After all is said and done: Experimental study of effects of post-election rhetoric on support for parties in power.

Transcription

1 After all is said and done: Experimental study of effects of post-election rhetoric on support for parties in power. Elina Lindgren Department of Political science University of Gothenburg Paper to be presented at the 2014 ECPR General Conference, September 3-6, University of Glasgow. Abstract During recent years many scholars have pointed to a negative trend in political support in Western democracies, growing cynicism and a perception of governing parties as promise breaking. This cynicism has, among other things, been explained as a reaction to a growing negativity in media coverage of political actors, governmental performance and policy outcomes. Since declining political support have negative impact on democratic societies, ultimately threatening the legitimacy of the system itself, governments need to find ways to handle such exposure. This paper studies whether persuasive rhetoric can help governments to defend post-election actions against negative media coverage, and hence gain more stable support. The theories of persuasive rhetoric that are used, are taken from the philosophy of language and linguistic semantics literature, research disciplines that for decades have worked on developing theories on how language can be used to persuade people. There are yet no studies combining theories from this research field to research on political support, in a postelection context. Hence, this paper contributes both to the political science literature on political support and the philosophy of language literature. It applies the theories of persuasive rhetoric to a post-election political communication context, and tests them empirically within this specific context. The analyses are conducted on data from a vignette experiment conducted in the field, on a diverse sample of 120 persons living in Sweden. Keywords: Political support, political communication, persuasive rhetoric, election promises 1

2 Introduction Can post-election rhetoric help parties in power to contradict negative media exposure to policy outcomes, preventing disappointment and gain stable support? In election campaigns, parties make many promises to the citizenry, and to attract attention and interest they need to make these sounds as appealing as possible. This is necessary in order to gain pre-election support, but when parties come into power it puts them in a vulnerable position where they risk causing disappointment due to exaggerated expectations, and receiving critic for failing with keeping their promises. Several scholars have found that media reports on governmental performance are growing more negative today, and since many people receive most of their political information through the lenses of journalists and reporters, this can contribute both to disappointment with policy outcomes and performance, and declining political support and cynicism post-election (see e.g. Floss; 2010; Farnsworth, 2003; Pattersson, 1993, 1996; Cappella and Jamieson, 1997). Declining political support and trust is perceived as negative in a variety of ways; it threatens democratic legitimacy (Norris, 1999), diminishes incentives to follow democratic rules (Hetherington, 2004), decreases political participation and turnout (Teixeira, 1992), has negative impact on trust for public authorities (Easton, 1965, 1975; Norris, 1999), and hamper governments potentials to solve problems (Hetherington, 1998). As a consequence, negative media coverage requires strategies for governments to defend their policy actions and to enable stable support in a long-term perspective. However, while parties are highly visible in the campaigns and formulate much of the information themselves, they tend to be less visible after the election. Hence, negative information from other actors such as the media will remain less challenged, leading to higher influence for these actors. The point of departure in this paper is that governing parties needs to engage in an ongoing dialogue with their constituents, and explaining their post-election actions to the citizenry in a trustful and convincing way. We know from previous studies that rhetoric can influence support for political parties and policy proposals, pre-election (e.g. Brader and Corrigan, 2005; Jerit, 2004; Lakoff and Wehling, 2012). However, we know less about the effects of post-election rhetoric on support, and whether this is effective in order to justify policy actions and outcomes. This paper test effects of persuasive post-election rhetoric in response to negative information, more specifically whether persuasive rhetoric can help incumbents to defend post-election actions and policy outcomes, and gain stable support. Drawing on theories from Philosophy of 2

3 language and linguistic semantics, the paper poses two hypotheses. Persuasive rhetoric in response to negative media exposure post-election can help governments to 1) convince voters that they have acted in accordance with their promises, and 2) gain more stable support. The political science literature on political communication and support, and the philosophy of language literature on persuasive rhetoric have developed separately, and there are few attempts to apply the theories of persuasive rhetoric to a political communication context. Thus, this paper contributes to both literatures; it applies the philosophy of language theories about persuasive rhetoric to a post-election communication context, and tests the theories empirically in this specific context. The methodological approach is experimental, and the analyses are employed using data from a field experiment including a diverse sample of 120 persons living in Gothenburg, Sweden s second largest city. Support for parties in power and information effects During the latest decades, many studies have pointed on a general decline in support and confidence in political actors among Western democracies (e.g. Pharr, Putnam and Dalton, 2000), and in America, public unpopularity of congress has long puzzled political scientists. (E.g. Stimson, MacKuen and Erikson 1995; Hibbings et al, 1995). Studies have shown how the US congress score lower in evaluation during the later part of the 20 th century than in the earlier parts (Cooper and McKenzie, 1981), and in 1994, the Washington post journalists Morin and Broder published results from a survey indicating that public confidence in the American Congress and the government had reach new lows. In addition, scholars that investigate publics perception of whether governments fulfill their promises have also found that, despite the fact that many Western governments fulfill most of their campaign promises, significant parts of the electorates have a general view of governments as being promise breaking (see e.g. Naurin, 2011; Thomson 2011). According to Marra, Ostrom and Simon (1990) can variations in public support for the president often be explained in three different ways. First as a law of inevitable decline, secondly as a function of environmental connection between chief executives and macro features in the political landscape, and finally as due to political dramas (e.g. speeches, foreign trips and diplomatic agreements). Another explanation to declining support for the US congress, has been that citizens compare their expectations on their representatives in Congress to how they perceive the actions and behavior of the Congress. This has been called the expectations-perceptions differential by Congressional scholars (Kimball and 3

