Shaping anti-immigration attitudes:
unravelling the dynamic between emotion and cognition
Master thesis Political Science, Political EconomyUniversity of Amsterdam Emiel Lijbrink (11011777)
June, 2017
Supervisor: Dhr. dr. G. Schumacher Second reader: Dhr. prof. dr. W. van der Brug
2 Table of contents Abstract ... 3 Introduction ... 3 Theory ... 6 Conceptualizing emotions ... 7
How to measure emotion? ... 10
Affective intelligence and political behaviour ... 10
The distinct effects of anxiety and anger ... 13
Framing effect and emotion ... 15
Motivated reasoning: affective tags and attitude polarization ... 17
Immigration: threats and emotions ... 19
Data & Methods ... 22
Results ... 29
Conclusions & Discussion ... 33
References ... 38
Appendix ... 44
3 Abstract
Earlier studies in political science examined the effects of framing, motivated reasoning and affected intelligence theory separately. This thesis integrates these relevant theories to investigate how anti-immigration attitudes are shaped by both the cognitive and emotional effects of anti-immigration frames. In a between group survey experiment (total N=329), I have shown that anti-immigration frames elicit significantly higher levels of anxiety, fear and anger compared to a balanced frame. The negative frame did not, as hypothesised, result in more negative attitudes towards immigration. Moreover, this thesis finds no confirmation for attitude polarisation, as predicted by motivated reasoning theory. I theorised - in line with the affective intelligence theory – that anxiety should increase anti-immigration attitudes. In contrast, the experience of anger and fear is not expected to boost attitude change. My results indicate that anxiety and anger do not stimulate attitude change, while fear significantly contributes to more negative attitudes towards immigration, when controlled for anxiety. The implications of these findings are discussed in this thesis.
Introduction
The influence of emotion over human reasoning has captured the interest of political thinkers for centuries. Aristotle, for example, stated that emotion - in his words ‘passion’ - could potentially distort the political judgement of citizens (Korsgaard, 1986). Emotion was portrayed as an undesirable distraction, meant to be consciously controlled by the rationale of individuals. The field of political psychology, which investigates the effects of emotion on political behaviour, is still relatively young. In the past, the field of political science was mainly concerned with cognitive issues (Marcus, Neuman, & MacKuen, 2000). Citizens were considered as rational individuals that took decisions based on the critical evaluation of information. One of the most famous scholars to adopt this view was Anthony Downs, who developed a rational voter theory which described political participation in terms of benefits and costs (Downs, 1957). In brief, the theory illustrates that citizens invest more time and energy in decision making and are more likely to vote, when they believe their vote to hold significant influence over the outcome of an election. The likelihood of voting decreases when citizens expect that their influence will not outweigh the costs of voting, often the case during large elections. This rational choice perspective can also be applied to political information processing, including the formation of attitudes, which forms the focus of this study. For
4
example, when confronted with political information, citizens will process this thoroughly and alter their attitudes accordingly solely when they can conceive the information as being pivotal for future political decisions
In Down’s work, and the work of many other rational choice scholars, the role of emotion was neglected or portrayed as the opposite of cognition. Nevertheless, research in recent years has shown that emotions are strong predictors of political participation and engagement and should therefore be included in models that intend to explain political behaviour more comprehensively. Research in this field has focused on the effects of emotions on voter turnout (Brader, 2005), other forms of political participation (Weber, 2013), the evaluation of political candidates (Chang, 2001), political interest (Huddy & Mason, 2008) and issue attitudes (Brader et al.,2008). Some scholars even suggest that emotion can contribute to democracy by solving collective action problems (Groenendyk, 2011). Such problems are related to voter turnout and other forms of political participation. Several empirical studies show how specific emotions, elicited by political messages, can counter this problem by stimulating voters to get off the couch and engage in political life (Groenendyk & Banks, 2014; Valentino, Brader, Groenendyk, Gregorowicz, & Hutchings, 2011; Weber, 2013).
The affective intelligence theory of Marcus, Neuman & MacKuen (2000), which emphasizes the positive role of emotions, has raised the interest of many contemporary scholars. The title of this theory is somewhat provocative, because it suggests that emotions are not undesirable, but a vital component of political life and, specifically, political judgement.Marcus et al. (2000) show that political decision making and, more generally, attention paid to political affairs are guided by two brain systems: the disposition system and the surveillance system. The disposition system is activated when citizens experience enthusiasm, resulting in a reliance on predispositions formed by previous experiences. The surveillance system is governed by anxiety and heightens the attention of citizens, making them more open to new information and thereby more inclined to adjust their position. More recent research indicates that the effects of anxiety, anger and enthusiasm are deliberately used by politicians in their election campaigns to strengthen the support of their loyal electorate and persuade undecided voters (Ridout & Searles, 2011).
In their initial study, Marcus et al. (2000) solely make a distinction between negative and positive emotions and hence expect the same effects for the negative emotions of anxiety, fear and anger. Later (political) psychology research suggests that these emotions have different
5
effects on political behaviour. While anxiety is expected to boost critical evaluation of new information and attitude change to counter future threats, the experience of fear is associated with fleeing behaviour (aversion) (McNaughton & Corr, 2004; Öhman, 2008) and anger with the reliance on habits and less critical evaluation of new information (Huddy, Feldman, & Cassese, 2007; MacKuen, Wolak, Keele, & Marcus, 2010).
The work of Marcus et al. (2000) has stimulated many scholars to examine the dynamic between cognition and emotion. For example, a handful of scholars in the field of framing studies, which were – and still mainly are – concerned with the role of cognition, have decided to incorporate the role of emotion into their work. Several new framing studies offered support for the hypothesis that emotions can mediate the impact of framing effects on citizens attitudes (Lecheler, Bos, & Vliegenthart, 2015; Lecheler, Schuck, & De Vreese, 2013). Framing studies strongly focus on the persuasiveness of political messages and suggest that views of citizens can be altered by presenting them with a message conveyed at the right angle.
A field of study which partly challenges this assumption - but also compliments framing theory- is the field of motivated reasoning theories. Empirical work, testing this theory, found that citizens do not process new information unbiasedly. Indeed, individuals prefer to search for information that reaffirms their opinions (Valentino, Banks, Hutchings, & Davis, 2009). All political concepts, stored in individuals’ long-term memory, are affectively loaded. These affective tags are activated within milliseconds of individuals being exposed to related political stimuli (Lodge & Taber, 2005). Subsequently, they accept arguments that are congruent with their views without much consideration, while fiercely counterarguing incongruent information. This results in the strengthening of prior attitudes, which Taber & Lodge (2006) characterize as ‘attitude polarization’.
