FRAMING AND AGENDA-SETTING 0 Framing and Agenda-Setting: Two Parallel Processes in Interaction František Kalvas, Jan Váně, Martina Štípková University of West Bohemia in Pilsen Department of Sociology Univerzitní 8 306 14 Pilsen Czech Republic +420 775 640 158 / +420 377 635 663 kalvas@kss.zcu.cz ; vanejan@kss.zcu.cz ; marsti@kss.zcu.cz Martin Kreidl Masaryk University Faculty of Social Studies Joštova 10 602 00 Brno Czech Republic kreidlm@fss.muni.cz František Kalvas is a head of the Department of Sociology at the University of West Bohemia in Pilsen. His research interests include public opinion, media effects, and survey methodology. Jan Váně is an Assistant Professor at the Department of Sociology at the University of West Bohemia in Pilsen. His research interests focus sociology of religion, and social theory. Martina Štípková is a Ph.D. student at the Faculty of Social Sciences, Charles University at Prague and lecturer at the Department of Sociology at the University of West Bohemia in Pilsen. Her research interests are social stratification, family, and health. Martin Kreidl is an Associated Professor at the Faculty of Social Studies, Masaryk University in Brno. His research has concerned social inequalities and mobility, social demography, sociology of health, and quantitative applications in social sciences. FRAMING AND AGENDA-SETTING 1 Abstract We interconnect the framing and agenda-setting theories of mass-communication effects in this paper. We postulate that the framing process creates conditions for the agenda-setting process. We argue that differently framed news have different effect in the agenda-setting process. We hypothesize that issue-specific frames, episodic frames, and value frames have a stronger agenda-setting effect. We suggest an explanation of the role of frames in the agendasetting process through the theory of cognitive dissonance. We use matched panel survey data and media content analysis regarding one particular issue – church restitutions. We show that indeed differently framed news have distinctive effects on setting the personal agenda. Some frames have a strong positive effect, some slightly negative effect, while some other have no effect. Key words: framing; agenda setting; cognitive dissonance; panel data; quantitative content analysis. FRAMING AND AGENDA-SETTING 2 Framing and Agenda-Setting: Two Parallel Processes in Interaction Agenda-setting The agenda-setting theory describes a process in which society prioritizes through coming to a consensus as to which issues should be dealt with primarily. It is logical that society first allocates resources for dealing with these priority issues. The term issue means “a conflict between two or more identifiable groups over procedural or substantive matters relating to the distribution of positions or resources” (Cobb & Elder, 1983, p. 32). All issues are a conflict, which can be recognized on the following three levels: 1) whether the issue even exists, 2) whether the issue should be solved and 3) how the issue should be solved. But not every conflict becomes an issue. For a conflict to become an issue it has to be identified as such and there must be a requirement for its solution (Dearing & Rogers, 1996). A set of issues hierarchically arranged according to their importance is called an agenda. An agenda changes dynamically in time – either because the issue is solved or its most thorny stage is warded off, or because society got used to the originally thorny condition (i.e. it is no longer perceived as urgent), or because other issues occurred which were more urgent. An agenda is not distinguished only on the level of the entire social system – its subsystems have their own agendas as well. When deconstructing the whole system, we can get all the way to an individual who has own agenda – we call it a personal (or intrapersonal) agenda (McLeod, Becker, & Byrnes 1974/1991). In terms of agenda-setting theory, three subsystems are crucial: media, public, and policy. All three have their own agendas among which the theory recognizes a dense network of mutual interrelationships. These agendas are also influenced by events in the non-mediated world, personal experience, and by interpersonal communication. The theory also fragments crucial subsystems and describes how agendas of subsystem parts mutually influence one another, and how a consensus concerning the agenda of the whole subsystem is created FRAMING AND AGENDA-SETTING 3 (Dearing & Rogers, 1996). In this text, we only focus on the relation of public and media and we omit the rest of the system. We perceive the media and public as one unit and do not structure these subsystems any further. Clarification is still required on how the agenda-setting theory determines the position of issues in the media agenda or public agenda. The intensity of attention is the common denominator for indicating the position of an issue on any agenda. Attention by the media is usually understood as the number of news dealing with the respective case. Attention by the public is usually understood as the proportion of the public that considers the issue to be one of the most important (Dearing & Rogers, 1996). If research works with data on the microlevel (as we do), we talk about personal agenda that is the agenda of an individual reflecting what this individual considers to be a significant issue (Kalvas 2009a). This is how we will understand the terminology as well – see methodology. Framing The framing theory lies on the borderline of three paradigms concerning media studies: cognitive, critical, and constructivist (D’Angelo, 2002; Matthes, 2007). The framing theory describes how the complexity of issues that are difficult to understand can be reduced. This theory claims that news contains not only plain facts but also a value frames that helps us understand these facts. According to Entman (1993) the frame contains four fundamental components “problem definition, causal interpretation, moral evaluation, and/or treatment recommendation” (p. 52). Entman’s definition is what we stem from as well. Entman (1993) states four locations where we can find frames “the communicator, the text, the receiver, and the culture” (p. 52). Framing must be perceived as a process in which frame-building happens first, followed by frame-setting, and this process has its consequences as well (de Vreese, 2005). Frame-building takes place mainly in the interaction between journalists, elites and social movements, and “the outcomes of the frame-building process are FRAMING AND AGENDA-SETTING 4 the frames manifest in the text. Frame-setting refers to the interaction between media frames and individuals’ prior knowledge and predispositions” (de Vreese, 2005, p. 52). In the framing process, many participants try to force their own perspective on the issue, “so that it resonates with their needs” (Trampota, 2006, p. 123). However, these participants have dissimilar capabilities to