Priming Issues, Party Visibility, and Party Evaluations: The Impact on Vote Switching

Voter volatility has become a hallmark of Western democracies in the past three decades. At the same time short-term factors—such as the media’s coverage of issues, parties, and candidates during an election campaign—have become more important for voters’ decisions. While previous research did look at how campaign news in general affects electoral volatility in general, it has omitted to explicitly test the mechanisms underlying these effects. Building on theories of agenda setting, (affective) priming, and issue ownership, the current study aims to explain why certain news aspects lead voters to switch their vote choice. We theorize it is the visibility of a party, the evaluation of a party, and the attention for issues owned by a party that primes voters to switch to a certain party. We use national panel survey data (N = 765) and link this to an extensive content analysis of campaign news on television and in newspapers in the run up to the 2012 Dutch national elections. The results show that issue news leads to vote change in the direction of the party that owns the issue. Even stronger is the effect of party visibility on vote switching. Our results, however, find the strongest support for the effect of party evaluations on vote change: More favorable news about a party increases switching to that party.


Introduction
In the past decades the number of floating voters in Western democracies has increased considerably.The linkages between parties and citizens are of declining importance (Dalton & Wattenberg, 2000) and voters make more independent political choices instead of relying on traditional loyalties (Van der Meer, Van Elsas, Lubbe, & Van der Brug, 2013).As a consequence, citizens switch more often between parties than they did before, not only between elections, but also during election campaigns.With the decline in voter loyalty, sociological determinants have lost much of their predictive power (Dalton, 2013).Instead, short-term factors such as media attention for and media evaluations of parties and candidates have become more important for voters' decisions.
More recent studies have investigated whether different types of campaign coverage, most notably issue news and contest or strategic news, have an effect on individual-level vote switching (Adriaansen, Van Praag, & De Vreese, 2012;Takens, 2013).Adriaansen and colleagues (2012) found that issue news can induce voter uncertainty, whereas strategic news decreases voter uncertainty.This voter uncertainty would subsequently spark electoral volatility.In contrast, the study by Takens (2013) showed that voters become volatile in response to contest news, while issue news leads to stability in vote preferences.In this study we argue that prior research, by focusing on how campaign coverage in general affects volatility in general, has omitted to explicitly test the processes underlying these effects.
For issue news, theories of issues voting (issue ownership) are expected to explain the effect on vote switching.Rather than issue news in general affecting volatility in general, it are specific issues in the media, owned by specific parties, that lead to vote switching to a specific party.The missing link between issue news and volatility is thus the party: the impact of issue news on vote switching is dependent upon the issue discussed, because different issues are perceived to be owned by different parties.These assumptions have been tested on the aggregate level, looking at vote choice (Kleinnijenhuis & De Ridder, 1998;Kleinnijenhuis, Van Hoof, Oegema, & De Ridder, 2007;Sheafer & Weimann, 2005), but not on the individual level focusing on vote switching.
Contest or strategic news in itself also does not explain why voters might change their vote choice (as is assumed in Adriaansen, et al., 2012;Takens, 2013).Again, for these types of news the missing link between news coverage on the one hand, and vote switching on the other, is the party.It is the positive or negative portrayal of a specific party or candidate that causes voters to consider switching to that specific party (Balmas & Sheafer, 2010).Therefore, this study focusses on how the positive and negative evaluations of a party in the news affects vote change.
By focusing on issue news and contest or strategic news in general affecting volatility in general, prior work has neglected an essential explanation for vote switching.If the link between campaign coverage and vote switching is the party, then one should also take into account the visibility of each party in the news.This is even more important as previous research has shown that party visibility is a pre-condition for electoral success (Hopmann, Vliegenthart, De Vreese, & Albaek, 2010;Oegema & Kleinnijenhuis, 2000;Semetko & Schoenbach, 1994;Vliegenthart & Van Aelst, 2010).The current study builds further on the existing body of research studying the relationship between campaign coverage and vote switching, by testing the impact of issue news, party visibility and party evaluations in one model.We argue that it are the media that set the public agenda by paying attention to issues, parties and evaluations of these parties (McCombs & Shaw, 1972).This primes voters to switch their vote choice to a certain party, leading to electoral volatility (Iyengar & Kinder, 1987).
In this study we combine panel survey data and content analysis data to test the differential impact of campaign news on voters with varying levels of media exposure.
In the analysis we employ a model in which the voter-party combination is the unit of analysis, which allows us to examine at the individual level which voters change to which party.In the next section we will explicate how media content might affect vote change, linking the literature on voting behavior to political communication theories.

