Motivated reasoners or Bayesian updaters of political attitudes:
How the Need for Cognition leads to stronger attitudes and the Need to Evaluate to the motivation to preserve them
Ming M. Boyer, BSc. 5789311
University of Amsterdam Research Master Thesis
Dr. Bert N. Bakker 29-01-2016
Abstract
Democracies rely on rational citizens in order to function as supposed because this way citizens grant a government the legitimacy to rule. But are political attitudes formed
rationally? Political theory assumes that citizens adjust their political attitudes rationally, in a Bayesian manner, whereby attitudes are based upon new information irrespective of whether the new information is pro-attitudinal or counter-attitudinal. But motivated reasoning theory predicts that pro-attitudinal information has a stronger effect on one’s attitude than counter-attitudinal information, leading to strengthened in stead of nuanced attitudes. This study investigates whether individual differences, in the Need for Cognition (NFC) and the Need to Evaluate (NTE), condition who update their attitudes as a Bayesian updater and who defend their attitudes as a motivated reasoner, i.e. by actively counterarguing counter-attitudinal arguments. In two quasi-experiments, respondents were exposed to pro- or counter-attitudinal information and wrote down their thoughts, of which the counterarguments were of interest. In line with motivated reasoning theory, counter-attitudinal information led to more
counterarguing than pro-attitudinal information. Counterarguing led to less attitude change in one of the two experiments. As expected, counter-attitudinal information led to more
counterarguing for respondents with a high NTE compared to those with low NTE. Yet, NFC did not have an effect on counterarguing. Concluding, high NTE individuals are more likely to be motivated reasoners, while low NTE individuals are more adequately described as Bayesian updaters of political attitudes. People differ in how they form political attitudes. It is argued that both motivated reasoners and Bayesian updaters can contribute to a
well-functioning ‘marketplace of ideas’.
Introduction
Democracies rely on informed and rational citizens in order to function as supposed, because this way citizens grant a government the legitimacy to rule (Bobbio, 1987). According to the Bayesian updating model, rational citizens update their attitudes in light of new information and equally strong arguments should lead to equally strong attitude adjustment (Gerber & Green, 1998). But are citizens that rational? In their model of motivated reasoning, Taber and Lodge (2006) show how individuals update political attitudes with several biases. In general, people tend to avoid messages that are incongruent with their existing attitudes, leading to a confirmation bias in the exposure to political information. When they do face
counter-attitudinal messages, these are remembered and understood less than pro-counter-attitudinal messages and are actively counterargued in thought, leading to a disconfirmation bias in the processing of political messages (Jacobson, 2010; Lebo & Casino, 2007; Taber & Lodge, 2006).
Contrary to the Bayesian updating model, the motivated reasoning model suggests that attitudes are not updated rationally.
The question might not be whether citizens are Bayesian updaters or motivated reasoners. Instead, tentative evidence shows that there are individual differences in the strength of motivated reasoning effects (Arceneaux & Vander Wielen, 2013; Nir, 2011). Some people motivationally defend their existing beliefs, while others might actually exhibit Bayesian style updating of attitudes. Following Nir (2011), the most likely personality traits are considered in this study: the Need for Cognition and the Need to Evaluate. As previous research on motivated reasoning has not included its mechanism, the challenge remains to find out which individuals are motivated reasoners and which are Bayesian updaters.
The aim of the current study is to determine the role of individual differences in biased reasoning through the mechanism of selective judgment, i.e. the active counterarguing of attitudinally incongruent information (Jacobson, 2010; Lebo & Casino, 2007). The general
research question is: how, and to what extent, do the Need to Evaluate and the Need for Cognition influence selective judgment of political information? It is expected that counterarguing is more prevalent when respondents are exposed to counter-attitudinal information compared to pro-attitudinal information and that counterarguing leads to less attitude change. The Need for Cognition is expected to lead to more counterarguing, while the Need to Evaluate is expected to lead to more counterarguing against counter-attitudinal information, but not against pro-attitudinal information.
Because no study has yet investigated the mechanism of motivated reasoning, the current study is the first to examine the mediating role of counterarguing in motivated
reasoning. Moreover, the current study is, to the author’s best knowledge, the first to examine individual differences in counterarguing. By doing so, this study examines who can be best described as a motivated reasoner and for whom the Bayesian updating model is a more adequate representation.
Motivated reasoning in political attitude adjustment
Democracies rely on citizens to make rational political decisions. Specifically, using their informed and rational decision, citizens grant the best possible governments the legitimacy to rule (Bobbio, 1987). The traditional rational citizen of political theory collects and weighs arguments to make an informed political decision. This ideal type implies the existence of a memory-based model of decision-making (Lodge, Steenbergen, & Brau, 1995). After all, in order to make a rational decision, all arguments have to be obtained, understood,
remembered, evaluated and compared. Research with amnesiacs found that people without a well-functioning long-term memory showed similar candidate preferences as when they could still remember, but did not remember why (Coronel et al. 2012). Therefore, the memory-based model of the rational citizen should be reconsidered.
The alternative to the memory-based model is a Bayesian-style updating model, in which political evaluations of arguments are stored in attitudes (Gerber & Green, 1998). These attitudes are adjusted when one is confronted with new arguments. Negative information about a subject leads to a negative adjustment of one’s attitude towards this subject, while positive information has the opposite effect. Arguments similar in strength lead to equally large attitude adjustments. In this manner, citizens save precious cognitive energy and are still able to rationally form political attitudes.
Contrary to the Bayesian updating model, which predicts that rational processing of information should lead to equal attitude change for pro- and counter-attitudinal information, Taber and Lodge (2006) show that attitudes are adjusted more towards extremity than towards nuance, leading to attitude polarization. Accordingly, counter-attitudinal information is more often met with cognitively effortful reasoning mechanisms that lead to less attitude
adjustment than pro-attitudinal information (Gaines, Kuklinski, Quirck, Peyton, & Verkuilen, 2007; Lebo & Casino, 2007; Nyhan & Reifler, 2010; Taber & Lodge, 2006). Individuals accept pro-attitudinal information more easily than counter-attitudinal information. The explanation for this effect is called motivated reasoning (MR).
According to MR, biased reasoning is caused by a combination of motivations, defined as “any wish, desire, or preference that concerns the outcome of a given reasoning task” (Kunda, 1990, p. 480). In essence, individuals experience discomfort when they notice an inconsistency in their beliefs (Festinger, 1957). Therefore, humans naturally tend to avoid this cognitive dissonance. There are several ways in which cognitive dissonance is evoked. It occurs when one notices that his beliefs aren’t accurate. Moreover, it occurs when one’s beliefs are contradicted. In order to avoid these two instances of cognitive dissonance,
Accuracy motivations are defined as “the need to maintain a correct belief about a given issue” (Nir, 2011, p. 506). To avoid the cognitive dissonance of having incorrect beliefs, they lead to the use of more complex cognitive strategies (Tetlock & Kim, 1987) and reduce cognitive biases (Kruglanski & Freund, 1983). Overall, MR research shows that accuracy motivations lead to less biased reasoning through the mechanism of more and better deliberation (Kunda, 1990).
However, people often defend false beliefs (Nyhan & Reifler, 2010). After all, confronting individuals with an argument that contradicts their priors, leads to cognitive dissonance. The motivations to reduce this kind of cognitive dissonance are called directional motivations, which are defined as a need “to uphold and maintain a desirable conclusion” (Nir, 2011, p. 506). In MR they are expected to lead to reasoning, biased towards a desired outcome (Kunda, 1990). However, “[p]eople do not seem to be at liberty to conclude whatever they want to conclude merely because they want to” (Kunda, 1990, p. 482). As discussed, they also need to avoid the cognitive dissonance of having inaccurate beliefs. Therefore, they attempt to justify their desired conclusion rationally. Memory search and inferential reasoning is biased towards those memories and reasoning strategies that support the desired conclusion. Therefore, in general, directional motivations lead to biased reasoning.
