinions on these issues, as the priming hypothesis had implicitly assumed, but in fact did the opposite, i.e. they rationalized their vote choice by adjusting their own issue opinions so as to make them more consistent with their pre-established vote choice. Lenz (2012) concluded more generally that there was hardly any evidence for policy-voting.
Another important study of framing effects by Druckman, Fein, and Leeper (2012) confirmed the importance of this kind of rationalization in an experiment that tested the respective impacts of information repetition and information searching on opinion stability. In the absence of repetitive information, they found a recency effect, i.e. the most recent frame proved to be the most effective one (see also Chong and Druckman, 2010). However, if the original information was repeated or if the respondents had the possibility to search out information, a primacy effect prevailed, i.e. competing frames presented at a later point in time were rejected in favor of the originally presented frame.
The assumption that we generally strive to protect our beliefs is based on the insights of cognitive dissonance theory (Abelson et al., 1968; Aronson, 1969; Festinger, 1957), currently one of the best-founded theories in the field of social psychology. In this perspective, emotions and affect guide political thinking to a greater extent than is usually assumed. In line with this perspective, Lodge and Taber (2013) argue that unconscious, affect-driven thinking influences political attitudes and behavior. Their central tenet is that cognition is «hot», which means that affect enters the decision stream at every stage of the process of opinion formation. Thus, prior affect – our previously held positive or negative evaluation of target objects – biases attention to and the processing of information in ways that favor acceptance of affectively congruent arguments or evidence. This implies that citizens are motivated by a desire to maintain their prior-held beliefs and feelings rather than a desire to make accurate political decisions. In line with this «motivated reasoning» hypothesis, Lodge and Taber also show that supportive arguments are seen to be stronger than arguments that are inconsistent with prior-held beliefs, that people spend more time and cognitive resources on challenging information that is contrary to their beliefs (disconfirmation bias) and that they seek out confirming arguments when they get a chance to do so (confirmation bias).
Thus, more recent studies have come full circle and confirm the original skepticism of political communication studies regarding the ability of the elite to persuade voters. On the one hand, motivated reasoning limits the maneuvering space of political actors and the media in their attempts to persuade citizens. On the other hand, however, motivated reasoning provides a mechanism that accounts for the prominence of activation and reinforcement effects for which the early empirical studies of public opinion formation have provided evidence.
The effects of political communication on the opinion formation of citizens depends, however, on various context conditions. Among these are the characteristics of the overall political context, the characteristics of the political elites (the source of the messages) and of the public at the receiving end, as well as the characteristics of the issue (the object of communication). First of all, we need to distinguish between different political contexts. The effect of political communications depends on the institutional context. As Sniderman (2000) has observed, institutions are «organizers of political choices». Sniderman and Levendusky (2007) reiterate their key insight that, «In politics, citizens do not get their choice of choices. They must select from an organized menu of choices». Thus, the choice situation varies, both from one election campaign to another and between direct-democratic referendum campaigns, as well as in issue-specific public debates in both routine and crisis situations. General election campaigns in the US seem to be most inauspicious contexts for persuasion effects. Based on a systematic meta-analysis of no less than 40 field experiments and an additional nine original experiments of the same kind, Kalla and Broockman (2018) conclude that the best estimate for the persuasive effect of campaign contact and advertising – such as mail, phone calls and advertising – on Americans’ candidate choices in general elections is zero. Campaign contact seems to have an effect when it takes place early in the campaign, but these effects decay and, by election day, converge to zero as well. The chances of communication effects occurring are higher in direct-democratic campaigns, but they depend on the characteristics of the issue at stake, and the same applies to issue-specific debates. I shall come back to these settings when I discuss conditioning by issue characteristics.