Issue news and vote switching
To understand how issue news affects vote switching at the individual level we first need to understand how news about issues affects vote choice in general.We propose that agenda setting and priming theory, which address the salience of issues in the media and in voters' minds, should be combined with issue ownership theory.First, according to agenda setting theory, the media influence which issues voters recognize as most salient by emphasizing certain issues over others, making these issues more accessible in one's memory (McCombs & Shaw, 1972).Secondly, according to priming theory, individuals use the issues that are most salient and accessible in their memory to evaluate the performance of political actors (Iyengar & Kinder, 1987).Thirdly, according to issue ownership theory, citizens will vote for the party they consider to be the owner of the most salient issue -that is, the party they consider to be best in dealing with the issue (Budge & Farlie, 1983;Petrocik, 1996).
Broadly defined, issue ownership refers to the idea that voters consider some parties better able to deal with specific issues than others.The party with the best solution or track record for a given issue is considered to be the 'owner' of that issue, issues being for example employment, immigration, or health care.Since parties communicate their issue emphasis via the media (Walgrave & De Swert, 2007), one should not only take into account which party owns which issue, but also which issue received most emphasis during a campaign in order to empirically test all facets of the issue ownership theory.Thus, issue attention in news coverage should also be included in the issue ownership voting models.To our knowledge this approach has only been applied by Kleinnijenhuis et al. (2007), who mainly conducted analyses at the macro level.In the current study we test our model at the micro level to understand the effect of issue news on individual-level vote change, explained by issue ownership theory.We test agenda setting and priming implicitly by examining whether media attention for owned issues induces vote switching.
Theoretically, we expect issue news to be related to electoral volatility based on recent research showing that issue ownership also has a volatile component (Walgrave & De Swert, 2007;Walgrave, Lefevere, & Nuytemans, 2009;Walgrave, Lefevere, & Tresch, 2014).The stable component comes from parties' distinctive constituencies (Stubager & Slothuus, 2012).Yet, parties can also selectively (de)-emphasize certain issues, to re-activate old issue ownership, or to compete over new issues or steal issues from other parties (e.g., Damore, 2004).This dynamic part is affected by news coverage of issues and parties.A study on the sources of issue ownership showed that while party programs affect long-term perceptions of issue ownership, media coverage has a more immediate effect on issue ownership perception (Walgrave & De Swert, 2007).Additionally, recent experimental research (Bos, Lefevere, Thijssen, & Sheets, 2016;Dahlberg & Martinsson, 2015;Tresch, Lefevere, & Walgrave, 2015;Walgrave et al., 2009) has shown that news messages can affect issue ownership perceptions and vote preferences on the short-term.These findings are confirmed by a longitudinal study of the Dutch 2010 election campaign, which shows that "issue ownership is rather unstable at the individual level, due to news attention peaks for specific issues" (Kleinnijenhuis & Walter, 2014, p. 241).
Based on these findings we can assume that this volatility in issue ownership perception, at least partly, explains the volatility in party preferences.Although this is the underlying theoretical assumption explaining the relationship between issue ownership and electoral volatility, especially in the long run, this study captures issue ownership perceptions at one point in time in order to explain the impact of issue news on individual-level vote switching.We argue that this impact is dependent upon two factors.First, the amount of attention media devote to an issue during the campaign influences the issue salience in voters' minds.Since issues come and go more rapidly in the news, the media attention for issues, as well as the issue salience in voters' minds has become more volatile (McCombs & Zhu, 1995;Zhu, 1992).And secondly, the voter's perception of issue ownership determines which issue is perceived to be owned by which party.The issue owning party gains electoral success when the voter is exposed to this issue in the media.This leads to our hypothesis that: H1: Exposure to a party's owned issue increases switching to that party.