4 Patterson, 1997), and expectations gap by presidential scholars (e.g. Stimson, MacKuen and Erikson, 1995; Waterman, Jenkins-Smith and Silva, 1999). Also whether voters become winners or losers of elections have been shown affecting support for the government in a postelection context (e.g. Anderson et al, 2002; Kahneman 1994; Esaiasson et al, 2007). Effects of negative media exposure to governmental activity and policy outcomes In recent years, more and more focus has been directed to the relation between negative media reports and the declining support in Western democracies. There are many studies pointing on how support and trust for parties in power are affected negatively when voters become exposed to governmental performance and policy outcomes they regard as negative (Holmberg, 1999; Kumlin, 2002; Citrin and Green, 1986; Weatherford 1984), to framing of governments in terms of self-interested and strategic (Capella and Jamieson, 1996), and when media coverage induce an impression that political processes do not match individual preferences (Floss, 2010). A large part of what we know of governmental performance is brought to us by media (e.g Bennet, 2001; Lipset and Schneider, 1983), and together with studies showing that media seem to portray politicians, governments and policy outcomes more negative today (e.g. Patterson, 1996; Farnsworth, 2003; Bennet, 2001; Sabato, 1993; Holmberg, 1999), this brings many into considering media as an important actor contributing to peoples negative reactions to governmental performance (see e.g. Brody, 1991; Pattersson 1993; Cappella and Jamieson, 1997; Leone, 1994). That especially negative media reports have impact on how people perceive the government and policy outcomes are supported by political psychology studies, which has shown that information that causes negative emotional reactions are particularly efficient in seizing peoples attention and changing views (see e.g. Derrberry, 1991; May, Kane, and Hasher, 1995; Marcus, Neuman and McKuen, 2000). Thus, if voters become exposed to negative information about the governments performance and policy outcomes, it should make them extra sensitive and open to re-evaluate the government. 4

5 How can parties in power handle negative exposure to post-election performance and policy outcomes? In order to keep a stable support, governments needs to find ways to handle the increasing negativity in media coverage of post-election performance and policy outcomes, and to respond to the critic in a reliable and convincing way. This is particularly important, since the news reports often seem to not only inform about the details of political events themselves, but also contribute with subjective interpretations of these (see Brody, 1991). James Hoge (1994) formulates this as that the ubiquitous and emotional media puts pressure on politicians to give straightaway responses to news. Since the news is immediate, emotional and often taken out of its context, it requires a political leadership that is capable of constructing viable policies, to explain them accurately in a way that is easily grasped by voters, and that knows when to stand fast behind them. According to Hoge, people that are affected by news reports will still listen to politicians that are skilled in the art of persuasion. Since negative news that induces negative feelings also seem to create a situation where people are more open-minded and sensitive to new information, they may also as a side effect be more open to persuasion by the political actors themselves. In this sense, what seems to be important is that the incumbent manages to ease anxiety, anger or other negative feelings caused by negative information from external actors, turning them into something positive, and convince the electorate that their performance is in accordance with their pre-election promises. If, on the other hand, the government lacks such persuasive skills, they will end up in a situation where the [negative] media are catalytic for political events. Can persuasive rhetoric help governments to defend policy outcomes against negative media exposure? In a pre-election context, it has previously been found that rhetoric involving wording that evokes feelings, can help parties to influence political attitudes, support and behavior (e.g. Brader and Corrigan, 2005; Jerit, 2004; Lakoff and Wehling, 2012; Entman, 1993). However, this literature is less extensive regarding the effects of such rhetoric in a post-election context, and whether similar rhetoric that have effect in election campaigns also can be used postelection 1. This paper takes it standpoint in theories from the research field that is sometimes 1 In fact, there are at todays date not many studies of post-election rhetoric, although there are some studies enhancing importance of speeches when it comes to the American president. According to Esbaugh-Soha (2010) 5

6 called philosophy of language and linguistic semantics, which provides a theoretical framework for analyses of the mechanisms that makes rhetoric persuasive. Scholars in these fields studies a broad range of communicative situations, this paper focus specifically on the parts covering situations in which one actor aims at persuading another in a certain direction. Persuasive rhetoric, definition With persuasion means attempts to influence listeners by changing or redirecting attitudes, motivations, beliefs or behavior (Seiter and Gass, 2010), in situations as different as selling a product, convincing a jury, or fostering political support (Fautsch, 2007). Schacter, Gilbert and Wegner (2011) distinguish between systematic and heuristic persuasion. Systematic persuasion refers to processes through which attitudes or beliefs are changed by appeals to logic and reason, while heuristic persuasion is the process through which attitudes or beliefs are changed by appeals to emotions (see also for example Larson, 2010; Stevenson, 1963, 1944; Walton; 2001). This paper focuses on the latter, specifically when emotional reactions are evoked by use of certain words, henceforth referred to as persuasive words. I use three characteristics from the literature, to define the notion of persuasive words. The first characteristic is that people respond to them emotionally (they strongly like or dislike them). The literature on heuristic persuasion distinguishes between words and sentences descriptive and emotive meaning. The vast majority of words are mostly descriptive and emotionally neutral. Some words, however, have very strong emotional connotations, where the emotive meaning represents the feelings or attitudes that exposure to the word evokes (Stevenson, 1963a, Walton, 2001p. 117; Macagno and Walton, 2010, 2014). Persuasive words are those that have strong emotive meaning that outperform the descriptive meaning, with other words, terms that people like or dislike regardless of how they are defined (e.g. Walton, 2001, 2003; Stevenson, 1944; 1963; Aomi, 1985). Examples in kind can be freedom which is strongly associated with something positive, while oppression has negative emotive connotations. The second characteristic of persuasive words is that they often denote a social or political contextual value (Stevenson, 1963a, 1973; Finlay, 2005; Hare, 1952, 1997). This includes morals, ethics, preferences, belief systems and worldviews, and is used to identify the norms for individuals, groups and cultures. Using value-laden words such as democracy may impede opposition to a position, since opposing the position would equal opposing the speechmaking environment for presidents have changed with the emerge of the 24-hour cable news cycle, and speechmaking should be the Presidents primarily communication of their policy preferences to legislators and bureaucrats, and to address the public and the media. Further, Marra, Ostrom and Simon (1990) have found that prime-time foreign content speeches induce public support for the President. 6

