VALENCE FRAMING OF POLITICAL PREFERENCES AND RESISTANCE TO PERSUASION

This study tested the "valence framing effect": an assumption that negatively conceptualized attitudes (as opposing the non-preferred alternative) are more resistant to later persuasion attempts. In the experiment we created choice between two political candidates and experimental subjects were led to conceptualize their political preferences in one of two possible ways: either as supporting the preferred candidate or as opposing the non-preferred candidate. The data indicate that negative preferences show less overall change when exposed to counterarguments. This finding can be incorporated in two theoretical frameworks: dual process theories of attitude change (Elaboration likelihood model) and descriptive decision making theories (Prospect theory). Results are discussed for their implications for the efficacy of political communication.


INTRODUCTION
Decades of intensive study of judgment, reasoning and decision making unfailingly indicate that people err often and in a systematic manner -they perceive and remember selectively, give importance to irrelevant information, make conclusions that are not based on probability.As a matter of fact, these errors are aberrations from the normative system of propositional logic and aberrations from the expected rational behavior.Furthermore, there is an increasing body of evidence indicating that these errors are not idiosyncratic and random but rather common and systematic.This is exactly why experimental studies often focus on so called cognitive biases.
Theoretical and methodological bases of this research stem from two different research traditions that have largely developed independently from each other, but nevertheless have many points in common.One tradition is research on attitude's resistance to change.The other is research on principles of judgment and decision making, or more specifically: the effect of question framing on decision making.
Research on persuasion and attitude's resistance to change are essentially two opposing faces of the same issue.Clear distinctions between strong and weak attitudes are defined by the criteria of perseverance -strong attitudes are those surviving the test of time and are resistant to change; and -influence -relating to attitudes' ability to strongly affect cognition and behavior (Mc Guire, 1966;Chaiken & Eagly, 1993;Krosnick & Petty, 1995).Over the years, many factors affecting attitude strength and its resistance to persuasion were extracted experimentally (further elaboration of attitude's object, repetition of persuasive message, and additional knowledge about attitude's object...).
Another research direction, related to notion of framing has originated in the context of research on decision making.Until publication of the seminal paper of Kahneman & Tversky (1979) this field was dominated by a normative subjective expected utility theory (SEU), which was entirely based on the assumption of human rationality in decision making.According to this theory, decision making is defined as establishing a relationship between the probability that a chosen alternative would lead to a desired outcome and the benefit that is expected from this outcome (for details, see Eiser & van der Pligt, 1993;Manktelow, 1999).All that is needed is to combine the information about the probability of a certain event ("What is the probability that I will find something interesting on TV?"; "What is the probability that it will rain today?";"What is the probability that I will keep this apartment or find a better one, in the future?") with information about wishes and interests, or the outcomes of the decision ("How much do I really want to watch TV?"; "How much will I be bothered by taking an umbrella with me?", "Do I really need an apartment now?").Accumulated empirical evidence has questioned the ability of this theory to describe genuine decision making in humans.Historically, one of first major criti-cisms of subjective expected utility theory was made in the 1950's (Simon, 1955, cf. Holyoak & Morrison, 2005), and is based on limited human capacity for information processing.According to this criticism, it would be unrealistic to expect that people thoroughly calculate the expected utility value of each and every possible alternative, having in mind limitations imposed by their memory, attention span and time constraints.In reality, human thinking does not meet the requirements of this formal standard.This formal system of decision making would set a task that is too demanding for human mind.Furthermore, there are arguments supporting a thesis that requirements of this formal standard can not be achieved even in principle.Therefore, if we define rational thinking as thinking that is an absolute compliance with the abstract normative system "we are doomed to irrationality even before the judgment has commenced" (Manktelow, 1999).Discovery of subjective expected utility theory's inadequacies has stimulated research interest in how people actually do make decisions.A new approach to decision making was recognized as descriptive approach to decision making, and is best known for the Prospect theory of Kahneman and Tversky.In spite of absence of an universal agreement on how to define rationality, it is still possible to formulate two necessary criteria of rationality -coherency and consistency (Tversky & Kahneman, 1981;Evans & Over, 1996;Kahneman et al, 1991).From these criteria, it follows that a judgment based on same logical elements should be (a) independent from the way of their presentation; and (b) stable over time.
Tversky and Kahneman performed several studies in order to investigate the dilemmas that experimental subjects are facing when systematically making judgments that fail to meet the rationality criteria, and the conditions backing this failure.According to those who believe that rationality is the main postulate of decision making, when conditions of a problem are kept constant, varying the observation point should not affect decision making.A typical experiment of Tversky and Kahneman shows that exactly opposite is the case.Depending on whether subjects are told that the program A certainly saves 200 lives or certainly kills 400 people (out of 600), they will be more prone to accept or reject the program.The number of people that are either saved or killed is the same under both experimental conditions, the only difference being in either positive (saving lives) or negative (loosing lives) definition of the problem.With this experiment Tversky and Kahneman were the first to show that people make different decisions about equivalent but differently framed alternatives.This different framing of a problem (as either a gain or a loss) leading to afore mentioned peculiarities in human decision making is called decision framing.Most often, this notion relates to subject's perspective when approaching a problem.Decision framing is influenced by subjective factors such as attitudes, values, and personality traits but also by external factors such as framing of the problem.