Parties in the news and vote switching
In the introduction we argued that if the link between campaign coverage and vote switching is the party, one should also take into account the portrayal of each party in the news.According to priming theory, the most salient (attributes of) parties or candidates become important criteria for party evaluation (Kim & McCombs, 2007).Previous work on the impact of the media on party choice indeed shows that visibility, of parties and their leaders, affects party preference or support positively (Hopmann et al., 2010;Oegema & Kleinnijenhuis, 2000;Semetko & Schoenbach, 1994;Vliegenthart & Van Aelst, 2010).Following Hopmann et al. (2010) we argue that especially in a multiparty context the visibility of politicians or parties can be the straw that breaks the camel's back.A certain amount of visibility makes voters aware of a party and leads them to consider this party as a viable alternative.Even though the impact of visibility might be dependent on other factors (Bos et al., 2016;Hopmann et al., 2010), visibility is a pre-condition for being elected.This leads to the following hypothesis: H2: Exposure to news about a party, increases switching to that party.
Although priming causes the most salient party in the news (agenda setting) to be important for a voter's evaluation of a party, it does not take into account whether the evaluation of that party is positive or negative (Balmas & Sheafer, 2010).Sheafer (2007), therefore, introduced the concept of affective priming, which adds an affective dimension to priming.This notion entails that when a party (or candidate) is primed, people will evaluate that party based on the affective evaluation they attach to that party.This affective priming is in turn influenced by media content through the affective compelling arguments effect (McCombs & Ghanem, 2003;McCombs, 2004), in which journalists attach an evaluative tone to a party or candidate in the news.Consequently, it is very likely that a voter's evaluation of a party is influenced by the tone towards that party in the media.
The evaluative tone in the media can be attached to several aspects of the news coverage of parties: either to the standing in the polls, to performances in debates or to candidates' personal traits.For each of these news characteristics, it are the media's evaluations of a party and, consequently, the affective priming in a voter's mind that predicts if the evaluation of a party will be positive or negative.
Moreover, voters may rely on information about the evaluation of a party as a cognitive shortcut to make a voting decision without engaging in rational cognitive processing.As such, positive and negative evaluations serve as peripheral cues that can lead to a change of vote, without actively thinking about the parties and issues under consideration (Lodge, McGraw, Conover, Feldman, & Miller, 1991;Petty & Cacioppo, 1986;Zaller, 1991).Therefore we expect that: H3: Exposure to favorable news about a party increases switching to that party.

Method
The Dutch case Testing these hypotheses requires: (1) individual-level panel survey data, to assess whether respondents change their party choice during an election campaign; and (2) content analysis data, to distinguish between different content characteristics in campaign coverage.We focus on the Netherlands, a democratic corporatist country with a multi-party system and a public broadcasting system (Hallin & Mancini, 2004;Mair, 2002).This context is relevant for this study as ideological differences between parties in The Netherlands are small (Roberts & Wibbels, 1999) and parties' positions on certain issues are often fairly similar.Therefore, switching is likely to occur and both issues and party evaluations in the news are likely to induce vote change.In contrast to the US, the stability of vote choice in the Netherlands has decreased considerably since the 1960s (Mair, 2008), like it has in more Western democracies.As such, the Dutch context is an appropriate setting to test our hypotheses on the impact of media content characteristics on vote switching.
In this study, we focus on the 2012 Dutch parliamentary election campaign.In total, 21 parties participated in the elections on September 12, 2012.For 14 of them vote change was practically non-existent, which is why we include only the seven major parties (Liberal party, Labour party, Socialist party, Freedom party, Christian Democratic party, Democratic party, Green party) in the analysis.

Panel data
Sample.The panel survey dataset we used was collected by TNS NIPO in collaboration with the University of Amsterdam and de Volkskrant using computerassisted self-interviewing.These data were gathered in the campaign running up to the 2012 Dutch parliamentary elections of September 12. Initially, 2250 respondents were approached in May 17, 2012 (t-4: N = 1,537; AAPOR RR4 = 68.3%),and recontacted June 21 (t-3: N = 1,239; recontact rate AAPOR RR4 = 80.6%) , August 16 (t-2: N = 1,206; recontact rate AAPOR RR4: 78.5%), August 30 (t-1: N = 1,187; recontact rate AAPOR RR4: 77.2%) and September 14 (t: N = 1,162; recontact rate AAPOR RR4: 75.6%).i In this study we only included the respondents that participated in all waves (N = 765).ii We only used the data of the last two waves (t-1 and t), since we are interested in the influence of the media's campaign coverage which only started after t-2.Of the 765 respondents 66% remained loyal to the same party between the two waves, 24% switched their party preference, and 10% eventually did not turn out on Election Day.Our data are by and large representative of the Dutch population.iii Measures.Our dependent variable is vote switching, which is based on one variable in the panel dataset measured at two points in time.At t-1 respondents were asked which party they would vote for if elections were held today.At t, the post-election wave, respondents were asked which party they ended up voting for in the elections.For each respondent, we constructed 7 dummy variables, one for each party.The party dummy variables were assigned a '0' by default, and a '1' if the respondent switched to that party between t-1 and t. iv Our main independent variable in the panel dataset used to link the panel data to the media content data is media exposure.We asked respondents about their exposure to the various media outlets that are also included in the content analysis: "Can you indicate how often you read the following newspapers?","Can you indicate how often you watch the following television programs?".These media exposure variables were measured on a 5-point scale ranging from never (0) to (almost) daily (4).v Competence issue ownership was measured at t-1 for 9 issues (employment, immigration, safety, social security, health care, housing, European Union, education, environment) with an item that asked which party has the best solution for issue X according to the respondent.This measure of competence ownership is also used by other scholars (e.g., Lachat, 2014).
We also included several control variables, starting with the usual sociodemographic variables, measured at t-4: age (M = 51, SD = 17), sex (49.7% male, 50.3% female), and education (measured in 7 categories ranging from 'no education/primary education' to 'bachelor degree or higher', M = 4.07, SD = 1.81).In addition, we controlled for various individual predispositions measured at t-2.First, ideological proximity is measured by calculating the difference between each respondent's own left-right placement and left-right placement of each party as perceived by the respondent (1 = distant and 10 = close, M = 7.20, SD = 2.45).Second, political interest, which was measured with an item that asked respondents how interested they are in politics on a 7-point scale (1 = not at all interested and 7 = very interested, M = 4.35, SD = 1.66).Finally, we control for partisanship as the tendency to label one (preferred) party as issue owner on every issue by including a dummy variable which is assigned a '0' by default, and a '1' if the respondent mentions the same party as issue owner on all 9 issues.