7 democracy. (e.g Dancy, 1993; Gibbard, 1990; Stevenson, 1963, 1944; Schacter, Gilbert and Wegner, 2011; Petty et al, 2009) 2. The third characteristic I include in the definition of persuasive words is generality and abstraction. Persuasive words are often found at a rather high level of generality and abstraction, such as freedom and opportunities. Common for persuasive words are that they can be interpreted very differently, and also that they can be applied to many different situations, aspects of life and social phenomenon (Walton and Macagno, 2014; Walton, 2001, 2000; Stevenson, 1944). The level of generality is however diffuse. Many words can be stated at different levels of specificity, and can even be associated with one single descriptive meaning (Stevenson, 1944, 1963). Since changes in the descriptive meaning of the term, redefinitions, is difficult to notice for many listeners, they can easily be used for semantic manipulation (e.g Aomi, 1985). 3 Except for the above three characteristics, there are some additional components that can indicate that words are used to persuade. First, persuasive words are often preceded by words with enhancing effect, such as true, great and real, and the persuasive words are often repeated in the same messages. Persuasive words can also be associated with determinism. The use of the term unavoidable for example, is more or less generally accepted as a reason for an action. An avoidable action can be controlled by ethical judgments because the actor has a choice ; unavoidable actions cannot. (Stevensson, 1944; Walton, 2001). In order to achieve equality on the job market, we have no other choice than, to illustrate with an example. In sum, what is here referred to as persuasive words, are words that are emotive, often associated with values, is at a high level of generality and abstraction, and that can be applied to a variety of contexts and issues. They are also often repeated in the same message, preceded by enhancing words, and followed by statements such as there are no other options. First- and second order persuasion In 1944, Charles Stevenson presented a theoretical model for two different sets in which persuasive words can be used. He calls these different sets for first- and second-order 2 For a further discussion about self-affirmation, see Steele (1988). 3 Walton (2001, p 122) illustrates this with the following example. In an American presidential election campaign, the presidential candidate Ronald Reagan promises a social security fund to provide a safety net for the truly needy. When he after the election find it necessary to cut the program, he changes the definitions of truly needy and safety net. Because of the terms vagueness, many people did not notice the differences. This even though safety net and the truly needy now in some senses referred to notably different things than in the campaign. 7

8 persuasion, and which is best depends on the situation. First-order persuasion is when a persuader uses a persuasive word in a message, without defining what she or he means with the word. When the speakers meaning or definition of the word is left out, the audience can be considered to interpret the term due to their subjective comprehension, or to the general definition used in the social context by which they are surrounded. Hence, this use of persuasive words should be suitable when a persuader wants to hide certain aspects of an issue that might be perceived as negative. Second-order persuasion is when the speaker includes a definition of the persuasive word, when they formulate their message. In this sense, the word becomes closed for different interpretations. Since one of the characteristics of persuasive words is that people tend to keep their positive or negative attitudes towards it even when the descriptive meaning changes, such words can be used to push people to reject their personal idea of the word and instead integrate it with the new definition they are exposed to. (For later similar discussions, see Walton, 2001; Aomi, 1985). Applying persuasive rhetoric to a post-election political communication context According to previous findings, voters who become exposed to policy outcomes and governmental actions they regard as negative, can be expected to feel disappointed, loose support and confidence in the government, and even regard the incumbent as being promise breaking. In order to prevent disappointment and gain stable support, parties in power must therefore pay attention to what information that is send to the citizens after the election (as well as complaints made by citizens themselves), and respond to negative information and complaints in a way that convince the citizens that they are acting responsibly, reliable and in accordance with their promises. In this paper, I use theories of persuasive rhetoric from philosophy of language and semantic linguistics, and apply them to the post-election political communication context. In particular, I apply the strategies on situations were a government need to communicate with voters to defend their policy performance and outcomes against negative media exposure. In sum, what I do is testing whether persuasive rhetoric, in particular persuasive words and what Stevenson (1944) calls second-order persuasion, can be used by governments post-election to defend themselves against negative media exposure to policy actions and outcomes. I expect the strategy to work in three ways. First, since a characteristic of persuasive words is that people tend to like them independent of how they are defined, they 8