enforce their way of framing. Contemporary research deals intensively with the ability of media to set the frames in the minds of audience members and with the consequences that such frame-setting brings (Iyengar, 1991; Entman, 1993; de Vreese, 2005; Lee, McLeod, & Shah, 2008; Scheufele 2000). Interconnection of agenda-setting and framing McCombs and Shaw (1993) were the first to try connecting the theory of agendasetting and framing. They assumed that news carry information about an issue – this is how agenda-setting takes place. At the same time news emphasizes or conceals some attributes of the issue – this is how the framing of the issue takes place (McCombs & Shaw, 1993). The effort of McCombs and Shaw is disproved by Scheufele (2000) who claims that the frame is a complex pattern and guideline for a causal interpretation which can be substituted only through offering an alternative complex pattern. Today, the effect described by McCombs and Shaw is better indicated by the terms second level of agenda-setting or attribute agendasetting (e.g. McCombs, 2004). Now McCombs (2004) understands the frame according Entman’s definition, but still insists the frame is an attribute of a respective issue. Lecheler, de Vreese, and Slothuus (2009) work with a different idea: they consider the issue importance to be the moderator of the framing process. They measure individual evaluation of the issue importance (i.e. personal agenda) before participants are exposed to the manipulated impulses – variously framed messages. Their results show that if an issue is important, it blocks the possible effect of framing, while if an issue is not very important we can observe a strong effect of the framing process. The evaluation of importance definitely FRAMING AND AGENDA-SETTING 5 does not lead to framing. The authors only declare that if we consider an issue to be not very important, we incline to change frame if there is an appropriate impulse from outside. Our solution respects the Scheufele’s critique and finds inspiration in the approach of Lecheler et al. (2009). In our opinion, the framing process creates conditions for the agendasetting process. We do not argue that the framing process leads directly into the agendasetting process. In agreement with Scheufele (2000), we are convinced that these are cognitively different and parallel processes. At the same time, we believe that although the processes are separate, they interact whereby one creates conditions for the other. We presume that if some frames establish themselves in public discourse, the agenda-setting of the given issue is more successful than if some other frames do. In our opinion, different frames have a different ability to support the process of agenda-setting. This idea is in consensus with McCombs (2004) proposing that “certain ways of describing an object may be more compelling than other ways in creating object salience among the public” (p. 93). Cobb and Elder (1983) emphasize the key role of a problem definition for a public support of an issue. A similar idea to ours was given already by Wanta and Hu (1993), but we believe they expressed too strong a hypothesis that frames determine the effect of agenda-setting. To theoretically explain the relationship between framing and agenda-setting, we apply theory of cognitive dissonance (Festinger 1957/1962). If a person is exposed to frames that are not in compliance with his/her cognitive elements (knowledge, opinion, and reflection of own behavior), it causes a state of cognitive dissonance. The extent of cognitive dissonance is directly proportional to the number and importance of cognitive elements that are in contradiction with the frame. The pressure to reduce the present dissonance is then directly proportional to the extent of cognitive dissonance. A substantial way of reducing this pressure is drawing on new information or a confrontation with new opinions.1 To simplify the explanation of cognitive dissonance, we relate Festinger’s theory to frames. Where we use the term “frame” in this part of explanation, Festinger would use a general term “cognitive element”. 1 FRAMING AND AGENDA-SETTING 6 In other words, we think the frames in media messages could evoke a cognitive dissonance that motivate more intensive reception of information regarding the issue and intensified reception leads to setting the issue on the personal agenda. Iyengar and Kinder (1987) clarify the relationship between the intensity of information reception and importance of the issue by the memory-based model of information processing. The importance of an issue is connected with the easiness of recall and the amount of information we receive and store in our memory. The more information about issue we receive the easier we recall the issue and the more important the issue is for us. We should make several remarks on use of cognitive dissonance theory. The origin of dissonance is usually connected with a behavior (Cooper, 2007). But Festinger (1954/1999) originally assumes that information could evoke cognitive dissonance. Even in his canonical book, Festinger (1957/1962) does not exclude other sources of dissonance than behavior, though his examples concern only a behavior as a source of dissonance. Also in later work, Festinger (1962) highlights idea the dissonance evoked by behavior is only special case. Several empirical findings prove the cognitive elements could evoke cognitive dissonance as well (Dias, Oda, Akiba, Arruda, & Neuder, 2009; Donohew & Palmgreen, 1971). We are fully aware that the effect of one dissonant piece of news is incomparable with the effect of dissonant behavior. But we think a bigger sum of dissonant news could evoke cognitive dissonance, at least in weaker form. We also think the plain facts in the news are not only the source of dissonance but also the frames of the news. Frames that invoke cognitive dissonance in audience members stimulate the attention given to an issue, since it is unimaginable to discuss the frame and not to mention the issue itself at the same time. Through coping with cognitive dissonance (i.e. drawing on new information), a person strengthen the memory traces to the framed issue. And doing so, s/he increases the chance to set the issue on the personal agenda (according to the memory-based model). FRAMING AND AGENDA-SETTING 7 The theory distinguishes several typologies of frames and we will demonstrate how different types of frames should differ in the effect they have on the agenda-setting process. Iyengar (1991) defines thematic and episodic frames. Thematic frames provide a deeper insight into the context, emphasize general trends, and describe mainly the process. Therefore, they lead to attributing responsibility to society. Episodic frames focus on concrete events, and situations and they lead to attributing responsibility to individuals. We assume that episodic frames are more likely to invoke cognitive dissonance. Their more concrete