Content analysis
Sample.We conducted a content analysis of the last three weeks of the election campaign on television programs and newspapers (August 22 to September 12, 2012).vi All items with political content were coded in collaboration with the Dutch public broadcasting agency (NPO), by a team of four coders.In this study we only included those media outlets for which media exposure was tapped in the panel dataset (i.e., the most used media outlets in the Netherlands).Those are the news programs of the public broadcaster NOS Journaal, and two commercial stations RTL Nieuws and Hart van Nederland, the current affairs programs Eén Vandaag and Nieuwsuur, the talkshow Knevel and van den Brink (all public broadcasts), and the infotainment programs De Wereld Draait Door, PowNews and RTL Boulevard (only the latter is a commercial broadcast).For the newspapers we included two broadsheet/elite newspapers: de Volkskrant and NRC Handelsblad, two semi-tabloid newspapers: de Telegraaf and Algemeen Dagblad, and two (popular) free dailies: Sp!ts and Metro.Items were coded that satisfied the conditions of campaign news, in the sense that the story was about the elections, party leaders, or about the government.Items were identified based on content and form.
Measures.For the operationalization of issues, party evaluations and party visibility in the news we largely follow the suggested coding instructions by Hopmann, Van Aelst and Legnante (2011).
For issue news we coded all issues mentioned in a news item.For each item, the coders had to code whether an issue was mentioned or not.Coders could choose from a list of 22 issues.Some of the issues in the content analysis data were merged or renamed, so that they could be linked to the issues in the survey data.Issues that were coded in the content analysis, but could not be linked to the issues in the survey data are not included.These are issues that received little to no attention, like culture, infrastructure, agriculture.Eventually, 9 issues were included in the analyses: employment, immigration, safety, social security, health care, housing, European Union, education and environment.The average Krippendorff's Alpha across these issues was .74. vii The total issue attention per issue is the sum of all mentions across all items within a media outlet, see table 3.1.
Party evaluations are coded for all actors (max.8) in an item.We coded evaluations of parties, as well as evaluations of all party representatives.The evaluation in an item could either be neutral or mixed (coded as 0), unfavorable (−1), or favorable (1).The intercoder reliability for this measure is sufficient, with a Krippendorff's Alpha of .74.We use the average party evaluation across all items within a media outlet, see table 3.2.
Visibility is also coded for all actors (max.8) in an item.A party's visibility is relative to all other parties in an item and relative to the other content in an item.Coders were first instructed to code which part of an item is about all actors combined, ranging from 'very little' to 'very much'.In a second step coders had to code the share of attention for each actor.The Krippendorff's Alpha for this measure was .80.We use the average visibility across all items within a media outlet, see table 3.3.

Linking survey data to content data
For each respondent, the media exposure variables in the panel survey data were weighted on the basis of the media content variables in the content analysis data, computing individual exposure to issues, party evaluations and party visibility in the news (following e.g., De Vreese & Semetko, 2004;Schuck, Vliegenthart, & De Vreese, 2016): These weighted media exposure variables were thus determined by the media outlets each respondent uses and by the average attention for party evaluations and party visibility and the total attention for each issue in each outlet.viii The total exposure to issues, party visibility as well as party evaluations was divided by the total amount of media outlets included in the analysis.ix For example, a respondent reads the newspaper NRC Handelsblad on a daily basis (media exposure = 4) and the issue 'health care' is discussed 10 times in NRC Handelsblad.We link this issue attention to the respondent by multiplying the issue attention score for health care with the respondent's media exposure score for NRC Handelsblad, resulting in an issue exposure score of 40.This step is repeated for each of the 9 issues, each newspaper and each television program (total of 15 outlets).Subsequently, for each issue, the issue exposure score is first summed for all 15 media outlets and then divided by 15 to obtain an average media exposure score.
For the issue exposure variables a second step was added in which they were linked to the issue ownership variables.First, for each respondent and each issue, each party gets a separate competence issue ownership score (0=not owner, 1=owner).Second, for each issue X, the amount of exposure to issue X is linked to the party who is owner on issue X.This means that the average exposure to health care in the example above is linked to the party that owns the issue of health care according to the respondent, and is zero for all the other parties.Next, an aggregated issue variable was constructed, because the separate issue variables were highly correlated.This issue variable is created per party by taking the sum of issue exposure for all issues that a party owns.For example, out of a possible 9 issues, a respondent perceives the Liberal party as the owner of 6 issues and the Labour party as the owner of 3 issues, whilst the other five parties do not own any issue.For the Liberal party the exposure to the 6 issues is summed and for the Labour party the exposure to the 3 issues is summed.For the other five parties the value for the issue exposure variable is zero.