Despite the fact that the theories of affective intelligence, motivated reasoning and framing theory partly challenge each other’s assumptions, they all attempt to solve a similar puzzle: the interaction between emotion and prior views in processing political information. Yet very few studies, if any, combine these three theories. The goal of this article is to shed light on how these three theories tie into one another and, when combined, explain how anti-immigration attitudes of citizens are shaped by the interaction between predispositions and emotion.
6
In this study, a survey experiment will be conducted with two experimental groups, each confronted with a different message on immigration: a negative message or a balanced message. The topic of immigration is chosen because it has elicited a great deal of public attention in recent years and is often characterized in terms of a threat to society, the economy and national identity by anti-immigration parties (Hainmueller & Hopkins, 2014). Framing the issue in such a way can potentially elicit strong negative emotions, such as anxiety and anger (Brader, Valentino, & Suhay, 2008), which makes this a suitable topic for this investigation.
Multiple linear regression analyses and t-tests are conducted to test whether a negative message concerning immigration elicits higher levels of negative emotions than a balanced message (H1); if a negative message on immigration has distinct attitudinal effects on citizens with positive and negative views towards immigration (H2 & H3) and if the experience of anxiety contributes to higher levels of anti-immigration attitudes, in contrast to fear and anger (H4).
In support of framing theory, this study finds that citizens exposed to the negative message about immigration experienced significantly higher levels of negative emotions. My findings do not indicate a different attitudinal effect for respondents with positive and negative attitudes, as predicted by motivated reasoning studies. Moreover, fear was the only significant predictor of increased anti-immigration attitudes, but solely when controlled for anxiety. Furthermore, my research provides evidence for the distinctive role of anger. The implications of these findings are discussed in the conclusions & discussion section of this study. First, the underlying theories and concepts are presented in the theory section, followed by a description of the methods, data and results, upon which these conclusions are based.
Theory
The role of emotions has been intensively debated by political thinkers in the past. Classic thinkers such as Plato, Aristotle and Hobbes all found it necessary to understand the impact of emotions on political life. Emotions and passion were long considered detrimental to our use of reason. Plato held perhaps the most extreme views in this regard, claiming that emotions – labelled as passion – can hinder the ability to reason, making it necessary to exclude them completely from political life. Those citizens trained solely to serve as philosopher kings were,
7
according to Plato, capable of doing this, and were able to rule wisely without impediment to their capabilities to reason (Marcus, 2003).
Although the role of emotion was fiercely debated by political thinkers in the past, the topic has only recently raised the interest of scholars in political science. This can partly be explained by the complexity of conducting research in political psychology. Similarly, the marginal role of emotions in political science may also be related to the dominant role of rationality, emphasized by scholars such as Anthony Downs (1957). In his famous rational voter model, Downs claimed that the behaviour of voters could largely be explained by rationality and the weighted costs and benefits of political participation.
Later research has shown that political behaviour is much more complex and cannot solely be explained by rationality (Ragsdale, 1991). In recent years, scholars have increasingly emphasized that emotions should be included in voter models that attempt to explain political participation (Groenendyk, 2011); Groenendyk & Banks, 2014), information processing (Morris, Squires, Taber, & Lodge, 2003) and the formation of attitudes (Brader, 2005). The last two of which are investigated in this study.
Conceptualizing emotions
Before I describe the most important theories that form the basis of this study and reflect on the anti-immigration literature, I will first briefly conceptualize emotions, specifically anxiety, fear and anger. These three represent the core emotions examined in this theoretical framework. Even though emotion is an intensively studied concept within the field of political psychology, there is no clear consensus about a general definition of emotion. Emotions are conceptualised by scholars as feelings, affect, brain modes, etc. The reason for this is that emotion is multifaceted and not a unitary phenomenon or process (Izard, 2009). Although the concept of emotion is far from uncontested, scholars agree that emotions have a neurobiological basis, with a system dedicated to their processing, and that emotions ‘motivate cognition, action and recruit response systems’ (Izard, 2009, p. 7). Consensus has also been reached over two different forms of emotions: complex emotions, that demand cognitive components, and basic emotions, which are based on our evolution and challenges for survival (Izard, 2007). Basic positive emotions are joy and interest, while basic negative emotions include sadness, anger, disgust and fear. The neurobiological processes that let us experience emotion also create
8
feelings (emotional sensations). Izard (2009) defines these feelings as a ‘phase of neurobiological activity that is experienced as motivational and informational and that influences thought and action, a felt cognition, or action tendency’ (p. 3). This definition is particularly interesting for this study since subjective feelings are associated with a certain level of consciousness. This enables individuals to record their experience when taking part in an experiment, although with some consideration for the subtlety of certain feelings and their difficulty to be recognised. Feelings are especially useful for helping individuals to organize and simplify their impulses (Izard, 2009). It stimulates people to develop feeling-cognition-action patterns, which are helpful for identifying similar impulses in the future (Izard, 2009). This also makes them relevant for studies in the field of political psychology that examine the interaction between emotion and cognition, and do not treat these components as two completely different phenomena.
In this study, the effects of the negative emotions, namely anxiety, fear and anger will be examined. Fear and anxiety are closely related. Both emotions are experienced in the presence of a threat and lead to negative feelings and heavy emotional sensations (McNaughton & Corr, 2004). Although some scholars treat fear and anxiety as similar and find strong correlations between these emotions, the field of psychology describes several relevant differences. Fear is associated with an impending disaster and stimulates citizens to defend themselves, usually by fleeing. In contrast, anxiety is defined as an unpleasant and ineffable feeling of foreboding (Marks & Lader, 1973). Öhman (2008) describes the differences between fear and anxiety as follows: fear is experienced when individuals are confronted with a direct identifiable threat (stimulus), while anxiety is often (experienced) ‘pre-stimulus’ and is an anticipatory reaction. These differences make it irrational for anxious individuals to escape from a threat from which the location and nature are poorly defined (Öhman, 2008). This also suggests that the experience of fear is related to automatic behavioural responses, while anxiety should push for a more thorough evaluation of a threatening situation. This is in line with the two-dimensional neuropsychology model of defense by McNaughton & Corr (2004). Their model describes two different brain systems that can be activated when individuals are confronted with a threat: the fight-flee-freeze system (FFFS) and the behavioral inhibition system (BIS). The FFFS is controlled by fear and leads to direct aversion because it is deemed safer to simply avoid a critical threat. The BIS is activated when stimuli elicit anxiety, which boosts attention and is related to approach behaviour to counter a potential threat. The latter system requires more cognitive resources compared to the FFFS (McNaughton & Corr, 2004).
9
Although the field of psychology describes significant differences between fear and anxiety, political psychology studies often combine fear and anxiety in one negative dimension (Marcus et al., 2000) and therefore expect them to have the same effects on political behaviour. In this study, these emotions will be measured separately to investigate whether the effects differ – as predicted by the psychology literature.