and, hence, more seizable form is more likely to be in contradiction with individual’s cognitive elements. That is why we presume a more significant effect of episodic frames. H1: Thematic frames have a weaker agenda-setting effect than episodic frames.2 Frames can be also divided into issue-specific and generic. Issue-specific frames “are pertinent only to specific topics or events” (de Vreese, 2005, p. 54) while the generic ones are much more general. Generic frames “transcend thematic limitations and can be identified in relation to different topics, some even over time and in different cultural contexts” (de Vreese, 2005, p. 54). We distinguish five types of generic frames: conflict, human interest, economic consequences, morality, and responsibility (Semetko & Valkenburg, 2000). We anticipate a stronger effect of issue-specific frames because they focus the effect of invoked cognitive dissonance directly on the framed issue. Generic frames have the potential to focus attention on an issue different from the framed issue since they can be applicable to other such issues from the audience’s point of view, due to their generic character. H2: Generic frames have a weaker agenda-setting effect than issue-specific frames. The last typology is strategy and value frames. “A strategy frame organizes a policy conflict as a clash of political interests and competing strategies, typically highlighting political machinations of the contending parties, related to their objectives, strategies, and 2 Below we will provide a more detailed description and classification of frames used in our analysis. FRAMING AND AGENDA-SETTING 8 tactics” (Lee, McLeod, & Shah, 2008, p. 700). A value frame depicts policy conflict as a clash of moral principles or basic values where the parties enter the conflict on the basis of a different set of values. This type of frame is not very common, but when it occurs, it is “powerful and efficient in shaping audience reasoning processes, as values provide individuals with easily accessible heuristics that guide the understanding of complex policy issues without recourse to detailed information” (Lee, McLeod, & Shah, 2008, p. 701). Disillusion invoked by strategy frames should lead to an impression that the framed issue is not significant for a common member of audience. Thus, disillusion reduces the weight of cognitive elements (frame, issue, and framed information) and dilutes the degree of cognitive dissonance. Therefore, strategy frames inhibit agenda-setting effect. We anticipate a more substantial effect of value frames because they emphasize moral principles and values. H3: Strategy frames have a weaker agenda-setting effect than value frames History of the church restitutions case Before we describe the data with which we verify the validity of our hypotheses, we will explain the context of the issue that we chose – the dispute over church restitutions. Church restitutions were supposed to remedy the property injustices that the communist regime inflicted on churches. Already in the federal convention (1990-1992), there was the first unsuccessful attempt to enforce the restitution law that would enable the return of property to churches. Another attempt took place in 1996 and 1997 when the government passed a resolution concerning the performance of some so-called gratuitous transfers of the property belonging to churches, parishes, and some societies. The proposal for restitutional compensation, however, was not implemented. In 2004, there was a push from municipalities to solve the stalled situation, the socalled “Řečická výzva” (“Řečice Appeal”). It was submitted by mayors of municipalities that owned the property formerly belonging to churches. Such property is blocked by law, and FRAMING AND AGENDA-SETTING 9 municipalities are not allowed to manipulate it until a state-churches settlement takes place. In reaction to that, an agreement was prepared by the coalition committee in 2005, where churches would be paid rent totaling 50 million dollars plus inflation annually for the period of 50 years. But this agreement never became a government bill. Based on the government decree from January 2008, the Ministry of Culture of Czech Republic (MCCR) processed a bill on church restitutions that was approved by the government of the Czech Republic on 2nd April 2008. The bill counted on compensation for the confiscated property amounting to 7 billion dollars. Churches were supposed to be given a third of the confiscated property and the remaining two thirds were supposed to be compensated through a refund amounting to 4 billion, which would be paid out over a period of 60 years. In total, the proposed bill counted on the state paying approximately 12.5 to 13.5 billion dollars including interest. The bill was submitted for consideration in the Parliament of the Czech Republic. On 29th April 2008 the submitted bill on church restitutions did not pass the parliament. The discussion about the bill was postponed with the explanation that the “determination of scope of the given out property (listing or definition) and evaluation of the non-given out property, i.e. a determination of scope of the refund from state” (MCCR, 2008) needed to be filled in. This particular episode culminated on 6th May 2008 when the rebelling coalition member of parliament Vlastimil Tlustý refused to support this bill. The Chamber of Deputies established a temporary parliamentary committee that was supposed to evaluate and submit comments on the proposed bill. In May 2010, the activity of this committee was terminated together with the end of term of the respective Chamber of Deputies without any solution being adopted. One of the focusing events of the church restitutions case is the dispute over the ownership of the St. Vitus Cathedral. The core of the dispute is a polemic whether the statutory order from 1954 is valid, according to which the cathedral belongs to all Czech FRAMING AND AGENDA-SETTING 10 people. On 30th December 1992, the Devotional Array (Náboženská matice) submitted a declaratory action regarding the ownership of the St. Vitus Cathedral against the Office of the President of the Republic. Upon submitting this declaratory action, the dispute over the cathedral and the issue of unsolved church settlement gradually overlapped in the media. We mention this dispute mainly because the media attention paid to the case of church restitutions coincides significantly with the media attention paid to the dispute over the cathedral in the period that we studied (April – May 2008). Both cases coincide as well. On 30th April 2008, the Supreme Court definitively adjudicated the St. Vitus Cathedral to the state. That was a definite end to the dispute that had dragged on for more than 15 years. Data and variables The data file analyzed in this text originated through the merger of a panel survey of the Public Opinion Research Center (Centrum pro výzkum veřejného mínění [CVVM]) from