Data analysis
In order to estimate a model across parties, we created a stacked data set, so that the unit of analysis is the respondent-party combination.As a consequence, each respondent is included 7 times in the data set, one time for each party included in the analysis.In order to correct our estimates for possible errors introduced by the duplication of observations in the stacked data set (Steenbergen & Jones, 2002), we employ logistic regressions with clustered standard errors (respondents being the clusters).

Results
We begin our results section with a description of the media variables.Table 3.3 shows the average party visibility in the news for the included television programs and newspapers.Overall, party visibility is slightly higher on television than in newspapers, and across the different outlets the Liberal party (the incumbent party) is the most and the Green party the least visible.Table 3.2 shows the average party evaluations in the news for the television programs and newspapers.Overall, the tone towards parties is more negative than positive, and across the different outlets the Democratic party is most positively and the Green party most negatively evaluated.
As Table 3.2 and Table 3.3 demonstrate, both visibility and evaluations vary between parties as well as between media outlets.Therefore, we can assume that the degree of party visibility and party evaluations that individuals are exposed to depends on the media outlets they use.
Table 3.1 shows the total attention for each issue in each media outlet.In general, the European Union issue receives the most attention across all outlets.The second most salient issues in the media are employment, social security and health care.The issues immigration, safety, and the environment receive the least attention.
However, there are differences in issue attention between the various media outlets.Overall, newspapers pay more attention to issues than television programs, with the infotainment programs paying least attention to issues.More specifically, although the issue attention for immigration and safety is generally low, these issues are quite salient in the tabloid newspapers.Additional analyses on issue importance in voters indicated that the issue salience in the media largely matches the issue salience in voters.Only for the European Union we find a discrepancy: although the European Union receives a lot of media attention, it is one of the least important issues for voters.Note.Average value of party evaluations in each medium.Scale runs from -1 to 1, where -1 denotes 'unfavorable', 0 'neutral' and 1 'favorable'.Table 3.4 shows voters' perceptions on issue ownership at the aggregate level.It shows that there is no general agreement about issue ownership on most issues, except for the environment issue, which is clearly owned by Green party.Table 3.4 also shows that the incumbent Liberal party is on average most often perceived as issue owner.
The disagreement about issue ownership might partly be due to the endogeneity with party preference.A correlation analysis between how often respondents mentioned one and the same party as issue owner on multiple issues and party preference (measured as propensity to vote) showed a correlation of .61.This supports our decision to control for the tendency to name one (preferred) party as the owner of almost any issue.Table 3.5 presents the effects of issues, party visibility and party evaluations in the news on electoral volatility x .Model 1 includes only the socio-demographics and individual predispositions, which are assumed to be related to vote switching.We find that when people are ideologically close to a party, they are more likely to switch to that party.
In model 2, party visibility and issue news are added.We find a significant positive effect of issue news on vote switching.This confirms hypothesis 1, which stated that when one is exposed to a party's owned issue, one is more likely to switch to that party.We also find a significant positive effect of party visibility on vote switching, which confirms hypothesis 2.More exposure to news about a party, increases switching to that party.Partisanship has a negative effect on vote switching, meaning that people who name one preferred party as issue owner on every issue are more stable in their electoral choice.This effect is marginally significant in model 2 and significant at the .05level in model 4.    Model 3 includes party evaluations and party visibility in addition to the control variables.In this model the effect of visibility on vote change is still significant.Party evaluations also have a significant positive effect on volatility, which confirms hypothesis 3.More exposure to positive news about a party increases switching to that party.Figure 3.3 illustrates the marginal effect of party evaluations on vote change.Among respondents with an average level of exposure to party evaluations (M = -.07,SD = .09)the predicted probability of vote change was .03.By comparison, among respondents whose exposure to positive party evaluations was two standard deviations above the mean the predicted probability of vote change was .07.The full model, model 4, includes all three media variables -issues, party visibility and party evaluations-in addition to the control variables.xii In this model the effects of all three media variables remain significant: issues, party visibility and party evaluations do not cancel each other out when modeled together.We compare the standardized coefficients of the media variables within the full model to judge the relative importance of the media variables.The final model shows that the standardized coefficient for party visibility (0.330) is higher than the standardized coefficient for issues in the news (0.164).The graph in figure 3.2 also shows a steeper increase for the marginal effect of party visibility on vote change as compared to the marginal effect of issue news on vote change, which is displayed in figure 3.1.Party visibility in the news is thus a stronger predictor for vote switching than the news coverage of a party's owned issues.Additionally, visual inspection of figure 3.1-3.3suggests that party visibility and party evaluations play a larger role than issue news.The standardized coefficient for party evaluations (0.479) is higher than the standardized coefficient for visibility (0.330) and even more than twice as high as the standardized coefficient for issues in the news (0.164).This indicates that positive coverage of a party is even more important for changing party preference than mere visibility in the media.