MR theory distinguishes at least four ways in which individuals defend their existing attitudes: selective exposure, selective judgment, selective perception and selective memory (Jacobson, 2010; Lebo & Casino, 2007). Selective exposure entails a confirmation bias when partisans select information, in which pro-attitudinal information is preferred over counter-attitudinal information (Taber & Lodge, 2006). Selective perception entails that motivated reasoners understand pro-attitudinal information better than counter-attitudinal information and therefore more often misinterpret counter-attitudinal messages (Gaines, Kuklinski, Quirck, Peyton, & Verkuilen, 2007; Lebo & Casino, 2007). Selective memory entails that
motivated reasoners remember more pro- than counter-attitudinal information (Jacobson, 2010; Nir, 2011). Finally, selective judgment entails a disconfirmation bias in processing counter-attitudinal information, by means of counterarguing, downplaying or twisting arguments to fit prior attitudes (Lebo & Casino, 2007; Taber and Lodge, 2006). The current study investigates the role of individual differences in the best studied mechanism of
motivated reasoning: selective judgment.
Individuals tend to actively counterargue counter-attitudinal information more than pro-attitudinal information (Taber & Lodge, 2006). Accordingly, counter-attitudinal
information is generally examined longer (e.g. Ditto & Lopez, 1992; Taber & Lodge, 2006), and seen as weaker (Taber & Lodge, 2006), than pro-attitudinal arguments. Most importantly, Taber and Lodge (2006) find that counter-attitudinal political arguments are more often counterargued and downplayed in thought than pro-attitudinal arguments. As expected when pro-attitudinal arguments encounter less scrutiny than counter-attitudinal ones, Taber and Lodge (2006) also show that reading both pro-attitudinal and counter-attitudinal arguments in random order leads to more extreme attitudes. However, although the suggestion is made, Taber and Lodge (2006) do not test whether attitude polarization is actually caused by the counterarguing of counter-attitudinal arguments. Therefore, in the current study it is hypothesized that counter-attitudinal information leads to more counterarguing than pro-attitudinal information (H1) and, in turn, counterarguing leads to less attitude change (H2).
Individual differences in motivated reasoning
Contrary to what most motivated reasoning research implies, the effects described in the previous paragraphs are not equally strong for everyone. Indeed, there are limits to the rationalizing of prior beliefs (Petty & Cacioppo, 1986a, Redlawsk, Civettini, & Emmerson, 2010). MR is therefore conditional. Lodge and Taber (2000) divide individuals according to their accuracy and directional motivations, leading to four ideal types that differ in the
strength of MR. Taber and Lodge (2006) find stronger MR effects for more knowledgeable and more partisan individuals. They explain this by these individuals having the means – knowledge – and (directional) motivation – prior attitude strength – to counterargue counter-attitudinal arguments. Recent research has, however, suggested that individual differences in personality traits condition citizens’ reasoning in politics (Arceneaux & Vander Wielen, 2013; Nir, 2011). Specifically, in this study, the focus lies on individual differences in the Need for Cognition and the Need to Evaluate.
Both Nir (2011) and Arceneaux and Vander Wielen (2013) study the effect of the Need for Cognition (NFC) on biased reasoning mechanisms. NFC is the chronic tendency to engage in and enjoy cognitively effortful behavior (Cacioppo & Petty, 1982). It correlates with many attributes that coincide with accuracy motivations, like objectivism, complexity, information seeking for decision-making and curiosity (Cacioppo, Petty, Feinstein, & Jarvis, 1996). Moreover, it has been shown that people high in NFC are more often open to counter-attitudinal information (Petty & Cacioppo, 1986b) and are more open to ambivalence about their preferred political party (Arceneaux & Vander Wielen, 2013; Rudolph & Popp, 2007). Indeed, both Nir (2011) and Arceneaux and Vander Wielen (2013) find tentative evidence for increased accuracy motivations in high NFC individuals. Nir (2011) finds that high NFC individuals remember more pro-attitudinal and counter-attitudinal political information. Likewise, Arceneaux and Vander Wielen (2013) find that high NFC individuals comparably adjust their attitudes towards their preferred political party and their adversary, after being exposed to negative information about that party. Concluding, high NFC individuals experience more cognitive dissonance from being factually wrong. Therefore, they have increased accuracy motivations, leading them to scrutinize pro- and counter-attitudinal information equally and adjust their attitudes accordingly.
Besides the motivation to treat pro- and counter-attitudinal information equally, high NFC individuals adjust their attitudes less easily than low NFC individuals (Haugtvedt & Petty, 1992). According to the Elaboration Likelihood Model high NFC individuals are more likely to form attitudes via the ‘central route’ than low NFC individuals (Petty & Cacioppo, 1986b). The ‘central route’ entails the conscious and elaborate formation of attitudes, in stead of attitude formation by peripheral cues, like music in an advertisement. The ‘central route’ therefore leads to stronger attitudes that are more difficult to change than attitudes formed via the ‘peripheral route’. Moreover, because their attitudes are more often consciously formed, high NFC individuals should have more counterarguments available in memory than low NFC individuals, against any argument. Concluding, high NFC individuals have the motivation to be accurate and therefore to treat pro- and counter-attitudinal information equally. Moreover, they have an inherent motivation to exhibit effortful cognitive behavior and have counterarguments more readily available than low NFC individuals. Finally, high NFC individuals’ attitudes tend to change less. It is therefore expected that high NFC
individuals exhibit more counterarguing than low NFC individuals, regardless of whether an argument is pro- or counter-attitudinal (H3), leading, in turn to less attitude change.
Individuals high in the Need to Evaluate (NTE) form stronger political attitudes in various ways (Nir, 2011). NTE is the chronic tendency of individuals to form evaluative thoughts and judgments (Jarvis and Petty, 1996). High NTE individuals form more
evaluations of political candidates and are more likely to engage in political activism, to vote and to use political media (Bizer et al., 2004). Moreover, high NTE individuals are more constrained by ideology than low NTE individuals (Federico & Schneider, 2007). High NTE individuals are more often either left or right wing across all political issues and are less politically ambivalent. This ideological constraint is created through increased evaluations made by high NTE individuals, that make the cognitive bonds with their beliefs stronger
(Jarvis and Petty, 1996). Therefore, the cognitive dissonance when their beliefs are being challenged is also stronger and more directional motivations are provoked. This leads high NTE individuals to show more defensive cognitive behavior (Nir, 2011). Therefore, high NTE individuals are expected to put more cognitive effort in disproving counter-attitudinal arguments than pro-attitudinal arguments. Concluding, the difference in the amount of counterarguing between individuals confronted with counter-attitudinal and pro-attitudinal information is stronger for higher NTE individuals than for lower NTE individuals (H4). See figure 1 for a schematic overview of the hypotheses
Figure 1
All hypotheses in a causal model
Method
Design
Procedure. To increase external validity, topical news issues are used in this
experiment. Using topical issues means that prior attitudes are beyond control, and an argument being pro- or counter-attitudinal cannot be manipulated. Therefore, a left-wing or right-wing argument is matched with one’s prior attitude to form pro- or counter-attitudinal condition. In this manner, two online quasi-experiments are conducted, in which respondents
read a pro- or counter-attitudinal argument. The complete procedure is depicted in figure 2. First, the respondents’ prior attitudes are measured, followed by the random assignment to a left-wing or right-wing argument in the first experiment. Subsequently, respondents take part in a thought-listing exercise in which they write down the thoughts they had during reading the argument. Finally, among measures irrelevant for this study, respondents administer their attitudes again. This experiment is performed twice in succession for two different subjects: shorter or longer jail sentences and the acceptance or refusal of Syrian refugees into the Netherlands. In between the two experiments, respondents are asked to answer questions inquiring their Need for Cognition and Need to Evaluate.