the project “Public and Media Agenda”, quantitative content analysis of media performed by MediaTenor (CVVM, 2008), and our own content analysis of the corpus of news dealing with the issue of church restitutions. From the CVVM research, we use the information of mentioning the church restitutions as a significant all-society event and sociodemographic characteristics of respondents. From the content analysis of media, we use data on the media exposition concerning the topic of church restitutions and the dispute concerning the ownership of the St. Vitus Cathedral. Our own content analysis provided us with information on types of frames and the frequency of their occurrence in the corpus of news. In the panel survey, 648 Czech Republic citizens 18 years old or older were questioned in a 12 week period from 20th April until 6th July 2008. The sample was construed through a random sampling (see Vinopal, 2009). Respondents were instructed to fill out a questionnaire, always on Sunday at the end of the respective week, and then send it back by post. Within the questionnaire, the respondents’ task was to give two items every week that FRAMING AND AGENDA-SETTING 11 they considered to be “recently the most significant all-society events”.3 We created a dichotomous dependent variable “Mentioning of restitutions” which had a value of 1 if the respondent had an answer in the category “St. Vitus Cathedral” or “church restitutions” as one of the two most significant recent events. Since the answer in one wave of the survey can influence the answer in the following wave, we introduced a control variable “Previous mentioning of restitutions” which had a value of 1 if the respondent stated the restitutions or the dispute over the cathedral as one of the two most significant recent events in the previous wave of survey. The frequencies of these variables are stated in the Appendix 1. Why we constructed a joint variable for the mentioning of church restitutions and the dispute over the cathedral? The original codebook distinguishes these issues but in the whole (not yet reduced) set, the dispute over the cathedral was directly indicated only 24 times (7%) and church restitutions were indicated 319 times (93%). In our opinion, this noticeable disproportion shows that both the cases overlap in the respondents’ minds and, therefore, we created a joint variable in which we aggregate the evaluation of both issues as important. Sociodemographic characteristics of respondents include gender, age, education and religion. Respondents filled out these data only in the first wave and the values stayed the same for the whole research period. Gender is a dichotomous variable, male coded as 1. The variable age has 4 categories depending on which period of relation between the state and church the respondents experienced. The oldest category is for people 72 to 92 years. Another category is for people 52 to 71 years. The third age category consists of people who were 3151 years when the survey took place. The youngest age category consists of people aged 1830 years. The highest completed education is divided into 4 categories: basic and lower, vocational, secondary school certificate, and university. The original seven categories of religion are simplified into a dichotomous variable Christian vs. other. The value Christian 3 The question wording was: “Which recent all-society events do you personally consider to be most important?” FRAMING AND AGENDA-SETTING 12 religion includes respondents who acknowledged catholic, protestant or orthodox religion. The frequencies concerning categories of these variables are stated in the Appendix 2. Variables capturing the occurrence of church restitutions in news come from the content analysis of MediaTenor. This analysis monitored the content of nationwide dailies (Blesk, Hospodářské noviny, Lidové noviny, MF Dnes, and Právo), radio news (ČRo 1 Radiožurnál, and Impuls), and the main TV news (ČT1, TV Prima and TV Nova) from 24th March until 14th July 2008. The number of news in the categories “restitutions of church property” and “relation between church and state” were recorded on a daily basis. Since the dispute over the St. Vitus Cathedral ownership is sensu stricto not part of the church restitutions issue, neither can it be considered as the relation between the church and state, we supplemented the content analysis of MediaTenor with our own research. In the ANNOPRESS archive for the monitored media, we found all news concerning the dispute over the cathedral and we added their number from the respective days into the data set. If news covering the dispute over the cathedral mentioned church restitutions as well, we set a special attribute for this news within the data set. In this way we maintained reliable records of the total number of news, and whether the news relates to church restitutions, dispute over the cathedral, or both of these at the same time. We analyzed 240 messages in total. The variable “Media in total” assigns to each respondent the number of news related to the topic of church restitutions, relation between the state and church, or the St. Vitus Cathedral, which occurred in media during the 7 days prior to completing the questionnaire. We chose this time interval because Wanta and Hu (1994) proved that the agenda-setting effect occurs within one week (in the case of nationwide TV news). We also went through all news included in the variable “Media in total” and categorized them according to their main frame. Frame 1 understands the restitutions as a rectification of a historical injustice. Frame 2 presents restitutions as a field for a political FRAMING AND AGENDA-SETTING 13 battle that, however, does not solve the factual issue (“politicking”). Frame 3 argues the choice of suitable criteria and the way the church restitutions should be performed. Frame 4 supports the opinion that church does not have a moral right for property restitutions. Frame 5 defends the church’s economic activity as morally justifiable. And finally, frame 6 points out that the victims of unsolved restitutions are municipalities that cannot do anything with the blocked property. 18 news items in the monitored corpus (N=240) did not contain any frame. They were either very short (3 news items) or they only covered the dispute over the St. Vitus Cathedral and did not mention church restitutions (15 news items). 