A research into the influences of negative campaigning and populism on voting behavior and positive attitudes of Dutch citizens
Joel Larson (10772049)
Master Political Communication – Master thesis dhr. dr. R. (Rachid) Azrout – University of Amsterdam
Amsterdam, June 2020 – Word count: 7491
Abstract
The influence of negative campaigning during election times has been greatly researched in recent years. However, those researchers mainly focused on countries with two-party electoral systems, such as the U.S.A., also mainly focused on effects of negative campaigning towards the targets and not the sponsor, and furthermore, mostly focused on main effects of negative campaigning, in which possible moderating effects of negative campaigns were overlooked. This research focuses on all of this by: investigating the influence of negative campaigning on the support of citizens towards the target and sponsor, focusing on the negative campaigning in a multi-party system, namely the Netherlands, and looking into the possible moderation role of populism, which seems to go hand in hand with negative campaigning, according to recent literature. This research uses a 2x2 between subjects design, and consists of a dataset of 184 Dutch adult citizens. The main finding of this research is that seeing a negative campaign decreases the positive attitude of Dutch citizens towards the target of the campaign. Although the other findings were not significant, the results of “voting behavior” and “positive attitude” show similar findings, which shows there is some truth in these non-significant findings. Therefore, politicians in a multi-party system can create the lowest positive attitude and lowest voting behavior towards their opponents by creating negative campaigns combined with texts with populist elements. However, this also creates the greatest backlash effect for the sponsor. In order to gain the most votes, politicians should publish populist texts combined with neutral campaign videos. Next to politicians, also scientific researchers can benefit from this research. These results have important implications for the study of negative campaigns and populism in multi-party electoral systems. Furthermore, this research gives new insights into the role of populism and gives multiple possible ideas for future research.
Introduction
In the field of negative campaigning, in the past twenty years researchers have focused mostly on the effects that negative campaigning can have on the decreasing support of citizens for targets of the campaign (Lau, Sigelman, &; Rovner, 2007; Martin, 2004). However, there are theories that suggest negative campaigns could also increase the support of citizens for sponsors of negative campaigns (Carraro et Castelli, 2010).
These theories and results are mainly vivid in articles regarding the U.S.-presidential elections (Frimer and Skitka, 2018), which is logical because when the support for one presidential candidate goes down, then the support for the other is likely to go up. Many countries in the world however do not have a two-party electoral system. For the politicians and political parties in those countries it would be interesting to know if the usage of negative campaigns will decrease the support for the target and at the same time will increase support the sponsor of the campaign. If that will be the case, then politicians and political parties who use negative
campaigning could have great advantages during campaign-times. They could for example guarantee a victory by using negative campaigns. Also, using negative campaigning could have disadvantages: it could lead to decreasing support of citizens in a multi-party electoral system, which would mean that not the sponsor and target of negative campaigning, but other political parties receive the most votes. This is problematic for the sponsor and target of negative campaigns. Therefore, this research will look into the possible positive outcomes but also negative outcomes of using negative campaigns. In order to find results regarding countries with a multi-party electoral system, this Master thesis will focus on the Netherlands, which has multiple parties present in the political system. In order to measure the support towards the
sponsor and target of campaign videos, as Frimer and Skitka (2018) do by focusing on
“competence” and “warmth” this research will take a step further and will focus on the voting behavior and positive attitude of citizens towards politicians.
Many times, populism goes hand in hand with the usage of negative campaigning. Literature shows that populist politicians tend to criticize other politicians, because in order to create an in-group, they will try to exclude (dangerous) others, and in order to show themselves as “the voice of the normal man” they will criticize the elite establishment. And so, the elite are “usually
evoked in populist discourse as the source of crisis, breakdown, corruption or dysfunctionality”
(Moffitt and Tormey, 2014, page 391). Because of these vivid connections between populism and negative campaigning, when researching negative campaigns, it is interesting to also research if populism elements in messages might strengthen or might weaken possible effects of negative campaigns. This research intends to do this.
Also, the results of this research could enhance existing literature on negative campaigning in regard to a multi-party system, and specifically find out if a backlash effect is present. When a backlash effect is not found, then scholars could take this research into new directions in order to find out how the backlash effect differs between different electoral systems. Furthermore, this thesis will enhance existing literature by focusing specifically on the interaction effect between populism and negative campaigning.
Combining all of the information above leads us to the following research question of this Master Thesis:
RQ: What is the influence of online negative campaigns in the Netherlands by Dutch fictive
politicians on the voting behavior and positive attitudes of Dutch citizens for the target and sponsor
of the negative campaign, and what is the role of the amount of populism of Dutch fictive
politicians?