A crucial aspect of the general context that is applicable to all three types of settings refers to the configuration of the political elite. As Zaller (1992) has already shown, whether the elite are consensual or polarized has a great bearing on the effect of their messages on citizens. If the elite are in agreement, the most sophisticated citizens, who receive the most political information, tend to accept this information and follow the political elite. If, however, the elite are polarized, then public opinion polarizes as well, with the most sophisticated citizens polarizing the most, since they not only tend to receive the most information but also to resist information that is not compatible with their political pre-dispositions (see below). In an impressive survey experiment conducted in the US, Druckman, Peterson, and Slothuus (2013) document that polarization stimulates partisan motivated reasoning, which means that partisans will view their party’s frame as more effective than a frame not sponsored by their party or a frame sponsored by the opposition party. More precisely, they show that, in a non-polarized configuration, it is substance, not party endorsements, that carries the day, provided that the arguments involved vary in strength. When the situation becomes more ambivalent, i.e. when the arguments of both sides are equally strong, partisan cues have a strong effect as well. In a polarized configuration, however, substance becomes irrelevant and polarization wins out: when elites polarize on a given issue, citizens follow and polarize as well. Moreover, in a polarized configuration, citizens also view their own opinions as more valid, even though they are less well grounded.
These results speak directly to the relevance of the second conditioning factor to be discussed here, the characteristics of the source of political communication. In the context of the study of framing effects, Druckman (2001) had already shown previously, based on two laboratory experiments, that such effects critically depend on source credibility. He had already concluded that «perceived source credibility appears to be a prerequisite for successful framing» (ibidem: 1061). Moreover, he had taken these results to suggest that framing effects may occur not because elites seek to manipulate citizens, but because citizens seek guidance from elites they trust. The observed polarization effect provides a deeper level of understanding to these early results: citizens tend to follow sources they trust in conflict situations, in which the camps are clearly divided. In such situations, credibility and trust become the overwhelming determinants of opinion formation.
These results do not only apply to the US. Focusing on the interpretation of facts, Bisgaard and Slothuus (2018) show how party cues shape partisan perceptions. Based on both a quasi-experimental and an experimental study in Denmark, they show how elite partisan rhetoric shapes partisans’ perception of the state of the economy. They found that partisans respond to elite cues, but that they do so selectively: they only respond to cues from their own party. Bisgaard (2019) added a final twist to the effects of motivated reasoning: even if citizens get the facts right, they attribute blame in a selective way. In randomized experiments in both the US and Denmark during the financial crisis, partisans of different persuasions updated their economic perceptions in the same negative way. However, they used rationalizing strategies for unambiguous facts: they selectively attributed the blame for the undeniable worsening of the economic situation to the incumbents (if they were supporters of the opposition parties) or to other actors, such as the Central Bank, the financial sector or global trends (if they were supporters of the incumbent parties).
The next factor that conditions the impact of political communications concerns the characteristics of the receiving audience. As has already been alluded to in the previous discussion, the motivation and capacity of voters are crucial moderators for the effect of political communications. As Zaller (1992) has famously shown, the polarization effect is strongest among the most highly aware, who are able to resist messages that are not compatible with their own predispositions. Similarly, Druckman (2004) has shown that «experts» are less susceptible to framing effects than «non-experts». In line with these results, politically uninterested voters are more likely to be influenced by a direct-democratic campaign than interested voters (Kriesi, 2012). A striking example of the great resistance of the politically sophisticated is provided by Bartels (2008: 156), who shows that the proportion of conservatives in the US willing to admit that economic inequality has increased actually declined in the face of more politically sensitive information being made available. In other words, political sophistication not only produces sharp polarization in policy preferences, but also in perceptions of seemingly straightforward objective facts. Lodge and Taber (2013) also show that sophisticated voters have stronger attitudes and are more susceptible to biased information-processing than unsophisticated voters. In addition to this, perhaps more surprisingly, they demonstrate that sophisticated voters are more strongly influenced by subliminal affective cues than unsophisticated ones. They explain this result by citing the richer knowledge structure of the sophisticates, which provides priming events with more opportunities to deposit prime-congruent evidence into the decision stream. Most affected are sophisticates who do not share a candidate’s views and who are negatively primed. All these findings underpin the early observation of Lazarsfeld, Berelson, and Gaudet (1944: 69) that the people who switch parties during an electoral campaign are not the reasonable, thoughtful and conscientious persons who were convinced by the issues of the election, but rather «the least interested in the election; the least concerned about its outcome; the least attentive to political material in the formal media of communication; the last to settle upon a vote decision; and the most likely to be persuaded, finally, by a personal contact, not an issue of the election».