Figure 2
Stimulus material. Longer jail sentences and the acceptance of Syrian refugees are
chosen as subjects because they are part of the cultural ideological dimension (Van der Brug & Van Spanje, 2009) and their clear distinction between left-wing and right-wing positions. Explicitly, longer jail sentences are found to be the product of right-wing political influence (Jacobs & Carmichael, 2001), and resisting immigration is advocated mostly by right-wing parties (Van der Brug &Van Spanje, 2009). The stimulus material for each experiment consists of two arguments, equal in argument strength, but differing in valence.1 In the punishment experiment the left-wing text argues that prisons lead to more criminal activity and pleas for shorter jail time sentences. The right-wing argument suggests that potential criminals will choose crime less when punishment is harsher and proposes longer jail sentences. The left-wing argument in the refugee experiment argues that refugees can enrich society and appeals to welcome them into the Netherlands. The right-wing argument contends that, since refugees almost always enter the Netherlands through a safe country, they are not refugees but economic migrants and should not be let into the country.
Sample. For this experiment, a convenience sample was recruited via social media
websites and e-mail, through the researcher’s personal network and large social media groups. The initial sample consisted of 187 respondents. A drop-out rate of 23.5% in the punishment experiment led to a sample of 143 respondents. After excluding respondents who spent less than 5 seconds, or more than 10 minutes on the stimulus web-page or reported having had no thoughts at all, the final sample of the punishment experiment consisted of 123 respondents. A drop-out rate of 28.9% for the refugee experiment led to a sample of 133 respondents. Excluding respondents who spent less than 5 seconds, or more than 10 minutes on the
1 In a pre-test, respondents (N = 22) were asked to rate several right-wing and left-wing
arguments for argument strength, without taking their opinion into account. This led to left-wing and right-wing arguments that were regarded equally strong by left-wing and right-wing people. See Appendix 1 for detailed pretest results.
stimulus web-page or reported having had no thoughts led to 122 respondents in the refugee experiment. In the final sample young and highly educated people were overrepresented. The sample consisted of people between the ages of 19 and 74 and the average age was 33.92 (SD = 13.65) years old. Moreover, 50% of the respondents were below the age of 29. 78.8% of the sample had an applied sciences or university Bachelor degree. Gender was approximately equally distributed: 55.3% of the respondents were female.
Measures
Attitude change. Respondents were asked their attitudes regarding longer jail
sentences and the acceptance or refusal of Syrian refugees into the Netherlands, using the same items before (t1) and after (t2) seeing the stimulus material.2 The punishment attitude
scale questions were adopted from De Keijser, Van der Leeden and Jackson’s (2002) penal attitudes scales and consisted of three items from the “harsh treatment” paradigm.3 A principle component analysis (PCA) indicated that the t1 scale was uni-dimensional with an
eigenvalue of 2.20 and an explained variance of 73.33%. All items loaded above .80 on the component and the scale was highly reliable, Cronbach’s alpha = .82. A second PCA indicated that the t2 scale had an eigenvalue of 2.33 and explained 77.8% of the variance in
the items. Again, all loadings were above .80 and the scale was highly reliable, Cronbach’s alpha = .86. At both times the scale ranges from (1) left-wing (preferring shorter jail sentences) to (9) right-wing (preferring longer jail sentences). Both at t1 (M = 5.11, SD =
1.65) and t2 (M = 5.06, SD = 1.66) the respondents were centered around the ideological
middle of the scale, equally supporting longer and shorter jail time sentences.
2 The development of the punishment and refugee attitude scales is explained in more
detail in Appendix 2.
3 In the questionnaire, these items were mixed with items regarding the “social
constructiveness” paradigm, but these items turned out not to form a reliable scale. For the full analysis of the punishment items, see Appendix 3.
The Refugee Attitude Scale consisted of the immigration attitude questions from the LISS (Longitudinal Internet Studies for the Social sciences) panel administered by
CentERdata (Tilburg University, The Netherlands), adapted by changing the word
“immigrant” to “refugee”. This led to the exclusion of three items, forming a five-item scale.2
A PCA showed that the t1 items loaded on a uni-dimensional factor with an eigenvalue of
3.03, explaining 60.5% of the variance in the items. The scale was highly reliable, Cronbach’s alpha = .84. A second PCA confirmed that also the t2 items loaded on a single factor, with an
eigenvalue of 3.13, explaining 62.5% of the variance in the items. The t2 scale was also highly
reliable, Cronbach’s alpha = .85. All items loaded higher than .60 on the component at both t1
and t2. The attitude scale was measured from (1) right-wing (negative towards accepting
refugees in the Netherlands) to (9) left-wing (positive towards accepting refugees into the Netherlands). The respondents in the sample leaned more towards accepting Syrian refugees than refusing them, at t1 (M = 6.31, SD = 1.53) and t2 (M = 6.34, SD = 1.52).
The dependent variable, attitude change towards the argument (ACTA), is measured as the attitude change score of the respondent (t2 – t1). Subsequently, the change score is
adjusted to the valence of the argument to which the respondent was exposed. If the
respondent was exposed to the left-wing argument, ACTA is positive when a change is made to the left of the attitude scale (preferring shorter jail sentences or positive towards accepting refugees). If the respondent was exposed to the right-wing argument, ACTA is positive when a change is made to the right of the attitude scale (preferring longer jail sentences or negative towards accepting refugees). In this manner, a positive score on ACTA entails a change of one’s attitude in the ideological direction of the argument that the respondent was exposed to. An attitude change score is preferred over a lagged dependent variable model because of this ability to remove the valence of the stimulus material from the analysis.
Counterarguing. Counterarguing is measured by a thought listing exercise in which
respondents are asked to list every thought they had while reading the argument they were exposed to, in a maximum of ten text boxes. Following Taber and Lodge (2006), these
thoughts are coded to counter or bolster the argument in the stimulus material, or to be mixed, unclear or unrelated to the argument. Note that the thoughts are coded as compared to the argument the respondent had read. The same thought can therefore be a counterargument for one respondent and a pro-argument for another. 15% of the thoughts are coded by a second coder showing high the inter-coder reliability for the punishment experiment, Krippendorff’s alpha = .85, and the refugee experiment, Krippendorff’s alpha = .81.4 The counterarguing variable consisted of the number of thoughts to argue against the argument in the text.
Counter-attitudinal argument. The quasi-experimental manipulation, the argument
being either pro- or counter-attitudinal, is measured by matching one’s t1 attitude with the
valence of the stimulus material. The t1 attitude scores for both experiments were divided into
a left-wing and right-wing attitude by a median split. Respondents with a left-wing t1 attitude
score that were exposed to a right-wing argument and vice versa are considered to be in the counter-attitudinal group. Respondents with a left-wing t1 attitude score that were exposed to
the left-wing argument and respondents with a right-wing t1 attitude score that were exposed
to the right-wing argument are considered to be in the pro-attitudinal group.