Nevertheless, most everyday problems consist of options that can not be quantified.Effect of decision framing is displayed under these conditions, as well.An illustrative example is cited by Shafir (1993, cf. Manktelow, 1999).When choosing between the two people who are competing for foster parenthood, experimental subjects' decision will differ depending on framing of the question: who should be awarded or who should be denied foster parenthood.In both instances experimental subjects choose a person who possesses both extremely desirable and extremely undesirable personal characteristics, relative to another candidate for foster parenthood who possesses 'a personal profile of a typical parent'.Thus, depending solely on question's framing, most experimental subjects deny or award foster parenthood to the same person (for details see Kostić, 2006).
Framing effect on decisions relating to gains and losses can be considered as a special case of much broader phenomenon that is usually referred as valence framing (presenting of normatively equivalent information in either positive or negative fashion, i.e. framing alternatives in either positive or negative terms).There have been many studies dealing with this issue, extending over essentially different phenomena that were all covered by the general notion of valence framing.Thus, a more precise differentiation and categorization of these phenomena was needed.One possible categorization (Levin et al, 1998, cf. Kuvaas & Selart, 2004) differentiates among three types of valence framing, with different consequences on decision making and different underlying mechanisms: • Effects of framing of risky choices become manifest when readiness to take risk depends on positive or negative framing of potential outcomes (most often as a gain or a loss).• Effects of goal framing differentially affect the impact of persuasive message, depending on whether it highlights positive consequences of performing of a goal-directed action or negative consequences of not performing of the same action.
• Effects of attribute framing are displayed when evaluations of objects or events become more positive if their key feature, regardless of the context, is formulated in positive rather than in negative terms.For instance, employment rather than unemployment, observance of the law rather than breaking the law, and the like.This categorization encompasses many empirical studies using different information displayed to experimental subjects in order to demonstrate distinct framing effects.In order to establish different framing of problem situations, those situations were described in a different way (using either positive or negative terms) to experimental subjects.
Empirical findings reliably point at fundamental asymmetry in the way that people perceive gains and losses, pleasant and unpleasant events.This rule has been confirmed across variety of research areas: formation of impressions, decision making, and nonverbal communication.Experiments on perception and memory show that negative information is given more weight.Negative information is perceived easier and faster, and its memory loadings are higher.People attribute more importance and give more significance to events leading to negative consequences.When meeting another person we pay more attention and significance to her/his negative features rather than to her/his positive features (Fiske, 1980).In nonverbal communication, observers are more sensitive to signals bearing negative connotation relative to signals bearing positive connotation (Frodi et al, 1978).Simply stated, in our cognitive system "losses are more visible than gains" (Kahneman & Tversky, 1984, p. 348).
The logical next step in studying the effect of negative framing was to further refine variation in problem exposition: present the content of the problem in an identical manner, and activate positive or negative cognitive framing with an ensuing question.The question facing the researchers was whether thinking in either positive or negative terms about absolutely identical information may be responsible for formation of attitudes of different intensity and resistance to persuasion.
This dilemma was a starting point of a recent study by Bizer & Petty (2005) who studied the effect of framing on attitude's resistance to change when choosing between two political candidates.Experimental subjects formed their preferences about two political candidates based on the information provided.After that, half of the subjects were led to think about their preferred candidate in an affirmative way, and the other half was led to think negatively about their non-preferred candidate.It was shown that the latter were more resistant to counter arguments.
Following basic methodological approach of Bizer and Petty, the objective of this study was to investigate whether greater resistance to change can be achieved by an experimental treatment that is much simpler than any manipulation technique that have been used so far: simply by varying the framing of the question regarding the preferred political candidate.That is, the objective of this study was to investigate whether people who formulate their attitudes in negative fashion (in antagonistic terms) are less susceptible to persuading influence of information that criticize, challenge or question those attitudes.
Justification for this hypothesis originates not only from the experiment of Bizer and Petty, but also from ample empirical evidence that has been accumulated while testing dual process theory of attitude change (Petty & Caccioppo, 1986;Chaiken & Eagly, 1993).This theory also points at the effect of negatively conceptualized attitudes.Research on attitude change indicates that negatively conceptualized attitudes are acquired faster, are maintained longer and are more resistant to later persuasion attempts (Pratto & John, 1991).When learning about features of unknown objects people are more prone to make errors of the 'false negative' typeavoiding truly positive objects that they fear might be threatening -than to make errors of the 'false positive' type -approaching truly negative objects that may appear as friendly.Once learned, negatively conceptualized attitudes are easier to generalize (Fazio et al, 2004).Finally, experimental evidence shows that in order to shift somebody's attitude from negative to positive it is not enough to present her/him with 51% of positive attributes.Instead, at least 60% of positive attributes are needed (Petty & Wegener, 1998).