222 analyzed news items contained some frame as mentioned above, 196 related only to church restitutions, and 26 also mentioned the dispute over the cathedral. Variables “Frame 1” – “Frame 6” state for each respondent the number of news falling into the given frame that occurred in the seven days prior to completing the questionnaire. However, Frames 5 and 6 occur so sporadically that we decided not to include them in the analysis. We only do not analyze the variables “Frame 5” and “Frame 6”. The news containing frames 5 and 6 is included in the variable “Media in total”. The frequency of media variables are stated in the Appendix 3. In Table 1, we provide a comparison of the classification of frames 1 to 4 according to previously established theoretical typologies. Frames 1 and 4 are of a similar type – both are specifically aimed at church restitutions, they focus on the issue history and not on the up-todate events, both are value-anchored, and are related to a dispute over the very existence of the issue. However, they differ in the angle – Frame 1 evaluates the present situation as problematic and calls for its solution, while the Frame 4 denies the problematic character of the situation that churches were not returned the confiscated property. Frames 2 and 3 are of a similar type in that they focus on up-to-date events and they perceive the issue as a battle of tactics and strategies, not as value-anchored. However, Frame 3 focuses directly on FRAMING AND AGENDA-SETTING 14 restitutions unlike Frame 2, where restitutions themselves are not significant and they are instead a field for a politicking. Frame 2 is the only generic frame in our analysis. Frame 3 does not question the existence of the issue or the necessity of its solution, only the method for church restitutions is controversial. Frame 2 approaches as controversial not only the method for church restitutions but also the call for the necessity of their momentary solution. However, the center of Frame 2 relates to church restitutions only marginally since it mainly concentrates on the politicking itself. For analytical purposes, we only use the second to sixth wave of panel survey. We eliminated the first wave because, logically, it is missing the value of the control variable “Previous mentioning of restitutions”. The seventh to twelfth waves were eliminated because by this time the respondents mentioned the church restitutions and the dispute over the St. Vitus Cathedral very sporadically. Furthermore, we eliminated those respondents that did not fill out their personal characteristics, did not answer in all monitored waves, or answered at a different time than they were supposed to. In order to keep as high a number of observations as possible, we considered as a timely answer also the cases when a respondent filled out the questionnaire on the respective Friday, Saturday or Monday, besides the prescribed Sunday. The numbers of news (variables Media in total, Frames 1–4, and Cathedral) were assigned to respondents precisely according to respective days so they are not influenced by this shift in filling out the questionnaire. After all these modifications, there were 369 respondents left in the data set with a full set of five filled out questionnaires. Comparison of original and analyzed sample provides Appendix 2. Analysis Panel data have a multi-level structure and, therefore, we use multi-level regression models with a dichotomous dependent variable. In our data set, we have 369 respondents, i.e. 1845 observations. Respondents with their characteristics form macro-contexts, stable in time, FRAMING AND AGENDA-SETTING 15 for answers in individual waves of survey. Together with a development of media contents that change every week, individual answers form a micro-observation. Answers of one respondent in different weeks are mutually more similar than answers from different respondents. Contrary to classic linear or logistic regression, multi-level models can take this similarity into account and, therefore, they lead to more precise estimates of coefficients and, mainly, standard errors (Kalvas, Kreidl,Váně, & Štípková, 2009). However, methods of multi-level modeling are not totally stable and, therefore, we follow the recommendation of Allison (1999) and use more methods at once, where each has different advantages and disadvantages. We use marginal models (also population averaged or GEE models) and conditional logistic regression. Conditional logistic regression has a very beneficial attribute - it explains the variability of the dependent variable only through variables that change in time. Thus, it controls the effect of all monitored as well as nonmonitored characteristics of respondents that stay the same in the whole survey period. However, its problem is that it uses data from only those respondents who had a change in the dependent variable during the survey period. That means that conditional logistic regression excludes from the analysis all respondents who mentioned the church restitutions always or never. The marginal model, contrary to conditional logistic regression, handled the data effectively, but this method risked data deformation due to an influence of a non-monitored variable.4 In the next section, we present and interpret the results of both the modeling methods. All calculations are performed by the statistical software STATA 11. Results In total, we estimated six models whose test statistics are stated in Appendix 4. Table 2 provides estimated effects of all monitored variables for all models. Models 1 to 3 study the impact of the variable “Media in total” on respondents’ willingness to mention church 4 The options, advantages, and disadvantages of these methods were discussed in detail already in our previous text (Kalvas et al., 2009). FRAMING AND AGENDA-SETTING 16 restitutions as an up-to-date social event. They do not directly test any of our hypotheses, but provide insight on the effect strength of the number of news for a later comparison, and they also provide a rough test for the validity of the agenda-setting hypothesis. All three models confirm the agenda-setting hypothesis because it is validated that with every news related to the issue or focusing event the odds are increased that a respondent would mention restitutions as an important issue. This result is consistent with both statistical methods of estimation, in control of the respondent’s personal characteristics as well as in control of the answer from the previous wave of survey. Let us also look at the strength of the effect. The estimated logit 0.05 (or 0.04, or 0.03) means that if 14 (or 18, or 23) news items related to the topic of church restitutions or the dispute over the cathedral appear in media in the past seven days, it will double the odds (logit 0.7 = 0.05 * 14, or 0.72 = 0.04 * 18, or 0.69 = 0.03 * 23) that a person will determine church restitutions to be an important issue. Similar or higher values are reached in the media exposition during the third, fourth and sixth wave of the survey (see Appendix 3). The media exposition during the second and fifth wave is statistically significant but it is not substantive significant because there were not enough contributions from the media to allow a doubling of the odds that the church restitutions would set on the respondent’s personal agenda. Models 4 – 6 work with framing of news and, thus, they provide material for the evaluation of the validity of the first to third hypothesis and our basic research