Discussion
The stability of voters' electoral choices has declined over the past decades and short-term factors like media coverage on issues, parties and candidates during the campaign have become more important for voters' decisions.Previous studies have shown that different types of campaign news can induce change in voting intentions (Adriaansen et al., 2012;Takens, 2013).However, research focused on explaining how the different aspects of campaign news lead to volatile voting behavior is lacking.Building on theories of agenda setting, priming, affective priming and issue ownership, this article examined how issues, party visibility and party evaluations in the news lead to vote switching.

The impact of issue, party visibility and evaluations on vote switching
We found that media exposure to issue news, linked through issue ownership, leads to vote switching.Thus, (1) when an issue received much attention in the media which a voter is exposed to, and (2) a voter considers a certain party to have the best solution for that issue, he or she will probably switch to that party.The results corroborate the assumptions of previous studies on an aggregate level (Kleinnijenhuis & De Ridder, 1998;Kleinnijenhuis et al., 2007) and suggest that issue ownership indeed underlies the effect of issue news on electoral volatility.The finding that issue news leads to vote switching challenges earlier findings by Takens (2013), who concluded that issue coverage stabilizes vote choice.Voters do not necessarily stay with their party when exposed to issue news.This study shows that it depends on which issues dominate the news and which parties are perceived to be owner of these issues.
It should be noted that issue ownership does not only link issue attention in the media to voting behavior, but issue ownership itself might also be influenced by media coverage (Kleinnijenhuis & Walter, 2014;Walgrave & De Swert, 2007).Our content analysis of three weeks campaign news is too short to examine how issue ownership functions both as an outcome variable of media coverage and as an explanatory variable for voting behavior.Examining this causal chain might be an interesting step for future research.
With regard to evaluative news coverage about parties, we found that party evaluations in the media indeed have an impact on electoral volatility.Exposure to favorable news coverage about a party makes a voter more likely to switch to that party.We also found that more party visibility leads to more vote switching, which supports the traditional agenda setting and priming hypothesis.This is in line with previous research which showed that party visibility affects party preference positively (e.g., Hopmann et al., 2010;Oegema & Kleinnijenhuis, 2000).Independent from other factors, like issues and evaluations, mere visibility has a strong impact on vote switching.This highlights the necessity to take into account the (visibility of a) party when studying the impact of campaign news on vote change.Party visibility in itself can affect voter volatility in two ways.First, voters switch to the most salient party in the news, as a consequence of priming.Second, in a multiparty system, when media pay attention to multiple parties, more parties become a viable alternative to vote for.
Still, the effect of party visibility was weaker than the effect of party evaluations.In line with Hopmann et al. (2010) we believe that visibility can change voters' decisions, if this leads to a process where they consequently evaluate the performance of the visible parties and candidates.We find, that it is the evaluative aspect attached to a party that has a stronger effect on voting for or against a party.Different from visibility, evaluations can function as a heuristic that gives voters an indication on whether to evaluate a party positive or negative and whether to vote for that party.As such, this study provides support for the notion of affective priming (Balmas & Sheafer, 2010;Sheafer, 2007).This is consistent with the general assumption (Adriaansen et al., 2012;Takens, 2013) that contest and strategic coverage, by evaluating the performance of parties, leads to volatile voter behavior.
From a party perspective, the findings of our study show that there is merit in employing campaign strategies to get a party's owned issues in the news and to increase ones' visibility in the media, in order to gain electoral success.Especially smaller parties profit during an election campaign, since they are more visible than they are in routine periods (Strömbäck, 2016).However, our results show that evaluative news has the most impact.It are journalists who choose to portray a party positively or negatively, which indicates a certain degree of media interventionism (Strömbäck & Dimitrova, 2011).As such, an important aspect of campaign coverage that induces vote switching is one that parties are less able to influence.
One can wonder how we should interpret our findings from a normative perspective.According to most normative democratic theories, citizens should be wellinformed on political issues and base their vote on their position on these issues and choose the party that best represents that position (i.e., Downs, 1957).Our finding that party evaluations have the strongest impact on vote switching could signal that volatile voters are indecisive and uninformed (Lazarsfeld, Berelson, & Gaudet, 1948), making random vote changes, using media's evaluation of a party as a heuristic to vote for or against a party.Yet, even if evaluations in the media function as a running tally, voters could still come to a 'correct' vote.Moreover, according to the 'interpreting' citizen model, voters can fulfill the expectations of democratic theory if these expectations are understood in terms of "citizens' ability to interpret political reality, as opposed to the demand of being well-informed" (Porto, 2007, p. 312).Whereas the informed citizen bases her or his vote on what she or he believes are the most important issues, the interpreting citizen uses evaluative information as an effective shortcut to interpret political reality and form preferences.