METHOD
Following exposure to newspaper articles about imaginary politicians, experimental subjects were asked to express their attitude towards the politicians and their programs as either approving or disapproving.In the next stage, the subjects were exposed to information questioning candidates' credibility.Finally, a possible change in subjects' attitude was examined.For this purpose, an experiment comprising a test-retest design was conducted.

Creating the message
In the first phase of this research, two texts complying with the usual newspaper style and format were composed.Each text presented one of the two candidates for the forthcoming local elections.The approximately same-length texts were based on political programs of Serbian political parties and their Internet presentations.The texts were presented to subjects as actual newspaper clippings of the candidates addressing their potential voters.Each text described the candidate, his personal characteristics, political program, intentions and objectives in such way that those differences between the candidates were clear and obvious.One candidate (Nikola Kovačević) was described as a conservative, right-wing politician, while the other (Aleksandar Radovanović) was described as the center-left liberal democrat.There was no mention of their respective political parties and their respective political orientations.Political priorities of the right wing candidate were as follows: strict enforcement of the Constitution and the law, control of public services, social welfare provisions for the needy, and increasing administrative taxes as a way to supplement municipal budget.Political priorities of the center-left politician were somewhat different: commercial loans for small and medium-size businesses, scholarships and student loans, supporting the NGO (non-governmental organization) sector and fighting grey economy (see Appendix 1 for the full-length versions of the texts).