The concepts that are present in the research question will be operationalized in the next pages.
Theoretical Framework
Negative campaigning
As Lau and Pomper say in their article (2001, page 805-806), negative campaigning is “Talking
about the opponent, criticizing his or her programs, accomplishments, qualifications, and so on.
Positive campaigning is just the opposite: talking about one's own accomplishments,
qualifications, programs, etc.”
If a campaign can be named negative or not, depends on two features of the messages that should be present in the campaign; Firstly, the tone of the message has to be negative, meaning that something or someone is criticized. Secondly, the direction of the message has to be towards the opponent.
When the tone of the message is more negative than positive, we speak of a negative message (Hopmann, Vliegenthart, & Maier, 2018). A message can be operationalized as “a statement”, which is mostly one sentence or a part of a sentence. As Alessandro Nai (2018) explains in his article, there cannot be a range of how negative a campaign is, for example: a campaign cannot
be a 6 or 7 on a positivity scale ranging from -10 (negative) to 10 (positive). A campaign is either negative, neutral or positive. Therefore, a campaign is only negative when the proportion
negative messages is higher than the proportion positive messages in the campaign.
As to the direction of the message, if the negative message attacks the sender of the message, it is not a negative campaign, but it is a reflective statement. However, if the negative tone in the message criticizes the rivals of the sender of the message, then the message contributes to a negative campaign. A message has a negative tone when, for example, the ideas, character or program of the rivals are criticized.
In order to operationalize the term ‘negative campaign’, it is good to note that the amount of harshness or using vulgar language in messages does not contribute to the negativity of a message. “Bad manners” can indeed be used by politicians whom publish negative campaigns (Carraro et Castelli, 2010, page 392). However, these are not mandatory in negative
campaigning.
To sum it all up: Publishing negative campaigns means publishing more negative messages than positive messages, in which the rival is criticized, regardless of how harsh or vulgar the content is.
Types of negative campaigns
In most recent literature, there are two different types of negative campaigns mentioned: the policy attack campaigns, and the character attack campaigns (Brooks & Geer 2007; Carraro & Castelli, 2010; Min, 2004). In policy attacks, the sender of a political message is promoted by
attacking the opponents on policy issues. With character attacks the sender is promoted by attacking the opponents on their character or their persona. The usage of character attacks
messages instead of policy attacks messages in negative campaigns is more effective according to research of Brooks & Geer (2007). However, according to research by Min (2004), character attacks are also more likely to decrease voter-turnout.
While discussing the net worth of negative campaigning it is important to look into the intended effects versus the unintended effects of negative campaigns. This is important, because a
politician does not only want to decrease the affect for the campaign target, but also wants to increase the affect for him- or herself. However, as the research of Carraro and Castelli (2010) show, in which the sponsor of negative campaigning had a decreased perceived warmth, it could happen that a backlash-effect will appear. This means that citizens will like the politician whom sponsors negative campaigns less than politicians whom do not publish negative campaigns. This concurs with Min’s (2004) findings that negative strategies generally do not benefit the
sponsoring politicians. However, there is a way how this could still work: when a negative
campaign is published, and the target and sponsor both receive less support, but the support of the target goes down more, than the sponsor of the negative campaign has lost less support. This would be a win in a two-party system. However, in the Netherlands losing support in a multi-party system could mean that not the sponsor and target of negative campaigning, but other political parties will receive the most votes, which is problematic for the sponsor and target of negative campaigns.
The difficulty of conducting research into the field of negative campaigns is that many scholars have contradicting results. Recent research shows however that negative campaigns do indeed
work, but that the net effect of using negative campaigns depends heavily on the nature of the receiver of the campaign. The net effect of negative campaigns differs for receivers with a different amount of tolerance towards negativity (Fridkin & Kenney, 2011) and with different personal traits (Weinschenk & Panagopoulos, 2014; Nai & Maier, 2020). We see therefore in these examples that negative campaigning can work, but that there are still many different scenarios in which it will or will not work.
Voting behavior and positive attitude
Recent literature suggest that negative campaigns could change more perceptions than only attitude. As Frimer and Skitka (2018, page 846) explain, there are two main theories in literature, named the “Big Two” theories of personalities: the first is that people judge each other in terms of ‘warmth’ (or ‘liking the person’), the second is that people judge each other in terms of ‘dominance’ or ‘competence’. The research from Carraro and Castelli (2010) shows that this can indeed be the case. In their research sponsors of negative campaigns had a perceived less warmth but got more positivity in terms of perceived capability.