9 should push the voters to become positive towards, or at least accept, policy outcomes. By using a term with a strong moral and ethical emotive meaning, such as freedom, justice, democracy, equality, the voters become forced into accepting the actions taken 4. Furthermore, repeating and defining persuasive words that were used in the election campaign helps the government convincing the voters that their actions are taken in accordance with what they said they would before the election, and in fact what the voters themselves voted for. Finally, by referring to determinism, the government should also be able to defend the outcomes and actions in a way that make themselves less responsible. This leads to the hypotheses that is posed and tested in the paper: Persuasive rhetoric in response to negative media coverage post-election can help governments to 1) convince their voters that they have acted in accordance with their promises, and 2) gain more stable support. I illustrate the idea with an example. In an election campaign, a political party claims that if they win the election, they will implement policies to induce peoples rights to weekend leave and vacation. A campaign appeal concerning the job market that involves wording in terms of right to free time and vacation can be expected to be associated with things such as more holidays a year, less overtime work and the right to consecutive daily rest. Imagine that this sounds appealing to some people; they become enthusiastic and their support for the party increase. After the election, the media starts reporting about the governments performance, indicating that the party focuses their work on getting unemployed people into work, and that those who do not have a job must undergo compulsory internship at different workplaces, - actions that actually leads to less free time and vacation. To reduce disappointment following these negative reports, the party can argue that they are working on fulfilling their promise, from the logic that without a job, people cannot have vacation. They can also choose to frame it using more persuasive rhetoric, such as; In order to feel the sense of having a weekly leave, a free day or a well-deserved summer vacation, people in first hand need to have a job. Without a job, a person will feel depressed, loose her sense of selfconfidence, she will feel that she do not contribute to the society, always feel stressed about finding something to do, which ultimately cause a feeling of never being really free. Hence, to fulfill what they promised in the campaign, induce the right to some days of weekly leave and vacation, they have no other choice than to push the population into the job market. 4 This requires that the government manage to convince the voters that their actions are taken in order to achieve what the words indicate. 9

10 Research design To examine the effects of persuasive rhetoric post-election, I used an experimental method. The experiment had a 2 factorial design, and the participants were exposed to one out of two different conditions. The treatments were designed as news articles, reporting about negative consequences following the implementation of a project proposed in a fictive election campaign. The treatment condition included persuasive strategies, while the other condition that served as control, did not. In order for the respondents to have a policy to evaluate post-election, they were first exposed to a fictive campaign pamphlet. The pamphlet was identical for all respondents, and it involved a project proposal to reconstruct the City center of a fictive city (the River city), as well as description of policy plans included in the project. The pledge maker was a fictive party, presented as the New Party. An existing proposal was used as inspiration, to make the proposal sound realistic and increase the incentives for the respondents to evaluate it seriously. The existing proposal involved a reconstruction of the City center of Gothenburg in Sweden, the City in which the experiment was employed. The project was at the time of the experiment mainly a vision and not yet implemented, and the respondents evaluations can therefore be considered unbiased by existing outcomes. To be able to test the theory, it was necessary to present negative information about the outcomes in the post-election treatment. If the respondents would associate the fictive proposal to the real one, this information could have negative spillover effects on the respondents attitudes towards the real proposal. To avoid this, the respondents were from the beginning informed that the texts were fictive. The experimental treatment was designed as a newspaper article three years after the election, reporting about the implementation and outcomes of the project. To be able to examine whether post-election rhetoric can be used to contradict negative information about policy outcomes and induce stable support, the treatment included two conditions. The report about the policy outcomes were identical in both treatments, and both included interviews with politicians involved in the project as well as some representatives from different Citizens organizations. What differed between the two conditions was that one included persuasive rhetoric when the responsible politicians were interviewed, while the other, which served as control, did not. The purpose with the stimuli was to see whether persuasive rhetoric could help parties in power to contradict negative information about policy outcomes, and gain stable support. Second order persuasion, which I hypothesize as being useful in a post-election 10

11 context, means that the information giver defines a previously used persuasive word. First, one of the proposals presented in the initial campaign pamphlet was that the project should involve the building of new residences characterized by different standards, to enable for persons with different economic background to live in the city center. In the post-election article the government is criticized for that the project in fact caused more segregation, and that 80% of the new residences that were built as a step in the project were condominiums or houses, meaning that only 20% of the new housing was rental units. In the control version of the stimuli, the interviewed politician answer to this critique by arguing that he do not see this as a problem, since the promise never was that it should be half / half. In the persuasive stimuli, the politician instead responds we promised new housing which includes different economic standards, though it was never the Party s intention that it should be half / half. The construction of more condominiums than rental units is unavoidable (determinism), he argues, and instead, we should regard it as positive that so much as 20 percent of the new housings are rental units, which persons with lower economic status can afford (secondorder persuasion). In the pre-election pamphlet, the party also stated that all residents would be invited to open meetings with political representatives, to participate and contribute with their insights. A representative from a Citizens Initiative criticized the way the party had handled this, and argued that the residents impact on the shaping of the project were very marginal. In the control version of the post-election stimuli, another politician admit this, but points out that these kinds of decisions must be made by politicians and public officials. In the persuasive stimuli, the politician instead respond that the idea with the dialog was to inspire politicians and create an exchange of ideas (second-order persuasion), but points out that politicians and public servants have to take the actual decisions, since they are the ones who have the bigger picture (second-order persuasion). The persuasive version of the article further includes deterministic arguments, in response to critic that several smaller businesses and industries had been closed down or moved outside the city, and that many had lost their jobs as a consequence. In response, the politicians end with that the development was unavoidable (determinism), and that if the party would have acted differently, the city could not have followed the international development and it would still have been stuck in old patterns (determinism, persuasive metaphor). Adding to above, the persuasive stimuli also included a repetition of persuasive words that were used in the pre-election pamphlet, such as development, inspire and 11