The effects of anger on the formation of attitudes are also examined in this thesis since anger has distinct effects on information processing and the formation of anti-immigration attitudes compared to fear and anxiety. Moreover, the characteristics of anger are somewhat different. In an article that tries to untangle the differences between anger and hostility, Barefoot & Lipkus (1994) use a very broad and multifaceted definition of anger:
‘Anger refers to a label given to a constellation of specific uncomfortable subjective experiences and associated cognitions (i.e., thoughts, beliefs, images, etc.) that have various associated verbal, facial, bodily, and autonomic reactions. It is a transient state, in that it eventually passes, and it is a social role, in that our culture or subculture allows for the display of certain kinds of behaviors associated with the internal experience, but punishes others. Thus, anger is felt in people’s conscious awareness and is communicated through verbalizations and bodily reactions.’ (p. 20)
Additionally, Lemerise & Dodge (2008) state in The Handbook of Emotions that anger is high-arousal emotion which is related to self-defence and helps individuals to overcome obstacles to reach a certain goal. Later in this theoretical framework, the different effects of anger and anxiety will be described and compared more extensively in a political context. In sum, the psychology literature suggests – in contrast to political science research – that the experience of fear and anxiety will result in different behaviour and can have opposite effects on the formation of anti-immigration attitudes. While anxiety is related to approaching a source of threat and the thorough evaluation of new information (leading to attitude change), fear is associated with the immediate aversion of a threat and should therefore not result in attitude change.
10 How to measure emotion?
A methodological problem that is apparent in the field of political psychology is the measurement of emotion. Barrett (2006) concludes, after a review of the empirical literature on emotion methodology, that it is ‘difficult, if not impossible, to find an objective means of measuring the experience of emotion’ (p. 22). The two most employed methods to measure emotions are psychological reactivity and self-report. The former includes skin conductance, facial movements and heart-rate monitoring (Renshon, Lee, & Tingley, 2015). Even though researchers can measure if participants experience positive or negative emotions with these methods, they are not structurally able to distinguish the same psychological patterns linked to discrete emotions such as anger and fear (Cacioppo, Berntson, Larsen, Poehlmann, & Ito, 2000). Other methodological problems arise when participants are asked to directly report any emotions they experience themselves. While some individuals can report their emotional experiences even in discrete categories such as anger, sadness and fear, others are only capable of reporting emotions in broad terms of pleasure and displeasure (Barrett, 2006). The fact that everyone believes that emotions are natural-kind entities, but that research has not produced methods and results that strongly support this claim, leads Barret (2006) to the conclusion that we face a fundamental paradox. He therefore suggests that we may have to accept that emotions are not biologically given, but are constructs that are created via the process of categorization.
Affective intelligence and political behaviour
One of the most influential theories concerning the role of emotions in political decision making is the affective intelligence theory (AIT) of Marcus, Neuman & MacKuen (2000). This theory rejects the classical dichotomy between cognition and emotion. In fact, Marcus et al. (2000) state that these processes need to be considered as complementary. Political behaviour is, according to the AIT, guided by two brain systems: the disposition and surveillance system. These systems are triggered by different emotional responses which lead to different decision making strategies. The disposition system is located at the limbic region of the brain, which manages previously learned behaviour. This system is governed by enthusiasm. When citizens experience positive emotion in a political context, for example when confronted with information that concurs with their views, they will rely on their routines and predispositions. It becomes more efficient for them to maintain their habits and make swift comparisons, rather
11
than evaluating the new information thoroughly and repeatedly. Decisions that led to success in the past will most likely lead to success in the future as well. The disposition system obtains four types of information: information about the body, information about the environment and plans/procedures. When citizens combine these types of information they are able to compare information rapidly and make decision based on their heuristics. Minor changes in circumstances will not alter their habits completely, although they will slightly adapt their habits to prepare for similar situations in the future. The experience of enthusiasm works in this system as a piece of information (Schwarz, 1990), which forms the foundation for basic judgements. Schwarz (1990) concludes that these ‘affective decisions’ will most likely occur under the following four conditions:
‘‘(1) When the judgment at hand is affective in nature (2) when little other information is available; (3 ) when the judgment is overly complex , and cumbersome to make on the basis of a piecemeal information -processing strategy ; and (4 ) when time constraints or competing task demands limit the cognitive capacity that may be devoted to form in g a judgment.’’ (Schwarz, 1990) (p. 538)
On the other hand, the surveillance system is activated when citizens are confronted with new, unexpected situations. This emotional system, which is governed by anxiety, enables citizens to counter potential threats by making them more attentive and open to new information. Explicit consideration of (new) information is vital to manage uncertain conditions, and the surveillance system can therefore be considered as an evolutionary way of survival. The uniqueness of the affective intelligence theory (AIT) lies in the combination of these systems. Unconscious appraisals of the (political) environment determine which of the two brain systems are activated, and if citizens can rely on their habits or must be attentive and possibly reconsider their preconceived ideas.
The implications of the AIT have been empirically tested by Marcus et al. (2000) and many others in both experimental and natural settings (Brader, 2005; Gadarian & Albertson, 2014; Huddy et al., 2007; Huddy, Feldman, Taber, & Lahav, 2005; Jonathan Mcdonald Ladd & Lenz, 2011; Lecheler et al., 2013). The empirical work of Marcus et al. (2000) is based on data from the 1980-1996 American National Election Studies (ANES) and two specific experiments. They conclude that anxiety, rather than enthusiasm, increases overall political attentiveness, which results in higher political participation and engagement of citizens during election campaigns. Moreover, the experience of anxiety decreases the reliance on partisanship.
12
An interesting addition to the studies that test the effects of AIT on political behaviour is the study of Ridout & Searles (2011). They examine the use of emotional appeals during an election campaign and conclude that political candidates use emotions strategically in their communications. Leading candidates use emotions related to the disposition system (pride and enthusiasm) to strengthen the predispositions of their structural electorate, while trailing candidates focus on emotion appeals that are associated with the surveillance system – mainly fear - to stimulate reconsideration of prior preferences and information seeking (Ridout & Searles, 2011).
Although many studies find support for the AIT and the unique roll of anxiety, it is far from uncontested. When replicating the study, Ladd & Lenz (2008) found little to support the role of anxiety in decreasing the role of predispositions. Instead, their findings indicate that emotions directly influence candidate evaluations and that candidate evaluations are strong predictors of emotions. They therefore conclude that their findings are more in line with two simpler theories: The Affect Transfer theory and the Endogenous Affect theory. The publication of this somewhat ‘provocative’ paper resulted in a critical reaction from Brader (2011) and Marcus, MacKuen, & Neuman (2011). According to Brader (2011), Ladd & Lenz (2008) based their conclusions on a very narrow view of the available literature and evidence, ignoring several experimental studies that support the AIT. Moreover, Marcus et al. (2011) point to the fact that the operationalisation of affect and the dependent variable are different in the study of Ladd & Lenz (2008), which led to different results.