From a psychological perspective, it depends on how voters use issues or party evaluations in the decision making process (Lau & Redlawsk, 2006).Both can function as either a heuristic or substantive argument when deciding which party to vote for.Party evaluations in the news can be regarded as a peripheral cue if voters mindlessly vote for the party that is portrayed most positively (Lodge et al., 1991).On the other hand, it might as well be that voters are actually politically involved and use this information to cast a strategic vote (Irwin & Van Holsteyn, 2008).Likewise, voters who base their vote mainly on issue news, do not necessarily use this information substantively in the decision making process.An issue discussed in the news could merely function as a symbolic issue.In contrast to hard issues, symbolic issues are affectively charged and contain little substantive information on the possible desirable end states and the various policies to achieve these ends (Wojcieszak, 2014).For these issues voters are more guided by the moral value attached to the issue than the substantial arguments for or against a certain policy.
Meanwhile, this study methodologically advances previous research on the effect of media exposure on electoral volatility.While previous studies were mostly conducted on the aggregate level (Balmas & Sheafer, 2010;Kleinnijenhuis et al., 2007;Sheafer & Weimann, 2005), we were able to tell how voters switch their vote choice on a micro level by taking into account individual differences in media exposure, and by using individual measures of issue ownership.Yet, even though our rich data set allows us to test our hypotheses in a real life setting, we do acknowledge the difficulty of isolating the independent impact of the separate content characteristics with observational data.For instance, issue performance in the media is also part of evaluative news (Walgrave et al., 2009).Accordingly, there might be some overlap between our operationalization of evaluative news and issue news, causing the effect of issue news to be somewhat underestimated.Unfortunately, our data do not allow us to test to what extent the evaluative tone in the media refers to the issues discussed.Future research should therefore test our hypotheses in an experimental study in which the different elements (evaluations, issues and parties/candidates) are manipulated both in absence and presence of each other.Such an experimental design would also allow a test of the stages of agenda setting, (affective) priming, issue ownership and voting, and an investigation of the extent to which voters are involved in the rational cognitive processing of (non)-substantive information.In such a design one can also examine whether these effects are moderated by individual predispositions like age, education, partisanship and political interest (Hillygus, 2010).Finally, one can wonder whether a part of the large impact of party evaluations is tautological in the sense that more support for a party leads to more positive party evaluations.However, this study examines the change in vote preference between two waves.The content analysis of campaign news is conducted prior to the second wave of the panel survey.This temporal order indicates that party evaluations in the news are more likely to influence party support instead of the other way around.This does not mean that evaluations between the two waves are not affected by an increase in support prior to the first wave (see Green-Pedersen, Mortensen, & Thesen, 2015).Whether the relation between evaluative news and party support is reciprocal can be addressed in future studies.
The focus on the Dutch 2012 election campaign ensures high validity, the opportunity to do more in-depth research, and overcomes the problem of heterogeneity of cases (Gerring, 2006).The Netherlands is a typical example of Hallin and Machini's (2004) democratic corporatist media system and exemplifies a multi-party system with an open structure of party competition (Mair, 2002).Therefore, we are confident that our findings also extent to at least central and northern European countries.Future, comparative, studies could test this, and extend our research by including more than one election campaign, gaining more insight into the long-term changing relationship between media and political behavior.
In conclusion, this study adds important empirical evidence to our understanding of how specific aspects of campaign news affect electoral volatility.It extends earlier political scientific work on the same subject, which mainly focused on individual predispositions in explaining vote change and paid little attention to how people change their vote intention at the end of the funnel (Dassonneville, 2011;Kuhn, 2009;Van der Meer et al., 2013).As a consequence, campaign coverage as a possible explanation for vote switching has often been disregarded (e.g., Baker et al., 2006;Bybee et al., 1981).The results in this article clearly demonstrate that campaign news does matter.This study therefore not only contributes to research on electoral volatility, but also to the literature on campaign effects by showing, on the individual level, that people change their mind on which party to vote for as a result of exposure to campaign news (Erikson & Wlezien, 2012).