12 sustainability, as well as enhancing words, such as great. Please see appendix 1: 1, 2, 3 for a presentation of the full stimuli texts (translated to English). Treatment control To enable improvements of the texts before running the experiment, a small-scale pilot was conducted after which the respondents were asked to give comments on the texts and the questionnaire. After the major experiment post-tests were employed, to see whether the negative article actually reduced enthusiasm and triggered disappointment towards the project. Evaluation of the proposal pre-election and post-election were operationalized by indexes using the questions that were asked before and after the negative article; What is your attitude towards the project (scale 1-5) and What do you think about the project (scale 1-5), (Cronbachs α 0.87 respectively 0.92). Both indexes are standardized to scales ranging from 0=very bad, to 1=very good. A paired sample t-test showed that the evaluation of the project decreased with.23 percentage points, from.74 before the treatment to.51 after the respondents had been exposed to the negative article. The t-test showed that the differences in mean across the sample from before and after the treatment were significant on a 99% confidence level (t = 11.4, df = 118, p < 0.000, two-tailed). The effect size in terms of Cohens d 5 was also large (d = 1.0). Hence, it seems that the negative article, as intended, decreased the respondents evaluation of the project. 6 I proceeded by controls for that the negative article reduced support and trust for the New Party. For this, two paired sample t-tests were conducted on support and trust from pre- to post- the negative article. The first test showed that support for the New Party decreased with 17 percentage points across the sample, from.6 before the treatment to.43 after the respondents had been exposed to the negative article. The test showed that the differences were significant on a 99% confidence level (t = 10.17, df = 118, p < 0.000, twotailed). A calculation of standardized effect using Cohens d indicated that the size of the effects of the negative article was large (d = 0.9) 7. The second t-test showed that also the 5 The d- value in t-tests is used to interpret the effect size of the treatment, and gives a standardized measure of the extent to which the mean scores differ in terms of standard deviation. The formula for counting the d-value is: d = (x 1 - x 2 ) / Mean standard deviation. A d-value on 0.5 is considered as medium effects, 0.8 is considered large, and values below 0.2 is considered as very small. (Cohen, 1969, 1988; Olejnik and Alegina, 2000) 6 A control for regression to the mean was conducted, and a Pearsons correlation test showed a significant negative correlation between support pre-election and differences in support from pre- to post-election ( Support pre-election subtracted from Support post-election ), on a 0.01 confidence level. Since the analysis revealed significant negative correlations on support, regression to the mean effects should be taken into account when evaluating the results. However, the correlation is fairly weak (-0.304**, two-tailed). 7 A control for regression to the mean was conducted, and a Pearsons correlation test showed a significant negative correlation between support for NP pre-election and differences in support from pre- to post-election 12

13 respondents trust for the New Party decreased across the sample, from 5 before the treatment to 4 after exposure to the negative article. The test showed that the differences in mean scores before and after the treatment were significant on a 99% confidence level (t = 6.652, df = 118, p < 0.000, two-tailed), and the standardized effect size was medium in terms of Cohens d (0.6). Hence, I draw the conclusions that the negative article worked as expected, both concerning support and trust for the New Party; it decreased the respondents support and trust for the Party. 8 Table 1: Effects of negative exposure to governmental performance and policy outcomes on support and trust for the party in power. Support for the New Party (stand. scale 0-1) Pre-election Post-election t-value (sig.) Mean (.000) Std Number of respondents Trust for the new Party (scale 0-10) Pre-election Post-election t-value (sig.) Mean (.000) Std Number of respondents Notes: Paired sample t-test, two-tailed. ( Support for NP pre-election subtracted from Support for NP post-election ), on a 0.01 confidence level. Since the analysis revealed significant negative correlations on support, regression to the mean effects should be taken into account when evaluating the results. However, the correlation is fairly weak (-0.364**, two-tailed). 8 A control for regression to the mean was conducted, and a Pearsons correlation test showed a significant negative correlation between trust for NP pre-election and differences in trust from pre- to post-election ( Trust for NP pre-election subtracted from Trust for NP post-election ), on a 0.01 confidence level. Since the analysis revealed significant negative correlations on support, regression to the mean effects should be taken into account when evaluating the results. However, the correlation is fairly weak (-0.338**, two-tailed). 13