Ladd & Lenz (2008) are not the only scholars who questioned the claims of the AIT and, more specifically, the unique role of anxiety. Other scholars have found that anger and enthusiasm also enhance political attentiveness and stimulate citizens to remember campaign information (Civettini & Redlawsk, 2009). Furthermore, anger is defined by scholars as a ‘mobilizing emotion’, meaning that it should be related to higher levels of political participation. By conducting two experiments, Weber (2013) examined the effects of anger, fear, sadness and enthusiasm on political participation. Weber concluded that anger is the only emotion that systemically boosts different types of political participation. These findings are supported by a study from Valentino, Brader, Groenendyk, Gregorowicz, & Hutchings (2011). They employ three experiments and conclude that anger boosts different forms of participation such as voting, attending a political rally and contacting a political party, while anxiety only increases less time-consuming forms of political participation.
13 The distinct effects of anxiety and anger
The behavioural differences between anger and anxiety as well as their different aspects are described and explained by cognitive appraisal theories (Smith & Ellsworth, 1985). Anger is triggered when individuals feel they have control over a certain situation and can result therefore in risk-seeking and active political behaviour. Meanwhile, anxiety is associated with uncertainty and the loss of control over a situation, which leads to (risk-)aversion (Huddy et al., 2007). In their initial study, Marcus et al. (2000) do not measure the effects of anger and anxiety separately, instead opting for a two-dimensional valence model which combines these emotions. They therefore implicitly assume that these emotions lead to similar effects on political behaviour. In a later study, Marcus (2003) acknowledges the distinct effects of anger and anxiety and concludes, based on psychology research, that the experience of anger might – in contrast to anxiety - strengthen an individual’s reliance on previously held views. In a later paper, MacKuen et al. (2010) introduce an aversion dimension, which builds on the previously mentioned disposition system, further describing the distinct effects of anger. When citizens are confronted with a familiar threat, they might experience anger, disgust, contempt or hatred. Because the threat is well-known, individuals choose not to evaluate new information, instead attempting to ignore the situation and stand by their pre-existing habits and routines. MacKuen et al. (2010) investigate the aversion dimension by empirically testing whether citizens rely on partisanship or engage in open-minded deliberation. As hypothesized, citizens that experienced anger were inclined toward reliance on partisanship, while anxious citizens engaged in greater deliberation.
These findings are supported by a literature review of Huddy et al. (2007) on the chosen emotions. They conclude that anger and anxiety have distinct effects on risk assessment and action, political motivation and depth of cognitive processes. The latter is most relevant for this thesis because it is related to political judgement and the formation of attitudes. Anger leads to less thorough cognitive processes and therefore rapid decision making (Huddy et al., 2007), while anxiety is associated with increased attentiveness and boosts political learning and information seeking (Marcus et al., 2000). These characteristics are explained by deeper levels of cognitive processing that are activated when citizens experience the given emotion. This is more extensively described by a study of Larissa & Tiedens (2001), who investigated the effects of emotional certainty and uncertainty on judgement. Certain emotions – such as anger, disgust, happiness and contentment – are associated with certainty of judgement, while other emotions
14
(e.g. hope, surprise, anxiety and worry) are related to uncertainty of judgement. By conducting three experiments, they find that the experience of uncertainty-related emotions boosts systematic information processing, while certainty-related emotions lead to more heuristic processing (Larissa & Tiedens, 2001). Moreover, the depth of information processing associated with emotions can have an impact on how citizens determine their vote during an election campaign. Consistent with the aforementioned authors, Parker & Isbell (2010) demonstrate that the experience of anxiety predicates reliance on detailed, issue-agreement information, as a basis for voting decisions, while citizens that experience anger are more inclined to base their vote on general criteria. Although these studies suggest that anger leads to less critical evaluation of information, Moons & Mackie (2007) partly challenge these conclusions and find in an experimental study that ‘angry people have both the capacity and motivation to process and that their selective use of heuristics reflects the cue’s perceived validity and not the failure to process analytically’ (p. 706).
Despite the body of research suggesting that anxiety can enhance deep information processing, this does not necessarily mean that the experience of these emotions leads to unbiased information processing (Gadarian & Albertson, 2014). During an experiment, the previously cited authors confronted citizens with anxiety-loaded messages and gave them the opportunity to search for additional information on a website afterwards. News stories on this website were divided by threatening and non-threatening coverage. Citizens that reported anxiety had more often read, remembered and agreed with threatening news stories on the website. This suggests that anxious citizens might be biased information processors. This is also an important conclusion of Valentino et al. (2009), who state that anxiety does not simply boost information seeking: citizens will search for information that is useful to them, either to counter a certain threat or strengthen their own views when they are confronted with information that questions their predispositions. The latter can be related to motivated reasoning and cognitive dissonance, theories which will be further discussed later in this section.
In sum, the affective intelligence theory claims that anxiety can boost deep cognitive processes, which enhances political attention, openness to new information and attitude change. In contrast, anger is associated with less complex cognitive processes, which stimulate the reliance on predispositions and risk seeking. Although anxiety can boost critical evaluation of new information, several studies suggest that citizens are not able to completely ignore their predispositions and can therefore be classified as biased information processors.
15 Framing effect and emotion
While the affective intelligence theory combines cognition and emotion to explain political judgement, framing studies are traditionally solely concerned with the role of cognition. Yet, a handful of framing studies acknowledges that emotions – and specifically anxiety - can directly or indirectly influence the formation of attitudes. The basis of framing is the assumption that the attitudes of citizens can be altered by presenting them with a story set within a specific framework, where certain information is emphasized and other information ignored. Or as Matthes (2012) defines the act of framing:
‘To frame is to select some aspects of a perceived reality and make them more salient in a communicating context, in such a way as to promote a particular problem definition, causal interpretation, moral evaluation, and/or treatment recommendation for the item described’ (p. 249).
Scheufele (1999) defines several routes to classify frames and framing studies. Firstly, frames can be defined as media frames or individual frames. Media frames are created by journalists or other media actors and represent a certain storyline or presentation of daily events. To underline or address a problem, media actors emphasize certain aspects of the story over others. Secondly, frames can be considered as individual. These frames describe ‘how individuals make sense of political news’ mentally and process this information to form their political attitudes (Sheufele, 1999). A different way to classify framing studies is between research that examines frames as an independent variable versus frames as dependent variables. In studies that examine the former, frames are used to predict the effect on individual or group behaviour (e.g. on voting, candidate preferences or issue attitudes). In studies where a frame is used as a dependent variable, the factors that influence the frame exerted by journalists or other media professionals are analysed (Scheufele, 1999).