Individual differences. The Need for Cognition (NFC) and Need to Evaluate (NTE)
are eight item scales adapted from Cacioppo, Petty and Kao (1984) and Jarvis and Petty (1996), respectively.5 The NFC was measured on seven-point Likert scales and consisted of one factor with an eigenvalue of 3.67, which explained 45.0% of the variance in the items. All items loaded higher than .50 on the component and the scale was highly reliable, Cronbach’s
4 Two coders coded the same 15% of the thoughts. The codebook can be seen in
Appendix 4.
5 For a detailed description of the item-selection process of the shortened individual
alpha = .83. NTE was measured on five-point Likert scales. A PCA indicated that one item did not fit properly in the NTE scale. After this item was removed, the seven remaining items formed one factor with an eigenvalue of 3.10, which explained 44.3% of the variance. All items loaded higher than .45 and the scale was reliable, Cronbach’s alpha = .78. Both NFC and NTE were recoded to form scales from 0 (low) to 1 (high). Respondents scored an average of .63 (SD = .17) on NFC and an average of .57 (SD = .18) on NTE.
Analysis. After the sample description, the randomization and manipulation of the
experiment are checked. Following Baron and Kenny (1986), hypothesis 1 and 2 are tested in three OLS regression analyses for each experiment. First, the main effect of the counter-attitudinal argument, as compared to the pro-counter-attitudinal argument on ACTA is estimated, followed by its effect on counterarguing. In the final regression model, the effect of counterarguing and counter-attitudinal information on ACTA are calculated. To test
hypothesis 3 and 4, these steps are repeated, with an additional main effect of NFC and NTE and an interaction term of the counter-attitudinal argument and NTE.
Results
Descriptive statistics. Respondents spent an average of 38.76 seconds (SD = 45.02)
on the page containing the punishment argument, which led to 3.63 (SD = 1.80) thoughts. Of these thoughts, 1.15 (SD = 1.39) were counterarguments, on average. The average Attitude Change Towards the Argument (ACTA) was .08 (SD = .75) in the punishment experiment. Turning to the refugee experiment, respondents spent an average of 57.76 seconds (SD = 56.57) on the page containing the argument. On average, this led to 3.39 (SD = 1.88) thoughts, of which 1.17 (SD = 1.29) were counterarguments. The average ACTA was -.04 (SD = .59) in the refugee experiment. ACTA was not significantly different than 0 in either experiment, but as ACTA can be both positive and negative, the aggregate level of ACTA says little about possible attitude changes at the individual level.
Randomization check. The experiment was randomized over the valence of the
argument in the stimulus material, i.e. whether the argument was left-wing or right-wing. In contrast, the independent variable of interest is defined by the argument being pro- or counter-attitudinal. In order to test whether there are systematic differences between respondents in the left-wing and right-wing groups or between respondents in the pro- and counter-attitudinal groups, randomization checks were conducted. There were no significant differences in age, gender, education, political knowledge and interest, NFC, NTE and the t1 attitudes between
the pro- and counter-attitudinal conditions, for both the punishment and the refugee experiment (see Appendix 6). Similarly, there were no significant differences between the left-wing and right-wing stimulus material for both experiments.
The pretest of the stimulus material indicated that the left-wing and right-wing arguments were considered equally strong by both left-wing and right-wing respondents. However, an additional check is conducted to test the effect of the valence of the argument on reading time, the number of thoughts, counterarguing and ACTA (see Appendix 6 for more detail). A series of OLS regressions indicated that in the punishment experiment there was no effect of the valence of the argument on the reading time, total of thoughts and ACTA. However, a significant effect was found on counterarguing. The right-wing argument led to .51 more counterarguments than the left-wing argument, b = .51, SE = .25, t = .07, p =.041. Turning to the refugee experiment, the right-wing argument was read 33 seconds longer than the left-wing argument, b = 33.46, SE = 10.04, t = 3.33, p = .001. Moreover, the right-wing argument led to .71 more counterarguments than the left-wing argument, b = .71, SE = .23, t = 3.11, p = .002. The valence of the stimulus material is therefore included in every analysis as a dummy control variable, meaning the respondent was exposed to either the (0) left-wing argument or (1) right-wing argument.
Manipulation check. A manipulation check was conducted to test whether the
median split of the t1 attitude scales into left-wing and right-wing attitudes predicted whether
or not the respondents agreed with the left-wing or right-wing argument, i.e. whether the argument was pro- or counter-attitudinal. The respondents were asked whether or not they agreed with the argument they read and a Fisher exact test indicated that respondents in the pro-attitudinal group answered “yes” significantly more often than those in the counter-attitudinal group for both the punishment and refugee experiment, N = 123, p < .001 and N = 122, p < .001, respectively. Indeed, the counter-attitudinal group agreed with the text less than the pro-attitudinal group, providing the cognitive dissonance required for a defensive
motivated reasoning reaction.
H1 & H2: MR and attitude change. Following Baron and Kenny (1986), OLS
regression models are used to test the motivated reasoning mediation analyses, for both the punishment and the refugee experiment (see Table 1). In step one, the dummy indicating whether participants received pro- or counter-attitudinal information is regressed on ACTA. In order to test the hypothesis that counter-attitudinal information leads to more
counterarguing (H1), the pro- or counter-attitudinal information is regressed on
counterarguing in step two. Finally, in step three, the hypothesis is tested that counterarguing leads to less ACTA (H2). Testing for mediation, in step three, both the pro- or counter-attitudinal information and counterarguing are regressed on ACTA. All regression analyses include the valence of the argument as a dummy control variable.
In the punishment experiment, step one of the mediation analysis indicates a
significantly higher ACTA for the counter-attitudinal group than for the pro-attitudinal group, b = .44, SE = .18, t = 2.50, p = .014 (Table 1, Punishment step 1). Controlling the valence of the argument, the counter-attitudinal argument led to .44 points more ACTA than the pro-attitudinal argument. Turning to test whether counter-pro-attitudinal information led to more
counterarguing (H1), step two of the analysis indicates that the counter-attitudinal argument led to .71 more counterarguments than the pro-attitudinal argument, controlling the valence of the argument, b = .71, SE = .24, t = 2.94, p = .004 (Table 1, Punishment step 2). Therefore, the punishment experiment offers support for H1. In step 3, testing H2, counterarguing had a negative effect on ACTA, b = -.13, SE = .07, t = -2.05, p = .043 (Table 1, Punishment step 3). Controlling the valence of the argument, for every counterargument produced, respondents adjusted their attitudes to the argument .13 points less on the nine-point scale. This supports that counterarguing leads to less ACTA (H2). When controlling for counterarguing, the pro- or counter-attitudinal argument had an even stronger effect on ACTA than in step one of the analysis, b = .53, SE = .18, t = 2.97, p = .004 (Table 1, Punishment step 3). Although
counterarguing contributes significantly to the explained variance of the model, ΔR2 = .03,F (1, 119) = 4.19, p = .043, a test of the mediation effect indicates that there is only a marginally significant indirect effect of the pro- or counter-attitudinal argument, via counterarguing on ACTA, Sobel’s test = -1.67, SE = .06, p = .095. Concluding, there seems to be a suppressed effect taking place, where the counter-attitudinal argument leads to more counterarguing and, subsequently, less ACTA than the pro-attitudinal argument. This negative indirect effect suppressed partly the effect that the counter-attitudinal group actually scored higher on ACTA than the pro-attitudinal group. Therefore, the punishment experiment supports H1 and H2.