14 Questionnaire and operationalization of variables The experiment started with the campaign pamphlet concerning the City re-construction project. After reading the pamphlet, the respondents were handed a questionnaire including questions regarding enthusiasm and support for the project and the fictive party. After concluding this first questionnaire, the respondents were exposed to the treatment (negative article on policy outcomes), after which they were handed a second questionnaire. In this questionnaire they were asked to re-evaluate the project. The questions were the same as in the previous one, however now initiated by Regarding the information you have now. The independent variable in the study is rhetoric in response to negative exposure to policy outcomes, and it is operationalized in the negative article. The variable is dichotomous and has two conditions, 1) Persuasive rhetoric, and 2) No persuasive rhetoric. The dependent variables aim at measuring perception of the project as fulfilled or not, as well as support and trust for the party in power, post-election. Support for the party in power is operationalized by an index of two questions asked after the negative article (Cronbachs α.88). The questions were Given the information you have now, what is your attitude towards the New Party (scale ranging from 1=negative to 5=positive), and Given the information you have now, what is the probability that you would vote for the New Party in next election (scale ranging from 0=very small to 10=very high). Trust for the party in power posttreatment, were operationalized by the question Given the information you have now, how much do you trust the New Party? (Scale ranging from 0=very little to 10=very much). Perception of the project as fulfilled or not were operationalized by a question posed after the treatment, including three alternative options: No, the project is not fulfilled, The project is neither fulfilled nor not, and Yes, the project is fulfilled. To be able to measure the effects of negative media exposure, controls for support and trust for the party pre-treatment were included. The variable support for the New Party pre-treatment was operationalized by an equivalent index as the one for post-election support, using questions regarding probability to vote for as well as attitudes towards the party, that were asked before the respondents were exposed to the negative article (Cronbachs α.84). The variable trust for the New Party pre-treatment were operationalized by the question Given the information you have now, how much do you trust the New Party? (Scale ranging from 0=very little to 10=very much), that were asked before the respondents were exposed to the negative article. 14

15 Recruitment and description of sample To estimate a suitable number of respondents, I used the answers from a pre-survey on 12 persons. The effects in the pre-analyses turned out to be quite small and I therefore chose to use a relatively large sample (60 persons in each group, a total of 120 persons). The subjects were recruited at the central station in Gothenburg, Sweden s second largest city. The Central station in Gothenburg serve as transit hub for local buses and trams, trains with short- and long distance destinations, as well as airport shuttle buses, and can therefore be expected to cover travellers from different social and demographic backgrounds. During three weeks, I recruited participants from the central station to participate in the experiment. The respondents were told that they were going to participate in a survey concerning election pledges and public opinion. To avoid bias in the recruitment process, I systematically asked all persons that were sitting on the waiting benches at the station, one by one starting with the first person to the left on the first bench in each section 9. To randomize the groups, the treatments were randomly distributed to the participants using a list of a randomized order of numbers. After agreeing to participate, the respondents were informed about the conditions for their participation, that the participation was voluntary, and that all responses were going to be treated anonymously. The experiment started with an introduction questionnaire including some initial questions regarding background factors. Frequency analyses of these variables revealed an underweight of persons over 65, and an overweight of persons with post- high school level of education, and politically interested. Besides this, the participants were rather equally spread across gender, political trust and place of residence (rural-urban dimension) across the sample. Compared to a more representative sample from the 2010 Riks-SOM survey at University of Gothenburg 10, the level of education among the participants are in this experiment more skewed (more highly educated people in my sample). Except for this, differences from the more representative sample are fairly small. Please see table 1 in appendix 2 for description of the composition of the sample, and comparisons with a representative sample. To rule out alternative explanations from the analyses, randomization checks were conducted using t-test statistics and regression analyses. The analyses revealed no significant differences between treatment and control group concerning age, gender, 9 If the person was under 16, I kindly told them that since the survey concerned Swedish voters, and soon to be voters, persons under 16 did not fit the sample and that they therefore could not participate in the study. 10 The data from Riks-SOM, 2010 from the SOM-institute (Institute for research on society, public opinion and media) at University of Gothenburg, covers a systematic probability-sample, including 9000 persons. ( 15

16 education, residence, political interest, and support and trust for the New Party pre-election 11. However, a t-test on political trust (scale ranging from 1=low to 5=high) revealed unwishedfor significant differences between the treatment group (mean=3.05) and the control group (mean=2.73) on a 10% confidence level (p. 062). This means that the group that were exposed to the treatment were significantly more trusting than the respondents in the control group, and controls for political trust will therefore be included in the forthcoming analyses. Analysis and results The following sections present the results from analyses of the questions posed in the paper. It starts with the results from multinomial regression analyses of treatment effects on perception of the project as being fulfilled or not, including control for political trust in general. This is followed by bivariate OLS regression analyses of effects of persuasive rhetoric on support and trust for the party in power, post-election. Finally is presented the results from multivariate OLS regression analyses of support and trust for the incumbent post-treatment, including control for political trust in general and perception of the project as being fulfilled or not. Did the New Party fulfill the project? Table 2 summarize the results from multinomial regression analyses of effects of persuasive rhetoric post-election, on the respondents perception of whether the New Party fulfilled what they promised in the pamphlet, or not. The effects are presented in relative risk ratio, meaning the probability that one unit change in the independent variable leads to a perception of the project as fulfilled or neither fulfilled nor not fulfilled, compared to the reference category not fulfilled. The first model shows that being exposed to negative information about policy outcomes together with persuasive rhetoric, increases the probability of perceiving the project as neither fulfilled nor not compared to not fulfilled, 3.12 times when compared to the control group (99% confidence level). It also shows that being exposed to persuasive rhetoric increases the probability of perceiving the promise as fulfilled rather than not fulfilled times, compared to the control group (99% confidence level). This supports the hypothesis that using persuasive rhetoric increases the probability of perceiving the 11 T-test, differences between treatment and control: Age (p.517), political interest (p. 568), political trust (p. 067), Support for NP pre-treatment (p.128), trust for NP pre-treatment (p.239). Logistic regression, differences between treatment and control: education (high, low) (p. 153), gender (p.926). Multinomial regressions, differences between treatment and control: Suburb compared to countryside (p.481), City compared to countryside (p.568). 16