Another interesting distinction, relevant to this study, is made between equivalency framing effects and emphasis framing effects (Druckman, 2001). Equivalency framing effects are related to frames that consist of different but comparable words or phrases, used to influence individuals’ interpretation of events, issues or candidates. For example: ‘this product is 98% sugar-free’ over ‘includes 2% sugar’. Within this type of framing, the same exact information is presented, but in a positive or negative light. The second type of framing, defined by Druckman (2001), is an emphasis framing effect. Emphasis frames do not intend to present the
16
same information, but instead focus on certain details or aspects of an event to alter the attitudes of individuals. (e.g. a politician that focuses on the humanitarian responsibility of countries in the context of the refugee crisis versus a politician that solely describes the economic disadvantages of immigration). This thesis will mainly focus on emphasis frames as an independent variable that is regularly used and created by politicians and other elites. An elaboration on the frames used in this study will follow in the methodology section.
Despite the fact that researchers in this field seem to agree that minor changes in political frames can influence the way in which political events are interpreted (Juan Joselartya Igartua & Cheng, 2009), the underlying system concerning how frames influence attitudes remains largely unexplained. Petty and Cacioppo (1986) aim to fix this gap in the literature by prescribing two systems that can be activated when citizens are confronted with a certain frame: the central route and the peripheral route. The central route is strongly related to the idea of the rational voter, who thoroughly evaluates available information and compares this information with their prior knowledge. The peripheral route, also named ‘heuristic processing’, leads to automatic, less thoughtful processing. Citizens that are not motivated or able to process information extensively will often choose this route and will rely on salient cues.
Where the lion’s share of studies focuses on the cognitive effects of frames (e.g. how do frames influence the formation of attitudes), frames are also capable of eliciting emotions (Gross, 2008; Gross & D’Ambrosio, 2004), which can – according to the affective intelligence theory – influence cognitive processes, attitudes and political judgement (Marcus et al.,2000). As a consequence, a handful of framing studies combine cognition and emotion by examining emotions as mediators of framing effects on attitudes (Bos, Lecheler, Mewafi, & Vliegenthart, 2016; Juan José Igartua, Moral-Toranzo, & Fernández, 2011; Lecheler et al., 2015, 2013). This effect is not limited to negative emotions or specifically anxiety. For instance, Lecheler et al. (2013) find that anger and enthusiasm mediate news framing effects. An important point to emphasize here is that the mediating effect of emotions is highly dependent on the specifics of a certain frame, as well as issue selection. Even a frame about the consequences of immigration can potentially elicit positive emotions such as hope or compassion, which can function as mediators on attitudes (Lecheler et al., 2015).
17
In brief, framing studies indicate that frames used by the media and politicians can shape the way citizens interpret information and may elicit emotions that strengthen attitude change. From here I draw the following hypothesis:
H1: A negative message on the consequences of immigration will elicit more negative emotions compared to a balanced message.
Motivated reasoning: affective tags and attitude polarization
Although framing theory suggests that the views of citizens can be altered by presenting them information with a certain interpretation or angle, citizens do not accept frames mindlessly. Previously formed opinions undoubtedly play a strong role in the effectiveness of frames. In many situations, citizens will rely on their predispositions – as described by the affective intelligence theory - and are therefore not able to process new information without bias. The AIT builds upon the classic ‘hot cognition hypotheses’, which state that all previously evaluated socio-political information is affectively charged, and that these affective tags are activated from the long-term memory within milliseconds of citizens being confronted with political stimuli (Abelson, 1963). Two other scholars that use this as their starting point are Taber & Lodge (2005 & 2013). In their research – which is based on motivated reasoning theory - they conclude that hot cognition results in biased information processing, because balanced reasoning is prevented by automatic and implicit evaluations that are based on pre-existing knowledge and positions. In their later work, Taber & Lodge (2006) proposed and tested a model named ‘motivated scepticism’, attempting to further explain the reason why prior attitudes strongly influence information processing. Through two experimental studies, they show that citizens spend time and cognitive resources to fiercely counterargue information that is not in line with their views, while attitudinally congruent arguments are accepted without much consideration. An interesting element of this study is that of attitude polarization. They find that citizens who hold strong (positive or negative) attitudes on gun control will strengthen their attitudes after they read favourable or unfavourable arguments on the topic. The effect of prior attitudes is stronger for politically sophisticated citizens compared to less politically sophisticated citizens, since the former group already hold strong views as a result of thorough evaluation of both sides of the argument (Taber, Cann, & Kucsova, 2009; Taber & Lodge,
18
2006). These findings are supported by the earlier research of Taber, Cann, & Kucsova (2009) who specifically investigate the disconfirmation bias, concluding that citizens are unable ‘to escape the pull of their prior beliefs when considering policy arguments’ (p. 153). A similar case can be observed when citizens are confronted with information that attacks their favoured political candidate. This incongruent information stimulates attitude polarisation and therefore strengthens support for their candidate (Redlawsk (2002).
Studies that examine the effects of motivated reasoning are often partly in agreement with cognitive appraisal theories (Kemper & Lazarus, 1992). These theories posit that emotional reactions caused by certain events are not random, but depend on the available information, beliefs, predispositions and the goals individuals possess. Hence, individuals that are confronted with the same information will evaluate this differently and experience different types, and levels, of emotion.
The theory of motivated reasoning also shows a clear link to cognitive dissonance theory (Festinger, 1957). The basis of this theory is that individuals strive for a cognitive equilibrium (Westerwick & Meng, 2009) by dissolving conflicting information. Knobloch-Westerwick & Meng (2009) found that media users select and read attitude congruent messages and avoid information that challenges their beliefs. Stronger beliefs and habitual news observation resulted in stronger cognitive dissonance effects. Moreover, Donsbach (1991) & Stroud (2008) find that attitudes on several issues are strong predictors of their media selection (newspaper, radio, and internet news selection).
In sum, the literature on motivated reasoning suggests that the effects of negative political stimuli on citizens’ attitudes are dependent on their predispositions. Prior evaluations and experienced emotions largely determine how citizens process new information. Thus, a negative stimulus placed on immigration can have two different effects, seeing that citizens will engage in attitude polarisation. This leads to the following hypotheses:
H2: Citizens with negative attitudes on immigration will strengthen their current attitudes after reading an anti-immigration message
H3: Citizens with positive attitudes on immigration will strengthen their current attitudes after reading an anti-immigration message
19
Due to the fact that citizens are motivated reasoners and are inclined to exclusively process, seek, read and accept information which is consistent with their preferences, it is very difficult for elites to persuade citizens with information that is inconsistent with said preferences (Ditto & Lopez, 1992). Because individuals evaluate incongruent information very sceptically, politicians need to distribute high quality information (frames) to persuade individuals of an alternative conclusion (Ditto & Lopez, 1992).