The last set of moderating factors to be discussed refers to issue characteristics. While he generally observed hardly any indication of policy-voting, Lenz (2012) found that priming works when campaigns and news media place special emphasis on the economy: citizens give the economy greater weight in their judgment of a president’s performance, and in their decision to vote for or against the incumbent, when it is presented as evidence during a campaign (ibidem: 52). In other words, citizens do not so much change their opinion about the economy once the campaign focuses on the economy, but rather switch to the party from which they expect a better economic performance. Lenz (ibidem: 225) argues that «a growing body of evidence suggests that it is the economy and other performance domains, not ideology, that largely explain election outcomes». In discussing this exception, he puts forward certain possible explanations (ibidem: 226): First, and most importantly, he suggests that performance issues are cognitively easier for voters to digest. Policy issues require knowledge of politicians’ positions on the relevant policies, and require that voters hold views of their own concerning the policy or policies in question – conditions that are not often fulfilled. Second, politicians can more readily neutralize policy issues through a variety of strategies, including shifting to positions that are more popular. They are especially induced to do so when voters appear to develop strong views. Most of the time however, as Lenz notes, most voters lack strong views on policy, which gives politicians some leeway in their responses. Third, politicians might also attempt to blur their own position. They might be especially inclined to do so if they do not own the issue themselves and if the issue is a secondary matter within their own constituency (Rovny, 2013).
More generally, voters are more likely to hold strong policy views on issues that are familiar, which makes them less susceptible to priming effects. Note that those who already knew the respective party positions in Lenz’s (2009) study did not experience a priming effect, i.e. for them the issue did not become a more important factor in their voting considerations. In line with this point, Carsey and Layman (2006) found that voters for whom abortion was a salient issue tended to maintain their opinion on abortion and to switch to the party whose position was closer to theirs, while voters who cared less about abortion but followed politics closely enough to be aware of their party’s position on the issue tended to adopt the position of their party. Tesler and Zaller (2017) suggest that, for more familiar issues on which voters have more strongly held beliefs, they are more likely to change the party they support than change their opinion.
In a comparative analysis of three direct-democratic campaigns in Switzerland, it has been found that the impact of a campaign depends a lot on the familiarity and complexity of the issues that are submitted to the vote (Kriesi, 2012). The less familiar and the more complex the issue in question, the larger the share of undecided voters and of voters whose opinion is still malleable by the campaign. Two of the compared campaigns were concerned with familiar issues related to immigration – asylum law and naturalization law – while the third campaign related to an unfamiliar, highly technical issue – corporate tax reform. As it turned out, reinforcement and activation of political predispositions were much more prevalent in the familiar campaigns than in the unfamiliar one. Persuasion proved to be important in the case of corporate tax reform. While the overwhelming majority of voters in the immigration-related campaigns ended up making choices consistent with their predispositions, no less than one third of politically uninterested voters were persuaded by the corporate tax campaign to vote against their original predispositions. In this case, emotions played an important role: since these voters did not fully understand the arguments in favor or against the proposal, they relied on their feelings. Thus, feelings provide a heuristic tool both for voters who are not sufficiently motivated to process information, and for voters who are incapable of processing the complexity of the information required for making an informed choice. Positive emotions proved to be the strongest factor in determining the final voting outcome in the corporate tax campaign, stronger even than the prior vote intentions at the outset of the campaign.
The characteristics of an issue also play a key role in issue-specific debates and policy-making processes. In this regard, the salience of the issue is crucial. Only a limited number of issues are brought to the public’s attention, and on issues that are not publicly debated the public is likely to have no opinion at all. It is only if an issue becomes salient that the public will develop an opinion on it. To the extent that an issue becomes salient, it is also likely to become more familiar to the public and the public is likely to develop strong attitudes concerning the issue in question. Furthermore, when an issue becomes salient for the public, policy is likely to be responsive to the public. Thus, it is well known that policy tends to be more in line with public opinion on issues of high rather than low salience (Page and Shapiro, 1983; Page, 2002; Burstein, 2014).