The refugee experiment repeats the three steps of the punishment experiment. Contrary to the punishment experiment, the pro- or counter-attitudinal information had no effect on ACTA in step 1, b = -.02, SE = .11, t = -.17, p = .868 (Table 1, Refugee step 1). The counter-attitudinal argument did not lead to a different ACTA than the pro-attitudinal
argument. In order to test H1, step two indicates that there is a significant positive effect of the counter-attitudinal argument on counterarguing, b = .70, SE = .22, t = 3.21, p = .002 (Table 1, Refugee step 2). Controlling for the valence of the argument, respondents in the
counter-attitudinal group produced .70 more counterarguments than respondents in the pro-attitudinal group. Similar to the punishment experiment, the refugee experiment supports H1. Finally, unlike in the punishment experiment and rejecting H2, counterarguing had no effect on ACTA in the refugee experiment, b = .06, SE = .05, t = 1.40, p = .165 (Table 1, Refugee step 3). Similar to step one, there was no difference in ACTA between respondents in the pro- and counter-attitudinal group when controlled for counterarguing, b = -.06, SE = .11, t = -.55,
6 Note that the valence of the stimulus material and incongruence have the same standard
errors in every model, and in one model even have an identical slope. The identical standard errors are caused by the use of two binary variables in simple regression models that assumes constant variance. The identical slope in step two of the Refugee CMR model is caused by identical group means. In order to exclude the possibility of overly similarity between incongruence and the valence of the stimulus material, a crosstab between the two variables was made. In the congruent condition, 43 respondents saw the left-wing stimulus and 36 respondents saw the right-wing stimulus. In the incongruent condition, 37 respondents saw the left-wing stimulus and 27 respondents saw the right-wing stimulus. The similarity in the models is therefore due to coincidence.
Table 1
Unstandardized coefficients in the classical motivated reasoning mediation analyses for the punishment and refugee experiment6
Punishment (N = 123) Refugee (N = 122) Step 1; DV: ACTA Step 2; DV: counter-arguing Step 3; DV: ACTA Step 1; DV: ACTA Step 2; DV: counter-arguing Step 3; DV: ACTA
b (SE) b (SE) b (SE) b (SE) b (SE) b (SE)
Constant -.06 (.15) .54* (.21) .01 (.15) -.09 (.09) .67** (.18) -.13 (.09) Counter-attitudinal .44* (.18) .71* (.24) .53* (.18) -.02 (.11) .70* (.22) -.06 (.11) Valence -.19 (.18) .48* (.24) -.12 (.18) .12 (.11) .70* (.22) .07 (.11) Counterarguing -.13* (.07) .06 (.05) Adjusted R2 .04 .08 .06 .00 .13 .00 F 3.57* 6.60* 3.84* .58 10.35** 1.04 ΔR2 .03 .00 ΔF 4.19* 1.96 Note. *p < .05; **p < .001
p = .580 (Table 1, Refugee step 3). Counterarguing did not contribute to the explained variance in step three, F (1, 118) = 1.96, p = .165. Concluding, in line with the punishment experiment, the refugee experiment provides evidence for H1. However, and contrary to the punishment experiment, the refugee experiment fails to support H2.
H3&H4 individual differences in MR. H3 puts forward that a higher NFC leads to
more counterarguing, regardless of whether the argument is pro-, or counter-attitudinal. This increased counterarguing should in turn lead to less ACTA. H4 predicts that the difference in counterarguing between the pro- and counter-attitudinal groups is stronger for people with a higher NTE than for people with a lower NTE. H3 and H4 are tested in the same moderated mediation model, that follows the same three steps as the mediation analysis as defined by Baron and Kenny (1986; see Table 2). In step one, the dummy indicating whether participants received pro- or counter-attitudinal information, NFC, NTE and the interaction between the dummy and NTE are regressed on ACTA. In step two, testing both H3 and H4, the same independent variables as in step 1 are used to predict counterarguing. Step three replicates step one, but adds counterarguing to the model as an independent variable. Again, the valence of the argument is included as a dummy control variable in all analyses and the three steps are repeated for the punishment and the refugee experiment.
First, starting with NFC (H3) in the punishment experiment, in step one of the analysis, there is a negative effect of NFC on ACTA, b = -1.07, SD = .54, t = -2.00, p = .048 (Table 2, Punishment step 1). Controlling the valence of the argument, higher NFC
respondents adjusted their attitudes towards the argument less than lower NFC respondents.
Step two of the analysis fails to find evidence that higher NFC respondents counterargue the arguments more than lower NFC participants (H3), b = .28, SD = .78, t = .36, p = .718 (Table 2, Punishment step 2). In step three of the model, the effect of NFC on ACTA - which was significant in step one - is still marginally significant, b = -1.03, SD = .53, t = -2.00, p = .053
(Table 2, Punishment step 3). Although a higher NFC seems to reduce ACTA, counterarguing does not mediate the effect. The punishment experiment therefore rejects H3.
Table 2
Unstandardized regression coefficients in the individual differences motivated reasoning mediation/moderation analyses for the punishment and refugee experiment
Punishment Refugee Step 1; DV: ACTA Step 2; DV: counter-arguing Step 3; DV: ACTA Step 1; DV: ACTA Step 2; DV: counter-arguing Step 3; DV: ACTA b (SE) b (SE) b (SE) b (SE) b (SE) b (SE)
Constant -.10 (.45) -.03 (.65) -.10 (.45) .22 (.33) 1.03 (.64) .14 (.33) Counter-attitudinal 2.12** (.58) 1.36 (.84) 2.30** (.58) -.05 (.39) -1.28† (.77) .05 (.40) NTE 1.32 † (.72) .76 (1.05) 1.42* (.71) -.27 (.43) -.87 (.84) -.21 (.43) NFC -1.07* (.54) .28 (.78) -1.03† (.53) -.24 (.34) .20 (.67) -.25 (.34) Counter-attitudinal *NTE -2.87* (.97) -1.11 (1.41) -3.02* (.96) .04 (.67) 3.54* (1.32) -.21 (.69) Valence -.25 (.17) .47* (.24) -.18 (.17) .10 (.11) .74** (.22) .05 (.11) Counterarguing -.13* (.06) .07 (.05) Adjusted R2 .11 .08 .13 -.02 .17 -.01 F 3.89* 3.04* 4.09** .49 6.01** .81 ΔR2 .02 .01 ΔF 4.50* 2.37 Note. †p < .10; *p < .05; **p < .001
The refugee experiment follows the same three steps to test H3. In step one, there is no effect of NFC on ACTA, b = -.24, SD = .34, t = -.69, p = .489 (Table 2, Refugee step 1). Higher NFC respondents adjusted their attitudes towards the argument similar to lower NFC respondents. Testing whether higher NFC respondents counterargued the arguments more than lower NFC respondents (H3), step two finds no effect of NFC on counterarguing, b =
.20, SD = .67, t = .31, p = .760 (Table 2, Refugee step 2). Again, hypothesis 3 is rejected. Similar to step one, step three in the refugee experiment confirms that there is no effect of NFC on ACTA, b = -.25, SD = .34, t = -.74, p = .460 (Table 2, Refugee step 3). In line with the punishment experiment, the refugee experiment showed no effects of NFC on
counterarguing. Both experiments reject H3.