Furthermore, the literature on motivated reasoning and framing both challenge the normative view of democracy, in which competent citizens are able to take political decisions based on a critical evaluation of new information. They question the normative competence of citizens in two ways. Firstly, individual attitudes should not be determined by the framework in which political information is presented, but by the actual interests of citizens (Druckman, 2001). Secondly, citizens should be capable of evaluating information that conflicts with their prior views and adjust them accordingly. The latter point – highlighted by motivated reasoning theory – could potentially be countered by the role of emotions. When we follow the reasoning of the affective intelligence theory, the experience of anxiety decreases the effects of motivated reasoning, making citizens more open to new information and therefore more inclined to adjust their attitudes. This effect could potentially be abused by elites who persuade citizens to adjust their attitudes and vote against their interest. The normative implications of these theories will be further discussed in the conclusions & discussion section of this thesis.
Immigration: threats and emotions
Many framing studies – including this thesis – specifically examine the effects of negative emotions in political messages on immigration (Bos et al., 2016; Fitzgerald, Curtis, & Corliss, 2012). The reason for this is that immigration has elicited a public debate wherein emotional language is often used to describe why immigrants form a threat. This makes it a very suitable issue to analyse the effects of emotions on the formation of attitudes.
Immigration – especially from non-western countries – has gained a great deal of media attention in recent years. Immigrants are, as one example, framed as ‘enemies at the gates of Europe’ and terrorists entering Europe disguised as refugees (Esses, Medianu, & Lawson, 2013). This is fostered by the rise of populist and anti-immigration parties in Europe that regularly voice anti-immigrant sentiments. Although many studies show that negative news
20
frames on immigration are significant predictors of anti-immigration attitudes (Juan José Igartua et al., 2011; Juan Joselartya Igartua & Cheng, 2009), this does not necessarily mean that media attention generally focused on immigrants always leads to stronger anti-immigration attitudes. In fact, Boomgaarden & Vliegenthart (2009) demonstrate in their study that the increased media visibility of immigrants between 1993 and 2005 is related to lesser anti-immigrant views. This is explained by the ‘extended contact theory’, which illustrates that citizens who come into contact with immigrant or other minority groups via the media are more inclined tosympathize with these groups (Boomgaarden & Vliegenthart, 2009). The same study finds strong support for the influence of news media evaluations of immigrants on anti-immigration attitudes. Negative coverage boosted anti-immigrant attitudes and the reverse holds for positive coverage, when controlled for national immigration levels.
As a consequence of immigration, several questions arise: What are the costs of immigration? How many immigrants can our county handle? Will we lose our country due to immigration? Most studies that focus on anti-immigration attitudes distinguish two types of predictors: economic threats (e.g. the loss of jobs, a wage decrease, a decrease in economic growth or a more expensive welfare system) and cultural threats (the loss of national culture or (European) values). The question is: which of these factors is the strongest predictor of a perceived immigration threat? Several studies test this by comparing the effects of the factors mentioned earlier. In an extensive literature review where twenty years of research is analysed, Hainmueller & Hopkins (2014) conclude that citizens’ personal economic situation is a poor predictor of anti-immigration attitudes. Cultural or symbolic concerns bear a much stronger relation to ant-immigration attitudes. These results hold in both North American and Western European studies (Hainmueller & Hopkins, 2014). Despite economic circumstances and cultural threats being the most extensively studied factors in this field, some studies suggest that consternation over crime levels is yet a stronger determiner of anti-immigration views (Fitzgerald et al., 2012).
A slightly different angle is chosen by Bos et al. (2016), who specifically focus on four different frames about immigrant integration and measure their influence on anti-immigration attitudes: the victimization frame (portrays women within immigrant groups as victims of ‘gender inequality and oppression’), the multicultural frame (poses that cultural diversity is a valuable asset), the emancipation frame (the position of immigrants in our society is problematic, civic participation among them should be enhanced) and the assimilation frame
21
(immigrants need to adopt our national norms and values). They conclude that the victimization frame most strongly increased anti-immigration attitudes, while the opposite holds for the multicultural frame.
Furthermore, several framing studies on immigration now incorporate emotions into their observations, paying specific attention to anxiety. Brader et al. (2008) analyse the interaction between the experience of threats and anxiety. Although these are different concepts – threats are related to cognition, whereas anxiety and fear are emotions – they are far from independent. A (cognitively) perceived threat can elicit emotions, although it is also possible that emotions can occur without conscious awareness of a threat (Damasio, 2000). Brader et al. (2008) conclude that citizens change their preferences, their inquisitiveness, and their behaviour, when a political stimulus elicits anxiety. An interesting element in this study is the role of emotional cues based on the ethnicity of individuals or groups. Respondents that were confronted with a stigmatizing outgroup cue about Latino immigrants experienced more anxiety and were more inclined to change their attitudes. These findings are supported by other studies that conclude that identity of immigrants influences opposition against them (Schneider, 2007). An interesting explanation for the influence of ethnic cues is given in a recent study by Aarøe, Petersen, & Arceneaux (2017). They describe how anti-immigration attitudes and the effects of ethnic cues find their origin in a behavioural immune system. This psychological system is triggered through the experience of disgust and is meant to protect individuals from disease. They conclude that this immune system leads to aversion of inter-group contact and the presentation of immigrants as outsiders and contagious diseases.
In sum, immigration is an issue that is part of many studies on emotion and attitude formation. There are several reasons why citizens perceive immigration as a threat and experience anxiety, fear and anger. Prior studies of this issue find – in line with the affective intelligence theory – that experiencing anxiety amplifies changes in attitude and can therefore strengthen anti-immigration attitudes. In contrast to Marcus et al. (2000), who implicate similar effects for anxiety and fear, I hypothesise - in line with the psychology literature (McNaughton & Corr, 2004; Öhman, 2008) - that anxiety and fear will have their own distinct effects. While anxiety is expected to stimulate critical evaluation of new information and induce approach-related behaviour, fear triggers threat aversion. Where anxiety leads to changes in attitude, fear will not. Furthermore, the experience of anger is associated with a reliance on prior views and
22
will therefore also forego any attitude change. This results in the following hypothesis, concerning the effects of anxiety, fear and anger:
H4: Citizens that experience anxiety will strengthen their anti-immigration attitudes, in contrast to citizens that experience anger and fear
Data & Methods
This study was undertaken through an online pre-test/post-test between-group survey experiment, developed in Qualtrics Online Thesis Tool. In the pre-test, respondents were asked to answer questions about their demographic (age, gender, education level), anti-immigration attitudes, political interest and left-right positioning (control variables). After the pre-test, respondents were randomly assigned to one of the two experimental conditions. Group 1 was confronted with a highly negative message on immigration which described immigrants as a threat to society. Group 2 was exposed to a balanced message that emphasized the responsibility of the Netherlands to help refugees, but also mentions the possible downsides of increased immigration. Both treatments can be found in the appendix of this study. After the experiment, respondents were asked – using items from the Discrete Emotion Questionnaire (DEQ) (Harmon-Jones, Bastian, & Harmon-Jones, 2016) - to report if they experienced fear, anxiety or anger. In the last part of the survey, respondents were expected to answer the same questions on immigration attitudes. An overview of the survey experiment is shown in figure 1.