Frimer and Skitka (2018, page 846) say that “if social approval is the goal, it is better to be seen
as warm than it is to be seen as dominant. In this way, the benefits of coming across as warm
could overwhelm any costs of being seen as slightly submissive.”
They then explain that the cost for a politician for being uncivil is that people would perceive the politician as less warm, and the politician would therefore have a decreased social approval. At the same time, their research shows that Donald Trump could win the American presidential elections by being perceived as less warmth and more as dominant. Therefore, social approval should maybe not be the goal for politicians who would like to win votes.
In our research, both concepts of the ‘Big two’ are investigated. The research will focus on the voting behavior of Dutch citizens (competence) but will also look at the positive attitude (the warmth) people have towards the politician.
The voting behavior is defined as ‘a form of electoral behavior’. In the Netherlands, Dutch citizens can vote during election times on the political parties, and on the politicians. The voting behavior will tell us what Dutch citizens think about the governing competence and capacity of the politician. As a result of the research by Frimer and Skitka (2018) and Carraro and Castelli (2010), this research assumes that using negative campaigns will increase the voting behavior towards the sponsor and decrease the voting behavior towards the target. Furthermore, as research by Min (2004) and Lau, Sigelman and Rovners (2007) suggests, negative campaigns will decrease the positive attitude towards the target but will probably also have a backlash effect on the sponsor of the negative campaign.
Following all of the information mentioned above, this Master thesis proposes the following four hypotheses:
H1A: Dutch citizens who see a negative campaign towards a targeted politician instead of a
neutral campaign will be less likely to vote on the target of the negative campaign.
H1B: Dutch citizens who see a negative campaign towards a targeted politician instead of a
H2A: Dutch citizens who see a negative campaign towards a targeted politician instead of a
neutral campaign will have a lower positive attitude towards the target of the negative campaign.
H2B: Dutch citizens who see a negative campaign towards a targeted politician instead of a
neutral campaign will have a lower positive attitude towards the sponsor of the negative
campaign.
Influence of populism
Populism is a multidimensional concept, which changes regarding who you ask about it. As Engesser, et al (2017), explain: when talking about populism researchers tend to go into the ideology, the actors, the style and the strategy. From an ideological perspective, Mudde (2004) came up with a minimalist definition of populism by saying that populists are like socialists, and that if populism is an ideology, then we can measure it.
The core idea of populism ideology became the following: “The world consists of the pure
people and the corrupt elite. Next to this, populism promotes and protects the in-group against
the out-groups” (Mudde, 2004).
The ideology of populism can be seen as thick ideology, namely when the message contains anti-elite features and promoting-the in-group features. Ideology of populism can also be seen as thin ideology, meaning that only one of these features is present. This is called thin ideology, because the ideology of populism is like a thin coat that people can put over other ideologies that they already have (Mudde, 2004).
Next to ideology, populist features can also be found in the style of politicians, and some
scholars, such as Jagers an Walgrave (2007), actually say that populism can be best defined as a style itself. Ernst, et al (2019), also looked into this and researched different styles that populist
leaders would use. They conclude that politicians could use populism as a style, and that the three main populist styles are: negativity, emotionality and sociability.
Because populist politicians tend to use negativity in their messages in order to attack the out-group and elites, populist politicians will use negative campaigns more than mainstream politicians. Also, it could be argued that citizens who are used to seeing negativity in the messages of populist politicians, will find it ‘a better fit’ when populist politicians use negative campaign instead of mainstream politicians using it. Therefore, this research suggests that the usage of negative campaigning will have more effects on citizens when the politician who sends the message is a populist, instead of a mainstream politician.
Next to ideology and style, populism is also seen as a feature of contemporary political discourse, regardless of the ideology of the actor. Populists can use mainstream language, and also mainstream candidates can use populists appeals. To do further research on populism, we should use an alternative approach in which populism is seen as discursive feature. (Nai, 2018) Populism becomes a communication frame in this alternative approach. In this approach, we look into three main components of populist communication: 1. It appeals to the people (horizontal: in-group versus out-group), 2. It is anti-elitism (vertical), and 3. There is usage of simple language and it is anti-intellectualism (Nai, 2018). The usage of all three of these components results in the usage of ‘Full’ populism as a communication frame. All three components will be used in this research to investigate if using populism could strengthen the influence of negative campaigning.
Looking at the relation between populism and negative campaigning, the following four hypotheses are formulated:
moderate the effect of a negative campaign on the voting behavior towards the target of the
negative campaign.