Turning to NTE (H4) in the punishment experiment, the same analysis is used as for NFC. In step one of the analysis, there is a significant interaction effect of the pro- or counter-attitudinal argument and NTE on ACTA, b = -2.87, SE = .97, t = -2.95, p = .004 (Table 2, Punishment step 1). Controlling the other variables, low NTE respondents adjusted their attitudes towards the argument more in the counter- than in the pro-attitudinal condition, while high NTE respondents did not change their attitudes at all (Figure 3). Testing whether the difference in counterarguing between the pro- and counter-attitudinal condition is greater for high than for low NTE respondents (H4), step two of the analysis indicates no interaction effect of NTE and the pro- or counter-attitudinal information on counterarguing, b = -1.11, SD = 1.41, t = -.79, p = .432 (Table 2, Punishment step 2). Controlling for the valence of the stimulus material, and rejecting H4, the difference in counterarguing between the pro- and counter-attitudinal condition is the same for low and high NTE respondents. Finally, in step three, the interaction effect between NTE and the pro- or counter-attitudinal argument on ACTA that was significant in step one, remains, b = -3.02, SD = .96, t = -3.14, p = .002. Although lower NTE respondents change their attitudes more than higher NTE respondents in the punishment experiment, counterarguing is not the mechanism through which this
Figure 3
Predicted ACTA with 95% CI, by pro- or counter-attitudinal argument and low (M - SD) or high (M + SD) NTE
Finally, the moderation by NTE (H4) is tested for the refugee experiment. In step one of the analysis, there is no interaction effect of NTE and the pro- or counter-attitudinal information on ACTA, b = .04, SD = .67, t = .06, p = .951 (Table 2, Refugee step 1). The difference in attitude change between the pro- or counter-attitudinal condition did not differ between low and high NTE respondents. In step 2, testing H4 for the refugee experiment, there is a significant interaction effect of NTE and the pro- or counter-attitudinal information on counterarguing, b = 3.54, SD = 1.32, t = 2.68, p = .008 (Table 2, Refugee step 2). As depicted in figure 4, controlling the other variables, low NTE respondents counterargued the arguments equally, while high NTE respondents reported significantly more
counterarguments in the counter-attitudinal than in the pro-attitudinal condition. The interaction effect offers support for H4, which stated that the difference in counterarguing between the pro- and counter-attitudinal condition is larger for high than for low NTE
-1 -. 5 0 .5 1 L in e a r p re d ict io n o f AC T A Pu n ish me n t Pro-attitudinal Counter-attitudinal
respondents. Finally, step 3 of the analysis shows again that counterarguing has no effect on ACTA, b = .07, SD = .05, t = 1.54, p = .126, and nor did the interaction between NTE and the pro- or counter-attitudinal information, b = -.21, SD = .69, t = -.31, p = .756 (Table 2, Refugee step 3). Concluding, in the refugee experiment, high NTE respondents were found to
counterargue more in the counter-attitudinal than in the pro-attitudinal condition, while low NTE respondents counterargued equally in both conditions. Because this effect was not found in the punishment experiment, this offers partial evidence for H4. The confirmed and rejected hypotheses for both experiments can be seen in table 3.
Figure 4
Predicted number of counterarguments with 95% CI, by pro- or counter-attitudinal argument and low (M - SD) or high (M + SD) NTE
Unexpected results. The reported analyses were used to test hypotheses 1 through 4,
but entailed more effects than were expected from theory. Firstly, MR expects pro-attitudinal information to have a stronger effect on one’s attitude than counter-attitudinal information, but in the punishment experiment, respondents who were exposed to the counter-attitudinal
0 .5 1 1 .5 2 2 .5 3 L in e a r p re d ict io n o f C o u n te ra rg u in g R e fu g e e Pro-attitudinal Counter-attitudinal
argument adjusted their attitudes more towards the argument than those exposed to the pro-attitudinal argument (Table 1, Punishment step 1 and 3). This effect only takes place for low NTE respondents, while high NTE respondents do not change their attitudes at all (Table 2, Punishment step 1 and 3; Figure 3). In contrast, in the refugee experiment, attitudes did not change at all when exposed to a counter- or pro-attitudional information (Table 1, Refugee step 1 and 3). Secondly, in the punishment experiment high NFC respondents adjusted their attitudes to the argument less than low NFC respondents, regardless of whether they were exposed to the pro- or counter-attitudinal argument (Table 2, Punishment step 1 and 3). This effect was also not visible in the refugee experiment, as respondents did not change their attitudes at all. Concluding, motivated reasoning effects of NFC and NTE were found that did not include counterarguing.
Discussion
This study investigated how, and to what extent, the Need for Cognition (NFC) and the Need to Evaluate (NTE) influence selective judgment of political information. The aim was to find out who is a motivated reasoner and who a Bayesian updater. Supporting Taber and Lodge’s (2006) motivated reasoning theory, this study shows evidence that
counter-Table 3
Rejected and supported hypotheses in the punishment and refugee experiment Punishment experiment
Refugee experiment H1 Counter-attitudinal information leads to more
counterarguing than pro-attitudinal information. √ √ H2 More counterarguing leads to less attitude change.
√ x
H3 A higher Need for Cognition leads to more
counterarguing, regardless of whether an argument is pro-, or counter-attitudinal.
x x
H4 The effect of incongruence on counterarguing is stronger for people with a higher Need to Evaluate than for people with a lower need to evaluate.
attitudinal information leads to more counterarguing than pro-attitudinal information (H1). Moreover, it presents partial evidence that counterarguing is indeed a mechanism in
motivated reasoning, by showing in one of the experiments that counterarguing leads to less attitude change (H2). Partial evidence is also shown that higher NTE individuals had stronger directional motivations than lower NTE individuals. Higher NTE individuals produced more counterarguments against counter-attitudinal information than against pro-attitudinal
information, while low NTE individuals did not discriminate between the two (H3). There were no differences found in counterarguing between low and high NFC individuals (H4). In conclusion it can be stated that, although selective judgment plays a role in the updating of political attitudes, it only does so for high NTE individuals and that NFC has no role in selective judgment. Therefore high NTE individuals can be described as motivated reasoners, while low NTE individuals more adequately resemble Bayesian updaters.
The increased counterarguing of counter-attitudinal information offers additional evidence for Taber and Lodge’s (2006) motivated reasoning theory. The mediating role of counterarguing found in the punishment experiment extends the findings by Taber and Lodge by offering experimental evidence that counterarguing is actually one of the cognitive
mechanisms that causes decreased attitude change by counter-attitudinal information. But in the experiment concerning Syrian refugees, counterarguing did not have an effect on attitude change. Clearly, the two subjects are very different. Promoting shorter or longer jail sentences is by no means a political controversy. It is a long-lasting discussion (De Keijser, Van der Leeden, & Jackson, 2002), with political attitudes slumbering in times of slow discussion. Conversely, the Syrian refugee crisis is topic of high salience and controversy (Albarahi, 2015). Counterarguing counter-attitudinal information should be expected to be present in both salient topics and less salient topics. But attitudes about salient topics should be more carefully thought through. Therefore these attitudes are more difficult to change. Perhaps, this
is also the case in the current study, where individuals were found to counterargue counter-attitudinal information more than pro-counter-attitudinal information in both experiments, but counterarguing only had an effect on the less salient punishment attitude, while the salient refugee attitude was robust to change.
In line with the findings of Nir (2011), the current study finds evidence that high NTE individuals are more prone to having directional motivations than low NTE individuals. In the refugee experiment, high NTE individuals counterargued the counter-attitudinal argument more than the pro-attitudinal argument, while low NTE individuals counterargued both equally. In the punishment experiment, though, there was no effect of NTE counterarguing. Again, the difference between the experiments can be explained by the salience of the topic. High NTE individuals might be strongly motivated to counterargue the counterargument on the salient refugee topic, leading to a strong difference with low NTE individuals. In contrast, the lower salience of the punishment topic may annul the difference between high and low NTE individuals in the motivation to actively counterargue the argument.