Figure 1: Survey experiment overview
Sample N=329 Group 1 Negative N=169 Experiment Pre-test Group 2 Baseline N=160 Post-test Group 1 Negative N=169 Group 2 Baseline N=160
23 Sampling and demographics
A convenience sample (N=329) was taken by recruiting respondents by e-mail and social media (Facebook, Facebook Messenger, Twitter and WhatsApp). In addition to this, train passengers at Utrecht Central Station were – during two days of the survey period (April 12 – April 19, 2017) - asked to fill in the survey on a tablet. Furthermore, respondents were asked to distribute the survey experiment to create a snowball effect and enlarge the total sample size. To recruit even more respondents, participants could win a Bol.com voucher when they completed the survey experiment and distributed their email address. The minimum sample size for this study was computed a priori with G*Power for an independent samples t-test (two-tailed) and resulted in a minimum sample size of 105 for each experimental condition (Total N=210) with a given power of 0.95 and an effect size of d= 0.5.
The described sampling method resulted in a sample which is relatively unrepresentative for the Dutch population. This is supported by the scores on the demographic items used in this study and reported in table 1 below. Respondents are relatively young (M=32.8), women are overrepresented (58.2%) and higher educated personnel form a substantive part of the sample (83.6% HBO/WO).
As described earlier, respondents were randomly assigned to one of the two experimental conditions. Group 1 (N=169), was confronted with a very negative message on immigration. Group 2 (N=160), was exposed to a balanced message. To compare these groups, it is important that respondents are similar in terms of demographics. The descriptive statistics for both groups are shown in the tables below (1-3).
Table 1: Frequencies gender & education level
Gender Freq. group 1 (negative
condition)
Freq. group 2 (baseline condition)
Male 40.2% 43.4 %
Female 59.8% 56.6 %
24
Table 2: Frequencies education level
Education level Group 1 (negative condition) Group 2 (baseline condition
Geen/lager- of basisonderwijs 0.6% - vmbo/mavo/lbo 2.4% 2.5% mbo/mts/meao 4.7% 4.4% havo/vwo 7.7% 10.6% hbo/wo 84.6 82.5% Valid N 169 160
Table 3: Descriptive statistics age
Age Group 1 (negative condition) Group 2 (baseline condition)
Mean M=32.4 M=33.1
SD SD=14.7 SD=16.1
Valid N 169 157
These descriptive statistics show that there are no meaningful differences between the two experimental groups in terms of demographics. Four dummy variables are computed to include education level as a variable in the linear regression analysis. The category geen/lager- of basisonderwijs is excluded because only one respondent reported this education level.
Anti-immigration attitudes
In this study, the attitudes towards immigrants are both measured in the pre-test (independent variable) and post-test (dependent variable) by using the same items. These items are inspired by items used in the Longitudinal Internet Studies for the Social sciences (LISS). Respondents were asked to react to six different statements about immigration by dragging a graphic slider: 1. Meer immigratie leidt tot een toename van criminaliteit (1=Volledig mee oneens,
5=Volledig mee eens)
2. Sommige sectoren van de economie kunnen alleen blijven draaien omdat er immigranten werken. (1=Volledig mee oneens, 5=Volledig mee eens)
3. Wanneer ik aan immigranten denk, denk ik aan alle problemen die zij veroorzaken (1=Volledig mee oneens, 5=Volledig mee eens)
25
4. Onze nationale identiteit gaat verloren door de komst van immigranten (1=Volledig mee oneens, 5=Volledig mee eens)
5. Het moet makkelijker worden om in Nederland asiel te krijgen (1=Volledig mee oneens, 5=Volledig mee eens)
6. Over het algemeen leveren immigranten een positieve bijdrage aan de Nederlandse samenleving (1=Volledig mee oneens, 5=Volledig mee eens)
The scores for these items are combined by calculating a mean attitude scale (1-5) for both the pre-test and the post-test attitudes. However, a reliability analysis and factor analysis were first conducted to assess whether these items are internally consistent and form one dimension. Although, these items formed a moderately reliable scale (α=0.52), item nr. 5 lowered the reliability considerably. This may be explained by the fact that this item is related to national immigration policy, while the other items measured general attitudes towards immigration and immigrants. In conclusion, I deleted item nr. 5, which increased Cronbach’s alpha significantly (α=0.78). Furthermore, a principal component analysis indicated that these five items form one dimension. The descriptive statistics for pre-test and post-test ant-immigration attitudes are described in table 4. The differences between groups are marginal, which indicates that the two experimental groups are similar in terms of anti-immigration attitudes.
Table 4: Descriptive statistics anti-immigration attitudes (Pre-test and post-test)
Variable Group 1 (negative condition) Group 2 (baseline condition)
Attitude pre-test M=2.46 M=2.47
SD=0.73 SD=0.63
Attitude post-test M=2.49 M=2.49
SD=0.75 SD=0.65
Valid N 169 160
Note: M=mean, SD=standard deviation
Emotions
Respondents were confronted with 11 items, measuring three negative emotions (anxiety, fear and anger), to test whether a link could be made between their emotions and their attitude on
26
immigration. Specific attention was paid to any feelings of anxiety and potential decreases in any individual’s reliance on their predispositions). The questionnaire used to measure these emotions is a Dutch translation of the recently published Discrete Emotion Questionnaire (DEQ) (Harmon-Jones et al., 2016). The advantage of this tool, compared to other self-report questionnaires, is that it consists of 3 or 4 subscales per emotion, which makes it more reliable than a single item. This is particularly important because individuals are not always capable of reporting their feelings in discrete terms (Barrett, 2006). Initially, this study measured anxiety by using four items (vrees, ongerustheid, nervositeit and bezorgheid), fear by three items (bangheid, angst and paniek) and anger by four items (boosheid, woede, razernij and kwaadheid) employing the following item:
In welke mate ervoer u [Emotie] tijdens het lezen van de voorgaande tekst? (1=Helemaal niet, 7=In zeer sterke mate)
Unfortunately, the number of missing values on these subscales was found to be relatively high (9.1% - 41.6%) (see table 5).