Testing of the stimulus material
Preliminary testing of the stimulus material was done in order to determine whether the texts contained issues that were familiar and recognizable to a student population.For this purpose, 47 Psychology seniors were asked to classify actual party affiliation of the candidates after reading their political programs.The outcome of the classification is shown in Table 1.Political program of the center-left candidate Aleksandar Radovanović was perceived as belonging to either DS, G17+ or LDP political party.Political program of the right wing Nikola Kovačević was perceived as closest to political programs of either SRS or DSS.Informal analysis of the data indicated that the style of presentation and the issues covered in the texts are quite befitting the level of political involvement of Serbian student population.Obvious differentiation of political programs was confirmed by the fact that there was no overlap and no confounding in classification of party affiliation of imaginary political candidates, although six respondents failed/refused to give a definite answer or entered only 'left' or 'right' on their answer sheets.In order to further validate the stimulus materials, the students rated the texts by using three bipolar semantic differential scales (range 1 to 7).Opposing poles were defined as comprehensible/incomprehensible, relevant/irrelevant, and as same/different.The average rating for the comprehensible/incomprehensible scale was 6.06 (SD=0.9),and 6.27 (SD=1.1)and 5.34 (SD=0.8)for the relevant/irrelevant and the same/different scale, respectively.This indicates that the students have successfully understood the programs and that they were fairly successful in differentiating between them.
An information challenging candidate's credibility was used as a counterattitudinal message.In a supplementary text that was presented to experimental subjects, the candidate was blamed for corruption and illegal gain of substantial amount of money (full text attached in Appendix 2).The strength of the counter-attitudinal message was assessed in the preliminary phase, before it was used in the actual experiment using the same preliminary phase sample of Psychology students.They were asked to rate their support for each candidate on a 7-point scale before and after exposure to the counterattitudinal message.Average shift in the strength of their support (difference in the attitude's intensity before and after exposure to the counterattitudinal message) was 1.52 (F=4.58;p<0.01).The obtained difference supported our initial assumption about the adequate strength of the counterattitudinal message: an argument that was either too strong or too weak would preclude the ensuing comparison of attitude's resistance to change.In each instance, there would be no sufficient room for demonstration of the framing effect.In the literature (for details see Petty & Wegener, 1998), the attitude shift of 1.5-2.5 on an 11-point scale was considered as satisfactory.In the preliminary phase, we also learned about the approximate preference for the candidates in the student population which was later used when planning the sample size in order to obtain adequate statistical power.

Experimental design and subjects
The actual experiment included a gender-balanced sample of 120 freshmen from the Faculty of Philosophy and the Faculty of Electrical Sciences, Belgrade University.One half of the candidates were tested for their support for the candidate Aleksandar Radovanović, the other half for their support for the candidate Nikola Kovačević.
The subjects were presented with two texts, each text describing one of the two only candidates in the final round of the elections for the President of one Belgrade municipality (both elections and the candidates were fictional).Following exposure to the texts, the subjects were asked to express their preference for one of the two candidates.That is, one half of the subjects were asked whether they were for or against the candidate AR, while the other half of the subjects were asked whether they were for or against the candidate NK.The 2x2 experimental design used in this study (Table 2) was defined by two independent variables: framing of the question (either positive or negative) and the preference for the candidate (either AR or NK).Within each group preferring either one of the candidates, half of the subjects have expressed their attitude through their support for the preferred candidate and the other half through their opposition to the other candidate.

Subjects' responses Subjects' attitudes
Following their declaration of preference (for or against the candidate), the subjects expressed the intensity of their attitude using a bipolar 11-point scale with statements "1 = I completely oppose" and "11 = I completely support" at its opposing ends.Although the use of an 11-point scale is potentially too complex for the subjects, and the assessments obtained on a 11-point scale are highly correlated with assessments obtained on a 7-and 5-point scales, we have decided to use it in order to make our data directly comparable with the data from the study of Bizer & Petty (2005).This decision was further supported by the fact that or experimental sample consisted of young and educated people who showed no difficulty evaluating on a 11-point scale.
In the second phase of the experiment, the subjects preferring the candidate A were exposed to counterattitudinal message that was presented to them as the continuation of the original newspaper article.This new article spoke about A's involvement in corruption and the criminal investigation against him.An analogue procedure was used on the subjects preferring the candidate B (Figure 1).Following exposure to the counterattitudinal message, the subjects were again asked to express their attitude about the respective candidate using the same 11-point scale.The dependent variable was defined as the difference between the two scores (before and after exposure to the counterattitudinal message) on the 11-point scale.