17 project as being fulfilled compared to not fulfilled, when there are negative information about policy outcomes. Since political trust in general varied significantly between the treatment group and the control group, a second model was run including control for political trust. These analyses showed that the probability of perceiving the project as partly fulfilled compared to not fulfilled decreased slightly from 3.12 to 2.96 times when exposed to persuasive rhetoric compared to the control group, but the results stay significant on a 95% confidence level. The results also showed that the probability of perceiving the project as fulfilled compared to not fulfilled decreased slightly from to 9.66 times when being exposed to persuasive rhetoric compared to the control, but the results remain significant on a 99% confidence level. 12 The third model includes control for the respondents evaluation of the proposal pre-treatment as well as post-treatment. In this model, the probability of perceiving the promise as partly fulfilled compared to not fulfilled increases to 3.66 times when the respondents are being exposed to persuasive rhetoric compared to the control, and the probability of perceiving the project as fulfilled compared to not fulfilled increases to times compared to the control. We can also see that the probability of perceiving the project as partly fulfilled as well as fulfilled, compared to not fulfilled, increases significantly as the evaluation of the proposal pre- and post-election increases. In sum, the results reveal that both persuasive rhetoric, general political trust and the evaluation of the project pre- and post-election have significant effects on the respondents perception of the project as being fulfilled or not. Since the effects of persuasive rhetoric remain significant in all models, we can also draw the conclusions that the effects of the treatment that appears in the bivariate model are not the results of hidden effects of the respondents higher levels of political trust in general, in the treatment group. 12 The model also revealed that general political trust increases the probability of perceiving the project as partly fulfilled or fulfilled, compared to not fulfilled. 17

18 Table 2: Multinomial regression analyses of perception of project as fulfilled, neither fulfilled nor not, and not fulfilled. Effects of persuasive rhetoric presented in relative risk ratio. Neither fulfilled nor not Fulfilled Model 1 Treatment (Ref=control) 3.12** (1.32) 11.15** (9.4) Model 2 Treatment (Ref=control) 2.96* (1.27) 9.66** (8.36) Pol trust 1.25 (0.29) 3.59* (1.81) Model 3 Treatment (Ref=control) 3.66* (1.89) 15.17** (14.88) Pol trust 1.14 (0.33) 3.05* (1.71) Evaluation project pre-treatment 0.04* (0.06) 3.83 (13.06) Evaluation project post-treatment 445.6*** (611) ** ( ) N Total N = 114 Notes: Reference category for the equation is Project Not fulfilled (n=42) Standard errors in parentheses. ( * ) < p0.1, * p<0.05, ** p<0.01, *** p< Effects of persuasive rhetoric post-election on support and trust for the New Party; Controlling for political trust in general and perception of the project as being fulfilled or not. Table 4 below summarizes the results from OLS regression analyses of the effects of persuasive rhetoric on support and trust for the New Party post negative media exposure. Since the randomization checks of background variables revealed an unwished for significant difference between the treatment group and the control group regarding general political trust, the second models include political trust as independent variable, to see whether the treatment effects hold under control for this. The third models include control for perception of the project as being fulfilled or not. 18

19 The first models in the table reveal significant effects of persuasive rhetoric on support and trust for the Party in power; the support as well as trust declines when going from the treatment group to the control group. These results indicate that persuasive rhetoric has significant effects both on support and trust for the Party in power, when it is used together with exposure to negative information about post-election performance and policy outcomes. Hence, the results from the first models indicate that persuasive rhetoric can help governments to alleviate losses in support and trust when people are exposed to negative information post-election. When general political trust is included, the effects of persuasive rhetoric postelection become smaller, both for support and trust (see model 2 in table 3). The effects are still significant on a 10% level on support for the New Party, but the treatment effects on trust become insignificant. These results confirm that the treatment effects in the bivariate model are not results of asymmetry in political trust between the groups, when it comes to support for the New Party. For trust however, the effects of the treatment become insignificant when general political trust is included, and it is therefore probable that it is the higher levels of general political trust in the treatment group that causes the positive effects on trust for the New Party, not the persuasive rhetoric. The third models show that support for the New Party post-treatment significantly increases from 0 to (scale 0-1) when the project is perceived as fulfilled compared to not fulfilled, and from 0 to 0.2 when the project is perceived as neither fulfilled nor not fulfilled compared to not fulfilled. It also reveals that the effects of persuasive rhetoric on support become smaller (the coefficients decrease from in the second model to 0.008), and insignificant. When it comes to trust for the incumbent Party, the effects of going from treatment- to control group unexpectedly become positive, however they are still insignificant. Taken together, the results indicate that persuasive rhetoric can alleviate negative effects on support caused by negative media exposure to policy outcomes, but not when it comes to trust. If the treatment would have had no effects on the perception of the project as being fulfilled or not, the conclusions to be drawn from the experiment would be that persuasive rhetoric have no effects on declining support due to negative media exposure, since the effects of the treatment disappeared when perception of the project as fulfilled or not were controlled for. However, since the treatment did had effect on the respondents perception of the project as being fulfilled or not, it seems like persuasive rhetoric have no direct, but indirect, effects on support post negative exposure to policy outcomes, in the way 19