Table 5: Missing values emotions items
Emotion subscale N Missing count Missing percent
Ongerustheid (ax) 288 41 12.5% Angst (f) 245 84 25.5% Woede (ag) 266 63 19.1% Boosheid (ag) 266 63 19.1% Nervositeit (ax) 249 80 24.3% Bangheid (f) 223 106 32.2% Razernij (ag) 210 119 36.2% Vrees (ax) 233 96 29.2% Bezorgdheid (ax) 299 30 9.1% Kwaadheid (ag) 235 94 28.6% Paniek (f) 192 137 41.6%
The high number of missing values may have multiple causes. Firstly, items later in the survey contained more missing values than items at the beginning, which suggests that
27
participants skipped later items because the survey took them too much time or they were confused by the similarity of the items. Secondly, participants had to drag a slider to indicate the level of emotion they experienced. The standard position of this slider was located to the left of the scale and indicated the lowest degree of emotion. Nevertheless, participants had to click on the slider to confirm this position. Lastly, respondents were not warned by the online survey when they forgot to answer one of the items and could continue to the next page without noticing.
This resulted in the decision to exclude five items which contained the most missing values. The explicit goal of this measure was to increase the number of useful cases. The correlation between the two remaining items for each emotion is strong and significant at the 0.01 level, similar to the findings of Harmon-Jones et al. (2016). For anxiety, a mean score on the items bezorgdheid and ongerustheid (r=0,55, p<.01) is computed, for fear on angst and bangheid (r=0.78, p<.01) and for anger on woede and boosheid (r=0.88, p<.01). The descriptive statistics for these combined scores can be found in table 6.
Table 6: Descriptive statistics emotion scores
Variable Group 1 (negative condition Group 2 (baseline condition
Fear M=2.49 M=2.05 SD=1.49 SD=1.28 Anxiety M=3.67 M=3.03 SD=1.68 SD=1.87 Anger M=3.70 M=2.27 SD=1.75 SD=1.34 Valid N 137 107
Note: M=mean, SD=standard deviation
The affective intelligence theory (AIT) by Marcus et al. (2000) combines fear, anxiety, anger and other negative emotions in one negative dimension and expect them to have the same effects on political behaviour. I expect – as described in the theory section - different effects of these emotions on the formation of anti-immigration attitudes and I therefore measure anxiety, fear and anger separately. Nevertheless, earlier studies found strong correlations
28
between these variables. For this reason, I included a correlation matrix to assess the relationship between these variables (table 7).
Table 7: Correlation matrix fear, anxiety & anger
**= correlation is significant at the 0.01 level
As expected, the self-reported emotions are highly correlated. Fear and anxiety show the highest correlation, r(258) = .70, p < .01. Yet, the results for fear and anger, r(244) = .55, p < .01. and anxiety and anger, r(271) = .58, p < .01 also indicate fairly high positive correlations.
Experimental conditions
To elicit negative emotions and potentially stimulate a change in anti-immigration attitudes, respondents were randomly exposed to either a negative message concerning immigration or a balanced message. The latter can be considered as the baseline experimental condition. The messages are inspired by real speeches from the Dutch anti-immigration party, Party for Freedom (PVV) (negative condition) and the Christian Party, ChristenUnie (CU) (baseline condition). Although these messages were relatively short (100 words each), respondents were compelled to read the message or stay on the active page for at least 25 seconds before the online survey allowed respondents to go to the post-test. The goal of this measure was to increase the suggested effect of the experiment by stimulating attentive reading.
As part of the negative condition, the message contained multiple threats posed by immigration as described in the theory section of this thesis, namely perceived threats to common values, national identity, national security and the economy. The second group of respondents were assigned a balanced message, which emphasized that the Netherlands should take on its responsibility to help immigrants, even though immigrants will not always hold the same views on equality between men and women, the freedom of religion and the freedom of speech as Dutch natives. To include the experimental treatment in the regression analysis, a
Emotion Fear Anger Anxiety
Fear - 0.55** 0.70**
Anger 0.55** - 0.58**
29
dichotomous variable has been created (0=baseline condition, 1=negative condition). The complete messages can be found in the appendix of the study.
Control variables
Two other variables that are worth mentioning are political interest, measured with the following item: ‘In welke mate bent u geïnteresseerd in politiek?’ (1=not interested, 7=highly interested) and the left-right position of respondents on the political spectrum operationalised with the following item: ‘Er wordt vaak gezegd van politieke opvattingen dat zij links of rechts zijn. Wanneer u denkt aan uw eigen opvattingen, waar zou u zich zelf plaatsen?’(1=left wing, 7=right wing). Both variables are used as control variables in the linear regression analysis, which will be further described in the results section. Firstly, political interest can potentially be a factor that influences the effect of the experimental conditions. Individuals that are highly interested in political affairs may hold stronger predispositions than less interested individuals and may therefore be less inclined to critically consider new information (Taber et al., 2009). Secondly, the left-right position of respondents may be a significant predictor of anti-immigration attitudes (Van Der Brug & Van Spanje, 2009). The descriptive statistics for both variables are found below (table 8).
Table 8: Descriptive statistics political interest & left-right position
Variable Group 1 (negative condition) Group 2 (baseline condition)
Political interest (1-7) M=5.08 M=4.98
SD=1.53 SD=1.38
Left-right position M=3.58 M=3.46
SD=1.52 SD=1.41
Valid N 164 153
Note: M=mean, SD=standard deviation
Results
To test the four hypotheses as formulated in the theory section, several analyses are employed. Firstly, an independent-samples t-test is conducted to test if the two experimental groups differ in terms of experienced negative emotion (H1). The results of this T-test are shown in table 9.
30
Table 9: Independent samples t-test experimental groups
Variable Negative condition Baseline condition T-test df
M SD M SD
Anxiety 3.67 1.68 3.03 1.19 3.83** 306
Fear 2.49 1.49 2.05 1.28 2.57* 256
Anger 3.70 1.75 2.27 1.34 7.70** 275
*=p<.05, **=p<.01 note: M=mean SD=standard deviation
There was a significant difference found in the scores for anxiety, fear and anger between the negative and baseline experimental condition. While the mean difference of fear was significant at the 0.05 significance level; t (256) =2.57, p=<0.05, the difference for experienced anxiety; t (306) =3.83, p=<0.01) and anger; t (275) =7.70, p=<0.01 was significant at 0.01 level. These results suggest that citizens experience higher levels of negative emotions when confronted with negative political stimuli on immigration, compared to a balanced message.
Hypotheses two and three– which are based on the motivated reasoning theory - predict that the effect of the experimental condition on post-test is strongly dependent on citizen’s prior anti-immigration attitudes. To test this, a dichotomous filter variable was created, indicating if respondents scored low (1-3) or high (3-5) on the anti-immigration variable in the pre-test. Subsequently, two paired-samples t-tests are employed to test if the attitudes in the post-test differ from the attitudes in the pre-test. The results are shown in table 10.
Variable Pre-test Post-test T-test SE Mean df
M SD M SD Anti-immigration (negative) 3.62 0.48 3.64 0.51 -0.80 0.03 54 Anti-immigration (positive) 2.23 0.44 2.26 0.47 -1.47 0.01 274
Table 10: Paired-samples t-test for negative and positive citizens