question, if there is any point in further examination whether the process of framing creates conditions for a parallel agenda-setting process. Test statistics of models in Appendix 4 show that the models better describe data if we introduce variables indicating frequency of frames in the news content. In our view, it proves important role of frames in the agenda-setting process. In these models, we always use variables “Frame 1” to “Frame 4” as well as the variable “Media in total”. The estimated effects of variables “Frame 1” to “Frame 4” FRAMING AND AGENDA-SETTING 17 determine how the effect of news increases or decreases if it contains a respective frame. The results are consistent with both statistical methods of estimation, in control of the respondent’s personal characteristics as well as in control of the answer from the previous wave of survey. An important result is that the effect of media exposition stays both statistically and substantive significant even if we control the analysis for news frames. Another important result is that effects of frames differ, although in our case this means that only “Frame 2” differs from all others. The negative effect of “Frame 2” markedly reduces the effect of the news itself. According to the estimated model, it diluted the effect of the news (Model 4), annulled its effect (Model 6) or even caused an overall weak negative effect (Model 5). A substantive significant effect of news containing “Frame 2” remains in Model 4, but the effect of this news is 4 times lower than if it contained a different frame. In Model 5, the negative effect of “Frame 2” (logit -0.18) is higher than the positive effect of the news itself (logit 0.16). If we add these values we get an overall weak negative effect of such framed news (logit -0.02). This effect will be factually significant if “Frame 2” appears in at least 35 news stories in the previous seven days. In such case, the odds of setting the church restitutions on the agenda drops at least by half (logit -0.7 = -0.02 * 35). With this or even higher frequency, “Frame 2” appears in news during the third or fourth wave of survey. The overall negative effect estimated by Model 5 does not have a substantive significance during the other waves. We should also mention an increase of estimated effect of “Media in total” in Models 4, 5, and 6. The effect is from three to five times higher than in respective Models 1, 2, and 3. While in Models 1, 2, and 3 variable “Media in total” refers to the effect of an ‘average’ news item, in Models 4, 5, and 6 it refers to the news in reference category: (a) news with frame 5 and 6, (b) news without any frame or referral, (c) and news referring to Cathedral trial without FRAMING AND AGENDA-SETTING 18 restitution frame. Insignificant effects of “Frame 1, 3, and 4” tell us the respectively framed news does not statistically differ from news in reference category and also this news has quite strong positive effect. So we could understand significant negative effect of “Frame 2” relatively – news framed as politicking has much more weaker effect than any other news. Now we look at sociodemographic characteristics of respondents, i.e. Models 3 and 6. Gender did not play a role. In the case of age, however, we do observe substantive as well as statistically significant differences in both models. The two oldest age groups of respondents (72 – 92 years, and 52 – 71 years of age) had approximately the same tendency to adopt church restitutions into their personal agenda, while for the younger groups these odds dropped rapidly. Respondents of 31 – 51 years of age had 3 times lower odds that they would indicate church restitutions, and in the youngest group (18 – 30 years) the odds were 5 times to 6 times lower. In the case of education it is only persons with basic education who differ factually and statistically. Contrary to the other groups, they have approximately three times lower odds that they consider church restitutions as important. The respondent’s religion also proved to be statistically significant for these two models. However, this effect is not highly substantive significant; if an individual endorses some form of Christian religion, they have 1.5 times higher odds that they consider church restitutions to be important. The last result is a negative effect of “Previous mentioning of restitutions”. Models 5 and 6 show dramatic decline of its effect if we add frame variables into models, but it is likely effect of specific composition of analyzed sample. Majority of respondents (199 out of 369) does not mention church restitutions in any wave. Effect of “Previous mentioning of restitutions” is strongly negative in Models 1 and 4 that analyze only 170 respondents who mentioned restitution at least once. (Leaving out respondents without change in dependent variable is trait of conditional logistic regression discussed above.) We also estimate very FRAMING AND AGENDA-SETTING 19 strong negative effect of “Previous mentioning of restitutions”, if we apply Models 2, 3, 5 and 6 on 170 respondents who mentioned restitutions at least once (results are not presented in this paper, please, ask authors). Discussion Kalvas et al. (2009) have assumed the introduction of frame variables could explain the strong negative effect of “Previous mentioning of restitutions”. But effect lasts even if we control the composition of frames. We think we have drained all possibilities how to explain this effect by media variables. This effect could describe the general tendency leave mentioned issues quickly in such a frequently panel survey. But Kalvas (2009b) shows on the same data file that the effect of previous mentioning differs significantly issue by issue. We suggest that the strong negative effect of “Previous mentioning of restitutions” could illustrate the specific relationship of the Czech public to the church restitutions. We will now evaluate the hypotheses postulated in this text. The key role has the overall weak effect of news framed as politicking. We offer two explanations. (1) From all the studied frames, this frame has the weakest relation to church restitutions. As we have stated above, in this frame church restitutions are only a field for a politicking on which the “Frame 2” concentrates primarily. (2) “Frame 2” was by far the dominant one in the monitored corpus of news – it is the main frame in 65 % of news items. It allows other frames a larger potential to cause cognitive dissonance in the news receivers because these frames brought a different perspective on the issue. “Frame 2” is generic, episodic, and a strategy frame at the same time. Therefore, the weak effect that we found within it supports Hypothesis 2 because it is the only generic frame and, at the same time, the only frame where the factually as well as statistically significant weak effect was proved. However, this result contradicts Hypothesis 1 because it assumed that an episodic frame would