Also irrespective of counterarguing, this study presents evidence that high NTE individuals were more defensive than low NTE individuals. In the punishment experiment, low NTE individuals adjusted their attitude towards nuance when faced with either pro- or counter-attitudinal information, while high NTE individuals did not change their attitudes at all. This means that, lthough not through counterarguing, in the punishment experiment high NTE individuals did defend their beliefs. Concluding, with regard to NTE, this study’s findings align with Nir (2011), who finds that high NTE individuals remember more pro-attitudinal than counter-pro-attitudinal arguments, while low NTE individuals did not differ in memory. High NTE individuals have more directional motivations than low NTE individuals.
In contrast to NTE, NFC did not have any effect on counterarguing in this study. This contradicts findings by Nir (2011) and Arceneaux and Vander Wielen (2013) who find that
high NFC individuals have increased accuracy motivations. Unexpectedly, in the punishment experiment, high NFC individuals changed their attitudes less than low NFC individuals, regardless of counterarguing. However, the finding that higher NFC change their attitudes less is in line with the Elaboration Likelihood Model, that predicts high NFC individuals’ attitudes to be stronger and more robust to change (Petty & Cacoppo, 1986b). The assumption of Arceneaux and Vander Wielen (2013) that decreased attitude change of high NFC individuals is caused by motivated reasoning is hereby drawn to question. High NFC individuals’
attitudes seem to be inherently more difficult to change, more likely caused by earlier deliberation than increased scrutiny of a presented argument. As motivated reasoning takes place at the moment that one is confronted information, it can not be the reason for attitude change differences between low and high NFC individuals. Future research should consider high NFC individuals’ attitudes are more difficult to change than low NFC individuals, regardless of defensive cognitive mechanisms.
As indicated by the differences between the two experiments, future research on motivated reasoning effects should put a stronger focus on the salience of the topic. The discussion of the current study suggests that on the one hand a high salience topic can lead to more counterarguing against counter-attitudinal information, and therefore make the
mechanisms of motivated reasoning more visible. On the other hand, attitudes on high salience topics are more difficult to change than attitudes on low salience topics. As a result, attitude change on high salience topics are likely to reach such low levels that a very large sample is needed to be able to study the effects. Therefore, low high salience topics are more suited to study the effect of motivated reasoning. In order to expand knowledge about
motivated reasoning mechanisms, the effect of the salience of the topic is essential and future research should proceed to experimentally manipulate topic salience.
This study relied on existing attitudes on recent topics to determine whether information was pro- or counter-attitudinal. Besides increasing the external validity of the experiment, this choice has two downsides. Firstly, the individuals’ prior attitude being either left-wing or right-wing was based on a median split and was therefore relative. One can never be entirely sure that the information one saw was pro- or counter-attitudinal. It forms a crude division between pro- and counter-attitudinal arguments. Future research could use fictional topics to manipulate prior attitudes by insinuating public support for, or protest against a policy. This way a less crude measure of pro- and counter-attitudinal arguments can be constructed.
The findings of this study have great implications for a previously grim democratic future. In an offense against the Bayesian updating model, Lodge and Taber (2013) offer a pessimistic view of a “rationalizing voter” that doesn’t nuance its attitudes, regardless of counter-attitudinal evidence. This view is already limited by scholars noting that the cognitive dissonance of inaccurate beliefs at a certain point surpasses the cognitive dissonance of being wrong and leads citizens to alter their beliefs (Petty & Cacioppo, 1986a, Redlawsk, Civettini, & Emmerson, 2010). The current study contradicts the pessimism by Lodge and Taber (2013) to some extent. The findings discriminate between a part of the population that uses cognitive defense mechanisms to preserve and strengthen previous beliefs, i.e. high NTE individuals, and a part of the population for which the Bayesian updater is a better representation than the motivated reasoner, i.e. low NTE individuals.
Looking at the motivated reasoners, high NTE individuals are the citizens with most clearly developed attitudes. Considering high NTE individuals are more likely to engage in political activism, to vote and to use political media (Bizer et al., 2004) it is mere logic that they would counterargue counter-attitudinal arguments more than pro-attitudinal arguments. As they are more likely to be politically engaged, they are more likely to have heard several
pro and con arguments and to have learned the counter arguments as well. For the same reason, it is also not very surprising that they don’t adjust their attitudes that easily: their attitudes are formed more extensively. Therefore, it is also not desirable that high NTE individuals’ attitudes are as easily changed as low NTE individuals. Because their attitudes are extensively constructed, the rational thing to do is to preserve them.
Turning to the Bayesian updaters, these are low NTE citizens. These are citizens who are generally not that interested in politics (Bizer et al., 2004) and are not restricted by a specific ideology (Federico & Schneider, 2007). Therefore low NTE individuals have little attitude to defend and their attitudes are changed relatively easily by both pro- and counter-attitudinal messages. Because low NTE individuals’ attitudes are not carefully constructed, the rational thing to do when confronted with counter-attitudinal information is to adjust them. Concluding, the roles of both the high NTE motivated reasoner and the low NTE Bayesian updater fit the notion of a ‘rational citizen’ quite well.
How does this new picture of the rational citizen fit into democracy? The representative democracy, as is common in contemporary Western society, implies
ideological pluralism, or the possibility to choose between different ideologies. Accordingly, the metaphorical ‘marketplace of ideas’ suggests several salesmen trying to convince the group of potential buyers that they offer the best product. In other words, there should be two groups in society: salesmen and potential buyers. In order for the best product to prevail in a marketplace, you need a rational and critical group of buyers. These are represented by the Bayesian updaters of political information: the low NTE individuals. But ideas shouldn’t be abandoned too quickly either. In order to offer a wide variety of ideologies, salesmen are needed that stick to their beliefs. These are represented by the motivated reasoners: the high NTE individuals. Concluding, there is a place for both the Bayesian updater and the motivated reasoner in democracy. They both serve an important purpose. The one by rationally updating
their beliefs, and the other by rationally holding on to what they believed before. Accordingly, this study shows that individual differences in personality serve an important function in shaping citizens’ behavior in a democracy and thereby the functioning of democracy itself.
Literature
Albahari, M. (2015). Europe’s refugee crisis. Anthropology Today, 31(5), 1-2.
Arceneaux, K., & Vander Wielen, R. J. (2013). The effects of need for cognition and need for affect on partisan evaluations. Political Psychology, 34(1), 23–42.
http://doi.org/10.1111/j.1467-9221.2012.00925.x
Baron, R. M., & Kenny, D. A. (1986). Moderator-mediator variables distinction in social psychological research: conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51(6), 1173–1182.
Belanger, J. J., Kruglanski, A. W., Chen, X., & Orehek, E. (2014). Bending perception to desire: Effects of task demands, motivation, and cognitive resources. Motivation and Emotion, 38, 802–814. http://doi.org/10.1007/s11031-014-9436-z
Bizer, G. Y., Krosnick, J. A., Holbrook, A. L., Wheeler, S.C., Rucker, D. D., & Petty, R.E. (2004). The impact of personality on cognitive, behavioral, and affective political processes: The effects of need to evaluate. Journal of Personality, 72(5), 995-1028. Bobbio, N. (1987). The Future of Democracy. Cambridge: Polity Press.
Cacioppo, J. T., & Petty, R. E. (1982). The need for cognition. Journal of Personality and Social Psychology, 42, 116–131.