In this process of issue-specific mobilization of public opinion, arguments and cues provided by the political elites and the media play a crucial role. The elite-driven «expansion of the scope of conflict» (Schattschneider, 1960) renders specific issues more salient. According to Schattschneider, one of the founders of the agenda-setting approach, the definition of issues is «the supreme instrument of power» (ibidem: 66) and the party that is able to make its definition of the issues prevail is likely to take over government (ibidem: 73). However, elite actors are likely to be constrained in their attempts to impose their views in the opinion-making process. First, they must compete with other elite actors. Framing effects are constrained by the competing frames promoted by other elite actors (Druckman, 2004). Early framing studies had overestimated possible framing effects because they had presented respondents with just a single frame. In addition, there is also competition from bottom-up mobilization efforts. Thus, recent contributions from the agenda-setting tradition emphasize the power of protest to signal discontent and raise the salience of certain issues in the elite’s debates (e.g. Vliegenthart et al., 2016). An analysis of Twitter messages by members of both the 113th American Congress and the public (Barberá et al., 2019) suggests that US legislators are responsive to changes in attention allocation of the attentive public, in particular to such changes among their own party supporters. More specifically, the controversial public debates that result from the expansion of conflict through protest mobilisation and tweets within the attentive public increase the legitimacy of speakers and allies of the movements in question, as well as tweeters from the public, in the eyes of journalists and decision-makers, who tend to closely follow these debates and exchanges (Gamson and Meyer, 1996: 288). Wolfsfeld’s (1997: 47) «principle of political resonance» formulates this relationship in a more concise way: challengers who succeed in producing events (and I would now add tweets) that resonate with the professional and political culture of important news media can compete with much more powerful adversaries in the agenda-setting process.
Increasingly, political science literature also pays attention to the social context in which citizens are embedded. Information does not only travel top-down and bottom-up, it also travels laterally, between peers in social groups and social networks. Thus, Druckman and Nelson (2003), Klar (2014) and Kertzer and Zeitzoff (2017) find that citizens’ conversations with one another can eliminate the effects of elite rhetoric. Studying different foreign policy issues in five experiments, Kertzer and Zeitzoff, to mention just one study, revealed an inconsistent effect from elite cues, but important effects from social cues (what «people like you» think) as well as from citizens’ political predispositions toward international affairs. While these studies suggest that people are likely to be more immune to elite manipulation than some of the more pessimistic observers suggest, they might not be entirely immune to social pressure, which might become more important in an age of social media.
Achen and Bartels (2016) draw on some of the literature I have tried to summarize here in order to debunk what they call the «folk theory of democracy» – the theory that celebrates the wisdom of popular judgments by informed and engaged citizens. Their more «realistic theory of democracy» leaves more scope for political leadership and makes fewer demands of citizens’ political knowledge and engagement. It draws on the twin idea that politics is essentially based on groups and that group identity is the key determinant of voting decisions. In thinking about politics, they argue, it makes no sense to start from issue positions. These positions are generally derivative of something else, and this something else is group identity, with the most important identities in politics being party identities. Voters choose a party based on their social and political identity (we have been told since childhood that a certain party represents «people like us») and then rationalize their choice with appropriate party-supplied reasons. The specific circumstances of a given election may occasionally inspire significant changes in voting decisions – especially circumstances related to economic downturns – but even these defections often reflect the salience of some specific identity.
This interpretation of how democracy works is not very appealing, but it iscompatible with the recent evidence concerning political communication surveyed in this short essay: in the «tribal» world of the realistic theory of democracy, voters protect their identities through motivated reasoning. In this world, there is little room for persuasion in terms of voting choice. Political communication serves to reinforce and activate party identifications, especially in polarized situations. Citizens are more open to arguments under less polarized circumstances and when they encounter unfamiliar issues or new facts, i.e. when they have something to learn. But even when faced with unfamiliar issues or new facts, they are influenced by party cues if the arguments do not clearly point in one direction or if the facts prove to be inconvenient because they clash with their preconceived views. Citizens’ identities put clear limits on the possible effects of political communication, since they are hard to change. However, the other side of the coin is that there is considerable room for persuasion in terms of issue positions. The upshot of the studies discussed here is that, except for opinions concerning economic performance, citizens adapt their issue-specific opinions to their party (or other) identities. They tend to believe what their identity tells them is right. Guided by their identities, citizens tend to resort to elites they trust and process information in a biased way in order to protect themselves. This opens the door for the «manufacturing of public will» within the confines of specific group identities.