have a stronger effect than a thematic frame. Furthermore, FRAMING AND AGENDA-SETTING 20 another episodic frame, “Frame 3”, has the same effect as “Frames 1 and 4” which are thematic. Hypothesis 3 is partially supported by the results. It predicted correctly the weak effect of “Frame 2” which is a strategy frame. However, “Frame 3” is also a strategy frame but its effect does not differ statistically from “Frames 1 and 4” that are value frames. From such evidence, it is not clear whether “Frame 3” does not have a negative effect, despite being a strategy frame, or whether for other reasons there is an effect of “Frame 2” which is, among others, a strategy frame. It is possible that the higher the number of our hypotheses a frame suits with its type, the weaker its effect is. “Frame 2” corresponds to two hypotheses (2 and 3), other frames only to a single one. However, we do not recognize this generalization as a general result as it is risky to generalize based on an analysis of a single issue and four frames. We understand it rather as an inspiration for further research where it would be possible to study a higher number of issues and a higher number of frames that will have larger variability in terms of examined typologies. Conclusion Our main goal was to determine if there is any point in further examination whether the process of framing creates conditions for a parallel agenda-setting process. Unfortunately, we were not able to directly test the causal mechanism of the link between both the processes, the role of cognitive dissonance. We only confirmed the effects of variables that describe the frequency of frames occurrence in the press, radio, and TV. In our opinion, however, the results are conclusive, and they show that frames can influence the setting issue on personal agenda. We conclude that further research is needed to address the problem of the relation between framing and agenda-setting processes. Other interesting result is that structure of frames is not able to explain strong negative effect of mentioning issue in preceding wave of study. We must explain this effect by other than the media content variables. FRAMING AND AGENDA-SETTING 21 Our results are not in contradiction with the research of Lecheler et al. (2009), who show that the framing process is blocked by the importance attributed to an issue by an individual. None of the texts contradict the assumption that both processes interact, creating conditions for each other and, therefore, that they influence each other. The framing process influences the setting issues on the personal agenda and, at the same time, the issue established on the agenda blocks the chance that a person will accept an alternative frame. Interaction between both processes is explained by the theory of cognitive dissonance. Frames invoke a person’s cognitive dissonance. The effort devoted to reduction of the dissonance will strengthen the memory traces to the framed issue. The effort also heightens perceived importance of the issue and cognitive elements of the frame that the person chose as a relevant interpretation of the issue. The strengthened importance of these cognitive elements then blocks the acceptance of an alternative frame. An alternative frame could replace the actual frame only if (1) the alternative frame weakens cognitive elements of the actual frame, or (2) the alternative frame brings stronger cognitive elements, or (3) the alternative frame is more consistent with other cognitive elements than the actual frame. FRAMING AND AGENDA-SETTING 22 References: Allison, P. D. (1999). Logistic regression using the SAS system: Theory and Application. Cary, NC: SAS Publishing. Centrum výzkumu veřejného mínění [Public Opinion Research Center]. (2008). Veřejná a mediální agenda [Public and media agenda] (Data file and code book). Cobb, R. W., & Elder, C. D. (1983). Participation in American politics: The dynamics of agenda-building. Baltimore, MD: Johns Hopkins University Press. Cooper, J. (2007). Cognitive dissonance. Los Angeles, CA: Sage. Dearing, J. W., & Rogers, E. M. (1996). Agenda-setting. London, England: Sage. de Vreese, C. H. (2005). News framing: Theory and typology. Information Design Journal + Document Design, 13, 48–59. Dias, Á. M., Oda, E., Akiba, H. T., Arruda, L., & Neuder, L. F. (2009). Is Cognitive Dissonance an Intrinsic Property of the Human Mind?. World Academy of Science, Engineering and Technology, 54, 784–788. Donohew, L. A. & Palmgreen, P. (1971). A reappraisal of dissonance and the selective exposure hypothesis. Journalism Quarterly, 48, 412–420. D’Angelo, P. (2002). News framing as a multi-paradigmatic research program: A response to Entman. Journal of Communication, 52, 870–888. Entman, R. M. (1993). Framing: Toward clarification of a fractured paradigm. Journal of Communication, 43(4), 51–58. Festinger, L. (1962). A theory of cognitive dissonance. Stanford, CA: Stanford University Press. (Original work published 1957). Festinger, L. (1962). Cognitive Dissonance. Scientific American, 207(4), 93–102. Festinger, L. (1999). Social Communication and Cognition: A Very Preliminary and Highly Tentative Draft. In E. Harmon-Jones & J. Mills (Eds.), Cognitive dissonance: FRAMING AND AGENDA-SETTING 23 Progress on a pivotal theory in social psychology (pp. 355–379). Washington, DC: American Psychological Association. (Original work published 1954). Iyengar, S. (1991). Is anyone responsible?. Chicago, IL: University of Chicago Press. Iyengar, S., & Kinder, D. R. (1987). News that matters. Chicago, IL: University of Chicago Press. Kalvas, F. (2009a). Nastolování agendy: Role masové a interpersonální komunikace, osobní zkušenosti a genderu [Agenda setting: Role of mass and interpersonal communication, personal experience and gender]. Plzeň, Czech Republic: ZČU v Plzni. Kalvas, F. (2009b). Ověření hypotézy o nastolování agendy pomocí panelových dat [Testing agenda-setting hypothesis by panel data]. In Škodová, M., & Nečas, V. (Eds.), Veřejná a mediální agenda [Public and media agenda] (pp. 76–97). Praha, Czech Republic: Professional Publishing. Kalvas, F., Kreidl, M., Váně, J., & Štípková, M. (2009). Modelování panelových dat s dichotomickou závisle proměnnou [Modeling panel data with a dichotomous dependent variable]. Data a výzkum, 3(1), 99–124. Lecheler, S., de Vreese, C., & Slothuus, R. (2009). Issue importance as a moderator of framing effects. Communication Research, 36, 400–425. Lee, N.