Notes
i For each wave, the initial 1,537 respondents were re-contacted.
ii Panel attrition does not seem to affect our findings.Most respondents dropped out between May and June.Those are probably the respondents who found it too much effort to participate in the whole panel survey.The respondents that we finally included in our study did not differ to a large extent from the dropouts on the most important variables, such as political interest and media use.iv Respondents who reported "refuse" and "no right to vote" on the dependent variable in the last wave were treated as missing.
v For newspaper exposure we also included exposure to newspaper websites.
vi Although we only look at vote switching from t-1 to t, it is likely that respondents were already influenced by campaign news that appeared before t-1.Therefore, we include campaign news as from August 22.Since the election campaign started later due to summer recess, we only use content analyses of the last three weeks of the campaign.
vii For certain issues the number of cases per issue were too sparse to calculate intercoder reliability properly.Two issues (income policy and public finance) were dropped from the analyses, because of low inter-coder reliability (values below .40).For two other issues (social security and education) the Krippendorf's Alpha was also fairly low.
Yet, we decided to keep these issues in the analyses as Holsti's method showed high inter-coder reliability (values above .90).As robustness check, we also estimated models with only 7 issues (excluding also 'social security' and 'education' from the analyses).
Results were similar to the ones reported in Table 3.5.
viii Although scholars are still debating on the most reliable and valid measure of media exposure, they agree that this measure of exposure per medium overcomes at least some of the limitations of conventional news exposure measures (for a more elaborate discussion see, Dilliplane, Goldman, & Mutz, 2013;Price & Zaller, 1993;Slater, 2004).
ix By employing the average exposure to media content instead of the sum, we control for potential overreporting of news exposure (see criticisms on self-reported news exposure measures, Prior, 2009).
x We also estimated models with a slightly different operationalization of electoral volatility: a model with vote choice as outcome variable and previous vote choice as control variable and a model with vote probability as outcome variable and previous vote probability as control variable.Results are similar.In addition, we estimated the same model for a smaller sample, including only the respondents who actually changed their vote between the two waves (N = 167).Results for media variables are similar.
Results for control variables are somewhat different: the impact of ideological proximity is no longer significant, education has a significant positive effect on vote switching and political interest has significant negative effect on vote switching.This also applies to the marginal effects of visibility and party evaluations based on model 3. xii We ran an additional analysis to test the interaction effect between political interest and the media variables.The findings revealed that the effect of party visibility and party evaluations on vote change are not moderated by political interest.We did find a small, but significant interaction between political interest and issue news: only voters with high levels of political interest do not change their vote when exposed to an average level of issue news.
xiii Standardized coefficients are obtained by using the command 'listcoef' in STATA.

Figure 3 . 1 :
Figure 3.1: The marginal effect of issues in the news on vote change.

Figure 3 .
Figure 3.1 illustrates the marginal effect of issue news on vote change.xi Among respondents with an average level of exposure to issue coverage (M = 2.09, SD = 9.19) the predicted probability of vote change was .03.By comparison, among respondents

Figure 3 .
Figure 3.2 illustrates the marginal effect of party visibility on vote change.Among respondents with an average level of exposure to party visibility (M = 12.01, SD = 9.83) the predicted probability of vote change was .03.By comparison, among respondents whose exposure to party visibility was two standard deviations above the mean the predicted probability of vote change was .06.

Figure 3 . 2 :
Figure 3.2: The marginal effect of party visibility in the news on vote change.

Figure 3 . 3 :
Figure 3.3: The marginal effect of party evaluations in the news on vote change.
iii A representative sample (2,250 persons) was selected.The respondent data of the 765 persons who completed the survey in all waves mirror census data by and large in terms of age, gender, and education.Older respondents (ages 65-80) are slightly overrepresented in our sample.

xi
Figure 3.1-3.3are based on the same model (model 4), enabling us to compare them to judge the relative importance of the media variables.The marginal effect of issue news (illustrated in figure 3.1) looks similar when estimated on the basis of model 2.

Table 3 .
1: Issue attention in the news (not related to actors).

Table 3 .
3: Party visibility in the news.

Table 3 .
5: Logistic regression for effect of issues, party visibility and party evaluations in the news on vote change.