Cacioppo, J. T., Petty, R. E., Feinstein, J. E., & Jarvis, W. B. G. (1996). Dispositional differences in cognitive motivation: The life and times of individuals varying in need for cognition. Psychological Bulletin, 119, 197–253.
Cacioppo, J. T., Petty, R. E., & Kao, C. F. (1984). The efficient assessment of need for cognition. Journal of Personality Assessment, 48, 306–307.
Coronel, J. C., Duff, M. C., Warren, D. E., Federmeier, K. D., Gonsalves, B. D., Tranel, D. & Cohen, N. J. (2012). Remembering and voting: Theory and evidence from amnesic patients. American Journal of Political Science, 56(4), 837-848.
De Keijser, J. W., Van der Leeden, R. & Jackson, J. L. (2002). From moral theory to penal attitudes and back: A theoretically integrated modeling approach. Behavioral Sciences and the Law, 20, 317-335.
Ditto, P. H., & Lopez, D. F. (1992). Motivated skepticism: Use of differential decision criteria for preferred and non-preferred conclusions. Journal of Personality and Social
Psychology, 63, 568-584.
Federico, M., & Schneider, C. (2007). Political expertise and the use of ideology: Moderating effects of evaluative motivation. Public Opinion Quarterly, 71(2), 221–252.
Festinger, L. (1957). A Theory of Cognitive Dissonance. Stanford, CA: Stanford University Press.
Gaines, B. J., Kuklinski, J. H., Quirck, P. J., Peyton, B. & Verkuilen, J. (2007). Same facts, different interpretations: Partisan motivation and opinion on Iraq. The Journal of Politics, 69(4), 957-974.
Gerber, A., & Green, D. P. (1998). Rational learning and partisan attitudes. American Journal of Political Science, 42(3), 794-818.
Haugtvedt, C. P., & Petty, R. E. (1992). Personality and persuasion: Need for cognition moderates the persistence and resistance of attitude changes. Journal of Personality and Social Psychology, 63(2), 308-319.
Jacobs, D. & Carmichael, J. T. (2001). The politics of punishment across time and space: A pooled time series analysis of imprisonment rates. Social Forces, 80(1), 61-91. Jacobson, G. C. (2010). Perception, memory, and partisan polarization on the Iraq War.
Political Science Quarterly, 125(1), 31–56.
Jarvis, W. G. B., and Petty, R. E. (1996). The Need to Evaluate. Journal of Personality and Social Psychology, 70, 172–94.
Kruglanski, A. W., & Freund, T. (1983). The freezing and unfreezing of lay-inferences: Effects on impressional primacy, ethnic stereotyping, and numerical anchoring. Journal of Experimental Social Psychology, 19, 448–468.
Kunda, Z. (1990). The case for motivated reasoning. Psychological Bulletin, 108(3), 480–498. http://doi.org/10.1037/0033-2909.108.3.480
Lebo, M. J., & Cassino, D. (2007). The aggregated consequences of motivated reasoning and the dynamics of partisan presidential approval. Political Psychology, 28(6), 719–746. http://doi.org/10.1111/j.1467-9221.2007.00601.x
Lodge, M., Steenbergen, M., & Brau, S. (1995). The responsive voter: Campaign information and the dynamics of candidate evaluation. American Political Science Review, 89(2), 309-326.
Lodge, M., & Taber, C. S. (2000). Three steps toward a theory of motivated political reasoning. In A. Lupia Jr., M. D. McCubbins, and S. L. Popkin (Eds.), Elements of Reason. New York: Cambridge University Press, 183–213.
Lodge, M. & Taber, C. S. (2013). The Rationalizing Voter. New York: Cambridge University Press.
Maio, G. R., & Esses, V. M. (2001). The need for affect: Individual differences in the motivation to approach or avoid emotions. Journal of Personality, 69, 583–615
Nir, L. (2011). Motivated reasoning and public opinion perception. Public Opinion Quarterly, 75(3), 504–532. http://doi.org/10.1093/poq/nfq076.
Nyhan, B., & Reifler, J. (2010). When corrections fail: The persistence of political misperceptions. Political Behavior, 32(2), 303-330.
Petty, R. E., & Cacioppo, J. T. (1986a). The elaboration likelihood model of persuasion. In L. Berkowitz (Ed.), Advances in Experimental Social Psychology (Vol. 19, pp. 123-203). New York: Academic Press.
Petty, R. E., & Cacioppo, J. T. (1986b). Communication and Persuasion: Central and Peripheral Routes to Attitude Change. New York: Springer-Verlag.
Redlawsk, D.P., Civettini, A.J.W., & Emmerson, K.M. (2010). The affective tipping point: Do motivated reasoners ever “get it”? Political Psychology, 31(4), 563-593.
Rudolph, T. J., & Popp, E. (2007). An information processing theory of ambivalence. Political Psychology, 28, 563–585
Taber, C. S., & Lodge, M. (2006). Motivated skepticism in the evaluation of political beliefs. American Journal of Political Science, 50(3), 755–769. http://doi.org/10.1111/j.1540-5907.2006.00214.x
Tetlock, P. E., & Kim, J. I. (1987). Accountability and judgment processes in a personality prediction task. Journal of Personality and Social Psychology, 52(4), 700–709. Van der Brug, W., & Van Spanje, J. (2009). Immigration, Europe and the ‘new’ cultural
Appendix 1: Pretest stimulus material
To ensure that results are not caused by differences in the perceived strength of the arguments made in the left-wing and right-wing stimulus material, a pretest was used to find out the strengths of different arguments. In order to do so, an array of 5 left-wing and 5 right-wing arguments for both experiments were composed and introduced to 22 respondents. Moreover, the respondents rated their political stances on a 10-point scale from left-wing to right-wing and were divided into left-wing (< 6) and right-wing (> 5) respondents. The respondents were asked to rate the different arguments on a 7-point Likert scale from weak to strong for
argument strength. To ensure that the effects in the pretest are not caused by motivated reasoning themselves, the respondents were specifically instructed to leave their moral judgments of the arguments out of this rating. Several respondents expressed their difficulty with this task and also several strategies of how they proceeded to manage, indicating that respondents actually did explicitly try leave their moral judgments out of the task.
The results indicated that the different texts showed many different perceived argument strengths for the refugee experiment stimulus material. A two-way repeated measures ANOVA with the two most similarly rated arguments for the refugee stimulus material as the within subjects factor and respondent ideology (left-wing/right-wing) as a between subjects factor indicated no significant differences between the argument strength of the left-wing and right-wing stimulus arguments, F ( 1, 20) = .75, p = n.s. Moreover, it showed no significant interaction effect with respondent ideology F (1, 20) = .23, p = n.s. It can therefore be concluded that there is no difference in argument strength between the right-wing stimulus argument (M = 3.27, SD = 1.88) and the left-right-wing stimulus argument (M = 4.05, SD = 2.10) for the refugee experiment.
The results also indicated perceived argument strengths differed strongly for the different punishment experiment stimulus arguments. Again, a two-way repeated measures
ANOVA with the two most similar arguments as the within subjects factor and respondent ideology as a between subjects factor indicated no significant differences between the argument strength of the left-wing and right-wing stimulus arguments, F ( 1, 20) = .04, p = n.s. Moreover, it showed no significant interaction effect with respondent ideology F (1, 20) = .14, p = n.s. It can therefore be concluded that there is no difference in argument strength between the right-wing stimulus argument (M = 4.77, SD = 1.67) and the left-wing stimulus argument (M = 4.73, SD = 2.07) for the punishment experiment. The full texts of the