20 that it affects the respondents perception of the project as being fulfilled or not. Perception of the project as being fulfilled or not seem to serve as mediator of the effects of persuasive rhetoric, on support. Table 3: Effects of persuasive rhetoric on support and trust for the New Party, post negative media exposure. Controlling for general political trust and perception of the project as being fulfilled or not. Support Post-treatment Trust Post-treatment Model 1 Model 2 Model 3 Model 1 Model 2 Model 3 Expgrp * ( * ) ( * ) (0.040) (0.038) (0.034) (0.416) (0.386) (0.352) Political trust 0.087*** 0.060** 1.011*** 0.733*** Project fulfilled Dum. Yes (0.02) (0.018) (0.209) (0.186) (ref: No) 0.357*** 3.684*** Dum. Neither (0.06) (0.629) yes nor no (ref: No) 0.2*** 1.79*** Intercept 0.571*** (0.063) 0.278** (0.091) (0.035) (0.365) (0.082) 5.01*** (0.659) ( * ) (0.928) ( * ) < p0.1, * p<0.05, ** p<0.01, *** p< Notes: OLS regression, std. Error in parentheses. Expgrp, 1= Persuasive rhetoric post-election, 2= Control (No persuasive rhetoric). Political trust in general, scale 0=very low, 10=very high. Dependent variable 1) Support for the New Party post- treatment, standardized scale ranging from 0=very low to 1=very high. Dependent variable 2) Trust for the New Party post- treatment, scale ranging from 0=very low to 10=very high (0.851) N Adj r

21 Summary results The paper aimed at testing two hypotheses, whether persuasive rhetoric can be used by parties in power in response to negative information about post-election performance to convince voters that they acted in accordance with their promises, and to gain stable support. First, multinomial regression analyses showed that including persuasive rhetoric as response to negative media exposure to policy outcomes increased the probability of perceiving the project as fulfilled or neither fulfilled nor not, compared to not fulfilled. The effects of persuasive rhetoric did hold under control for the respondents general political trust, as well as for evaluation of the proposal pre- and post- the negative article. This supports the hypothesis that persuasive rhetoric, when used in response to negative media exposure about policy outcomes, increases the probability of perceiving promises as being fulfilled compared to not fulfilled. Bivariate regression analyses further showed significant positive effects of persuasive rhetoric on support and trust for the New Party, after the respondents had been exposed to negative information about policy outcomes. For support, the effects held under control for political trust in general, which strengthen the conclusion that support for the party in power was affected by post-election rhetoric. This was however not the case for trust for the party, where the effects of the treatment became insignificant under control for political trust in general. In this case, it is likely that the effects of the treatment that appeared in the bivariate models are the results of asymmetry in general political trust between the experiment groups, i.e. the respondents higher levels of political trust in the treatment group. Finally, I wanted to see what would happen if perception of the project as being fulfilled or not was included in the analyses of support and trust after the negative article. These analyses revealed that the perception of the project as fulfilled or not had strong effects on support and trust, and further, they also took out the effects of the treatment on support for the New Party. The fact that the effects of the treatment on support disappeared when perception of the project as fulfilled or not was included in the model, indicate that perception of the project as being fulfilled or not serve as mediator of the effects of persuasive rhetoric, on support. Since the treatment had effect on the respondents perception of the project as fulfilled or not, the likely conclusions to be drawn is that persuasive rhetoric have indirect effects on support post negative exposure to policy outcomes, in the way that it affects the perception of the project as being fulfilled or not. 21

22 Figure 1 below illustrates the causal relationship between persuasive rhetoric in response to negative information about policy outcomes, the perception of the promise as being fulfilled or not, and support for the party in power. The causality that is indicated by the experimental results, is that persuasive rhetoric affects the perception of the promise as being fulfilled or not, which in turn affects support for the party in power. The analyses indicate that there are no direct effects of persuasive rhetoric on support for the party, meaning that perception of the project as fulfilled or not fulfilled is what actually determine the respondents support for the party. Figure 1. Causal relationship between post-election rhetoric, perception of promise as being fulfilled or not, and support for the party in power. Persuasive rhetoric as response to negative media exposure Perception of promise as being fulailled or not Support for the party in power Concluding discussion The paper started out with the question whether declining support for parties in power as due to negative media reports, can be reduced by persuasive rhetoric. Previous studies have shown how growing negativity in news media coverage contributes to cynicism towards politicians, and lacking support. Since declining political support is negative in many aspects, questions on how governments can handle negative information about their activities and policy outcomes become relevant. If the political leadership lacks persuasive skills and is incapable of explaining their policies in a way that can be accepted by voters, negative information about governmental performance from actors such as the media will remain unchallenged, leaving them as catalytic for political events. The point of departure in this paper were that governing parties needs to engage in an ongoing dialogue with their constituents, and explain their post-election actions to the citizenry in a trustful and convincing way. Using theories from the research fields philosophy of language and linguistic semantics, and applying them to a political communication context in a vignette experiment, the paper tested effects of persuasive post-election rhetoric on perception of promises as being fulfilled or not, and support and trust for the parties in power. 22