Procedure
The study was supported3 by computer facilities of both Faculties.Since the subjects freely chose their preferred candidate and since they spontaneously preferred Aleksandar Radovanović, the study was continued until the number of subjects who spontaneously preferred the less preferred candidate Nikola Kovačević, reached 60.Eventually, there were 88 subjects with an initial preference for Aleksandar Radovanović and 60 subjects with an initial preference for Nikola Kovačević.Average length of the experiment was 11 minutes.In order to maintain a balanced 2x2 experimental design, with 30 subjects per group, 28 subjects with an initial preference for Aleksandar Radovanović were randomly excluded from data analysis.Following completion of the experimental procedure, the experimenter explained that candidates' profiles were designed solely for the purpose of the experiment and thanked the subjects for their participation.

RESULTS
Difference in attitude change following exposure to the counterattitudinal argument was analyzed using 2x2 ANOVA.The subjects that preferred candidate (A or B) through opposition to the non-preferred candidate (A or B) were significantly less susceptible to attitude change (M=1.26,SD=1.69) relative to experimental subjects who were led to positively frame their attitudes; that is, as a support to their preferred candidate (A or B) (M=3.61,SD=2.59).This was evidenced by statistically significant main effect of factor "framing" (F /1, 116/=14.169,p<.001).Both the effect of the experimental factor "candidate" " (F /1,116/=0.44,p=.833) and the candidate x framing interaction (F/1,116/=3.47,p=.065) failed to reach statistical significance at the p = 0.05 level, thus supporting the conclusion that framing did not have any differential effect on subjects' preference of the candidates.That is, the effect of counterattitudinal message was object-independent.

Attitude change
Standard deviations from mean attitude change scores using 11-point scale.Counterattitudinal message effected all four experimental groups' attitudes, but the shift was significantly larger in the "negative frame" groups.
The data confirm previous findings of Bizer & Petty (2005) that leading people to negatively conceptualize their attitudes (as opposing the non-preferred alternative) is sufficient to induce greater resistance to later persuasion.Therefore, it is safe to conclude that individuals whose political preferences -when choosing between the two candidates -are framed as opposition to the non-preferred candidate, are less susceptible to change their attitude when facing counterattitudinal arguments.