-J., McLeod, D. M., & Shah, D. V. (2008). Framing policy debates: Issue dualism, journalistic frames, and opinions on controversial policy issues. Communication Research, 35, 695–718. Matthes, J. (2007, May). Looking back while moving forward. Paper presented at conference Creating Communication: Content, Control, Critique, San Francisco, CA. McCombs, M. (2004). Setting the agenda. Cambridge, England: Polity. McCombs, M., & Shaw, D. (1993). The evolution of agenda-setting research: Twenty-five years in the marketplace of ideas. Journal of Communication, 43(2), 58-65. FRAMING AND AGENDA-SETTING 24 McLeod, J. M., Becker, L. B., & Byrnes, J. E. (1991). Another look at the agenda-setting function of the press. In D. L. Protess, M. E. McCombs (Eds.), Agenda setting (pp. 47–60). Hillsdale: Lawrence Erlbaum Asociates. (Original work published 1974). Ministry of Culture of Czech Republic. (2008). Návrh majetkového narovnání (2007-2008) [Proposal of property rectification (2007-2008)]. Retrieved from: http://www.mkcr.cz/scripts/detail.php?id=2855. Scheufele, D. A. (2000). Agenda-setting, priming, and framing revisited: Another look at cognitive effects of political communication. Mass Communication & Society, 3, 297– 316. Semetko, H. A., & Valkenburg, P. M. (2000). Framing european politics: A content analysis of press and television news. Journal of Communication, 50(2), 93–109. Trampota, T. (2006). Zpravodajství [News]. Praha, Czech Republic: Portál. Vinopal, J. (2009). Konstrukce panelu respondentů a datový soubor v šetření CVVM [Construction of panel sample and data file for the CVVM survey]. In Škodová, M., & Nečas, V. (Eds.), Veřejná a mediální agenda [Public and media agenda] (pp. 124– 130). Praha, Czech Republic: Professional Publishing. Wanta, W., & Hu, Y-W. (1993). The agenda-setting effects of international news coverage: An examination of differing news frames. International Journal of Public Opinion Research, 5, 250–263. Wanta, W., & Hu, Y-W. (1994). Time-lag differences in the agenda-setting process. International Journal of Public Opinion Research, 6, 225–240. FRAMING AND AGENDA-SETTING 25 Table 1 Classification of frames according to theoretically substantiated typologies episodic/ thematic issue-specific/ generic value/ strategic Frame 1: Rectification of a historical injustice thematic issue-specific value Frame 2: Field for a policy battle (“politicking”) episodic generic strategic II and III / 0 Frame 3: Choice of suitable criteria episodic issue-specific strategic III Frame 4: Restitutions are not morally substantiated thematic issue-specific value Note: types of controversies present in the frame: 0 – present controversy does not relate to the issue I – whether the issue even exists II – whether the issue should be solved III – how the issue should be solved type of controversy I and II I FRAMING AND AGENDA-SETTING 26 Table 2 Estimated coefficients and (standard errors) of multi-level models concerning the occurrence of church property restitutions on personal agenda. Model 1 Model 2 Model 3 Model 4 Model 5 Model 6 Constant -3.42** (0.182) -3.23** (0.248) -3.28** (0.240) -3.08** (0.289) Previous mentioning of restitutions -3.25** (0.339) -2.26** (0.418) -2.27** (0.359) -2.98** (0.377) -0.25 (0.285) -1.01** (0.314) Media in total 0.05** (0.004) 0.03** (0.003) 0.04** (0.003) 0.17* (0.068) 0.16** (0.047) 0.17** (0.048) Framed news in preceding 7 days Frame 1 -0.16 (0.313) -0.19 (0.202) -0.15 (0.207) Frame 2 -0.13† (0.066) -0.18** (0.046) -0.17** (0.046) Frame 3 -0.29 (0.181) -0.10 (0.145) -0.13 (0.147) Frame 4 0.23 (0.409) 0.06 (0.284) 0.00 (0.294) Other or no frame (reference category) Respondent’s gender Man 0.22 (0.187) 0.22 (0.182) Woman (reference category) Respondent’s age 18-30 -1.80** (0.364) -1.69** (0.353) 31-51 -1.19** (0.205) -1.14** (0.200) 52-71 (reference category) 72 – 92 0.08 (0.314) 0.14 (0.302) Respondent’s education Basic -1.14* (0.575) -1.17* (0.568) Vocational school -0.19 (0.201) -0.17 (0.196) Secondary school (reference cat.) University 0.21 (0.241) 0.22 (0.233) Respondent’s religion Christian 0.47** (0.177) 0.47** (0.173) Other (reference category) Nmacro (Nmicro) 170 (850) 369 (1845) 369 (1845) 170 (850) 369 (1845) 369 (1845) Note: † p < .1; * p < .05; ** p < .01. FRAMING AND AGENDA-SETTING 27 Appendix 1 Number of respondents who indicated the restitutions of church property or the dispute over the ownership of St. Vitus Cathedral as an important event in the present or previous wave according to individual waves of research (N=369). 2nd wave 3rd wave 4th wave 5th wave 6th wave Mentioning of restitutions Number 20 121 39 8 15 Percentage 5.4 % 32.8 10.6 2.2 4.1 Previous mentioning of restitutions Number 0 20 121 39 8 Percentage 0% 5.4 32.8 10.6 2.2 FRAMING AND AGENDA-SETTING 28 Appendix 2 Respondents according to gender, age, education, and religion. Analyzed sample (N=369) Original sample (N=658) Number Percentage Number Percentage Gender Men Women Unascertained 123 246 0 33.3 % 66.7 0.0 230 425 3 34.9 % 64.6 0.5 Age 18-30 31-51 52-71 72-92 Unascertained 55 136 155 23 0 14.9 36.9 42.0 6.2 0.0 114 246 249 46 3 17.3 37.4 37.8 7.0 0.5 Education Basic Vocational Secondary school University Unascertained 19 128 155 67 0 5.2 34.7 42.0 18.2 0.0 44 232 277 102 3 6.7 35.3 42.1 15.5 0.5 Religion Christian Others Unascertained 144 225 0 39.0 61.0 0.0 401 254 3 38.6 60.9 0.5 FRAMING AND AGENDA-SETTING 29 Appendix 3 Number of media news which respondents were exposed to in seven previous days, according to the frame, reference to the Cathedral trial, and the precise date of the questionnaire being filled out. Date Frame 1 Frame 2 Frame 3 Frame 4 Frame 5 Frame 6 Without frame Cathedral Media in total nd 2 wave 25.4.2008 0 4 0 0 0 0 0 3 4 26.4.2008 0 5 1 0 0 0 0 3 6 27.4.2008 0 8 1 0 0 0 0 2 9 28.4.2008 0 9 1 0 0 0 0 2 10 3rd wave 2.5.2008 2 36 8 0 0 0 12 20 58 3.5.2008 3 39 8 2 0 0 13 25 65 4.5.2008 5 37 9 2 0 0 13 26 66 5.5.2008 6 39 11 3 0 0 13 26 72 th 4 wave 9.5.2008 7 53 9 5 1 1 3 10 79 10.5.2008 6 51 8 3 2 1 2 5 73 11.5.2008 4 54 8 3 2 1 2 6 74 12.5.2008 3 54 6 2 2 1 2 6 70 th 5 wave 16.5.2008 0 13 1 0 2 0 0 2 16 17.5.2008 0 11 1 0 1 0 0 2 13 18.5.2008 0 6 0 0 1 0 0 0 7 19.5.2008 0 3 0 0 1 0 0 0 4 th 6 wave 23.5.2008 0 19 2 0 1 0 0 0 22 24.5.2008 0 19 2 0 1 0 0 0 22 25.5.2008 0 19 4 0 1 0 0 0 24 26.5.2008 0 19 4 0 1 0 0 0 24 FRAMING AND AGENDA-SETTING 30 Appendix 4 Statistics of estimated multi-level models concerning the occurrence of church property restitutions on personal agenda. Test statistics chi2/ Wald LR chi2 M1: media in total, previous answer (conditional logit) 275.0 M2: media in total, previous answer (GEE) 153.9 M3: media in total, previous answer, characteristics of respondent (GEE) 193.1 M4: media in total, frames 1-4, previous answer (conditional logit) 284.2 M5: media in total, frames 1-4, previous answer (GEE) 179.9 M6: media in total, frames 1-4, previous answer, characteristics of respondent (GEE) 220.2 Indication and description of model d.f. Nmacro Nmicro p-value 2 2 10 6 6 14 170 369 369 170 369 369 850 1845 1845 850 1845 1845 < .001 < .001 < .001 < .001 < .001 < .001 Contrasts Extension with frames M4 – M1 9.2 4 170 850 = .056 M5 – M2 26.0 4 369 1845 < .001 M6 – M3 27.1 4 369 1845 < .001 Extension with characteristics of respondent M3 – M2 39.2 8 369 1845 < .001 M6 – M5 40.3 8 369 1845 < .001 Note: Respondent’s characteristics are: gender, age (3 dummy variables), religion (Christian vs. other), and education (3 dummy variables).