DISCUSSION
Since publication of pioneering papers of Tversky and Kahneman (Tversky & Kahneman, 1981;Kahneman et al, 1982) numerous experiments have shown that presentation of normatively equivalent information in either positive or negative fashion can systematically bias decisions and actions of decision makers.On the other hand, many experiments suggested that negatively conceptualized attitudes are more resistant to subsequent attempts of persuasion.Nevertheless, the processes underlying both phenomena are neither well understood nor conceptualized theoretically.
Different research traditions converge on basically offer similar assumptions about processes underlying the effect of valence framing.
An explanation originating from the decision making theory, assumes that question framing differentially affects subjects' selective search for information confirming their decision.Under one scenario they should confirm their decision about support of the preferred candidate.Under the other, they should confirm their decision about opposition to the non-preferred candidate.This will differentially affect their bias for information that is either confirming or refuting candidate's competencies or credibility.
Furthermore, most current models of information processing consider discrepancy between the desired (expected) and the real state of affairs as an activator for more rigorous and more systematic cognitive involvement.Proponents of the evolutionary approach (Kuvaas & Selart, 2004) claim that people have an inborn tendency for increased cognitive involvement and more thorough and more careful situation analysis when they encounter a negative information, since it may signal the discrepancy between the expected and the actual state of the affairs and also the approaching of undesired and threatening events.On the contrary, encountering positive information makes people less focused and less prone to systematic information processing.
Similar conclusions are made by researchers working within the context of the affective-cognitive model.The negative affect is viewed as a carrier of information that the actual situation is challenging and calls for a solution; therefore it stimulates attention and more detailed and more systematic information processing.On the other hand, the positive affect informs us that 'everything is OK' and that there is no need for greater cognitive involvement (Holyoak & Morrison, 2005).
Regardless of the precise nature of the mechanism underlying more systematic and more careful processing of negative information (and similarly, negatively framed information), there is much evidence that people receiving negatively framed information display higher degree of analytical thinking and better memory relative to people receiving positively framed information.
What happens next with an attitude that was formed following more thorough cognitive processing?Bizer and Petty (2005) hypothesize that people are more confident in negatively conceptualized attitudes simply because of more involving cognitive processing.This may, in turn, lead to greater resistance to later persuasion.According to them, two distinctive phases are involved in the process: (a) people perceive negatively conceptualized arguments in a different manner, and because of that (b) differentially process counterattitudinal messages.In the present experiment this means that experimental subjects who were asked 'whether they oppose one of the candidates' devote more cognitive effort to recognize and formulate their arguments relative to experimental subjects who were asked 'whether they support one of the candidates'.Thus, the former group (the "opponents") acquires a more solid attitude base which is better prepared to resist the counterattitudinal message.There are further assumptions about what happens in the next phase.Elaboration likelihood model (Petty& Caccioppo, 1986) proposes three different mechanisms that lie beneath greater attitude's resistance during the second phase (for details see Žeželj, 2005).It is possible that people use degree of their confidence in an attitude as a 'rejection sign' and hence a priori reject any relevant subsequent message, without any further cognitive elaboration.Another possibility is that people with negatively conceptualized attitudes pay less attention when processing counter-attitudinal messages.Finally, one can assume that people with negatively conceptualized attitudes attentively process the messages, but that their attitude is based on more information.This, in turn, enables them to be more resistant to counterattitudinal messages.
Attitude's resistance to change is only one component of its strength.Therefore, further research is needed in order to establish whether the experimental manipulation used in this study results in resistance to persuasion that is stable over time, and to what extent the observed group differences in resistance to change translate into behavior.
An obvious question that will hopefully be resolved in future studies concerns the relationship between the two types of framing.On one hand, it is possible to manipulate the subject-derived framing of the attitude -as was the case in this study.On the other hand, it is possible to manipulate the framing of the counter-attitudinal message: to use negative information about the preferred candidate or to use positive information about the non-preferred candidate.In specific, one group of experimental subjects could be presented with a text similar to the text used in this study, going against their preferred candidate.Another group could be presented with a text supporting the non-preferred candidate.Based on existing evidence, one could assume that discrediting will have a more powerful effect on the attitude change.
Findings reported here indicate two important things.First, negation is stronger than affirmation.Negatively conceptualized attitudes are more robust and more resistant to change.It seems that negative designations dominate organization of our preferences.It is so much easier to make a list of things that bother us, of things that we don't like and things that we oppose than to make a list of things that we like, and things that we prefer.Sometimes only one "for" implicates a whole series of "against".In many situations it is difficult to recognize positive choices.After a fall of an authoritarian regime, the citizens often have a problem to switch from an 'against' frame of mind to an 'affirmative' frame of mind and to use this mental shift when choosing their political representatives.
Another important finding is that it is relatively easy to elicit either a negative or a positive attitude frame, simply by differentially framing the question about the attitude's object.Broader implications of these findings are twofold, at least.We should start with responsibility of those who broadcast information.It seems that defining freedom of speech in terms of availability of information does not suffice.The most important political information is conveyed to us through different channels, and each of those channels adds its own 'twist' to the information.Here, we have shown that even very fine variations in 'twisting' significantly modulate message effect and consequently, decision making.Another question is whether affirmative thinking can be enhanced and taught?Whether systematic looking for reasons 'for' can strengthen positive attitudes?Similarly, are we capable to recognize and restrain our biases in order to make some kind of balance.Most cognitive scientists, probably because of their cognitive orientation, support this idea about controlled biases (Wilson & Keil, 1999;Quattrone & Tversky, 2000).The question is not any more whether people are always or ever rational when making decisions, but whether they recognize when they are not and how does that affect the quality of decisions that they make."…The candidate Nikola Kovačević/Aleksandar Radovanović was subjected to criminal investigation in 2001 under suspicion for corruption.Allegedly, he was illegally granting building permissions for numerous objects that have been built in the municipality, thus amassing an illegal financial gain estimated at 30,000 euros.However, there was not enough evidence for criminal indictment, all charges were dropped and further investigation was terminated…" REZIME UTICAJ NAČINA IZRAŽAVANJA POLITIČKIH PREFERENCIJA NA NJIHOVU OTPORNOST

Figure
Figure 1.Experimental procedure

Figure 2 .
Figure 2. The span of attitude change in four experimental groups