H3B: Dutch citizens who see a text with a higher amount of populist rhetoric will positively
moderate the effect of a negative campaign on the voting behavior towards the sponsor of the
negative campaign.
H4A: Dutch citizens who see a text with a higher amount of populist rhetoric will positively
moderate the effect of a negative campaign on the positive attitude towards the target of the
negative campaign.
H4B: Dutch citizens who see a text with a higher amount of populist rhetoric will positively
moderate the effect of a negative campaign on the positive attitude towards the sponsor of the
negative campaign.
Figure 1: Conceptual model of this research
+
Negative campaigning
Voting for the target of the campaign video Voting for the sponsor of the
campaign video
Attitude towards the target of the campaign video
Populist sponsor
Attitude towards the sponsor of the campaign video
-+
+
-
-+ + --+
+
-
+ H1B H1A H2A H2B H3 & H4Research Design and Method
Method
In this research the possible causal relations between the independent variable “seeing a negative campaign video” and dependent variables “voting behavior” and “positive attitude” towards politicians are investigated. When investigating possible causal relations, conducting an experiment is the best research method, because by conducting an experiment the value of independent variables are manipulated, so that it can be investigated if causal effects on
dependent values are present (Boeie et al., 2005). Therefore, in this Master thesis an experiment was conducted. In the following section a description is given regarding the data and materials that were created for measuring the relevant concepts in this Master thesis.
Design
An online experiment was conducted, which was a 2 x 2 between-subjects design. The four different conditions are shown below in table 1. The respondents did not do the research before, therefore they would not get bored while participating in the survey. Because different conditions were created, and the respondents were randomly assigned to the four different groups, there could be no assignment bias.
Table 1. Between- Subject Design with one control group; group 1
Neutral text Group 1: Respondents read a neutral text and viewed a neutral campaign video
Group 3: Respondents read a neutral text and viewed a negative campaign video Populist text Group 2: Respondents read a
populist text and viewed a neutral campaign video
Group 4: Respondents read a populist text and viewed a negative campaign video
Respondents
In June 2020, the distribution of the online survey experiment of this research was done through the usage of database Qualtrics. The participants were Dutch adults, who received the invitation to participate in the survey through personal Whatsapp- and Facebook-messages from the
researcher. Also, the method of snowballing and the online website “Surveyswap”, were used for this. The advantage of filling in the questionnaire online was that participants would have the surroundings they were familiar with, because they were in their own habitat. Because of these realistic surroundings, the external validity was high.
A total of 211 respondents participated in the online survey. These participants were randomly divided into the four conditions, mentioned in table 1. Out of all 211 respondents, two
respondents filled in that their Dutch was less than ‘good’. In order to get valuable results, it was decided that those two respondents were not included in the analysis.
Also, it was investigated how much time respondents spent on the research. The researcher investigated how fast the test could have been filled in by a serious respondent, which was six
minutes. It was decided that all respondents who spent less time on the survey were excluded, which were five respondents.
Therefore, the results of a total of 184 respondents were used for the analyses in this research.
Key concepts and measurements
Fictive politicians
The usage of fictive politicians in this Master thesis means that multiple persona were created from scratch and were used in the experiment as the main political characters. The usage of fictive politicians instead of real politicians has multiple reasons: When conducting an experiment in which real politicians are the main character, respondents might know the politicians already and therefore can have assumptions regarding these politicians that could influence the results. This will not happen with a fictive person.
A downside of the usage of fictive politicians is that the external validation of the experiment goes down, because the results cannot be generalized to all politicians.
When we look at the internal validation, we see that because of the usage of fictive politicians the internal validation of the experiment goes up. When certain results are found in the experiment, we can be sure that these results are valid. This would not have been the case when a real politician was present in the experiment.
Also, the Master thesis looks into the effects of the amount of populist features that are published by a politician. By using fictive politicians, the experiment does not have to go into questions such as: is the politician right-wing, left-wing, or in the middle of the political spectrum? By
using fictive politicians, the manipulation of the usage of a populist fictive politician, and a non-populist fictive politician will be more likely to succeed, because the attention should not be on left-wing, right-wing, or the middle, but only should be on the presence or absence of populist features that are used by the fictive politician.
Independent variable: Condition-group
During the experiment, the 184 respondents were divided into four conditions. The respondents in every group read a text about the fictive new politician Mark Timmers, who would run in the city-elections. After this, they viewed a fictive campaign video, sponsored by Mark Timmers. For “condition text”, a interview with Mark Timmers was written down, which resulted in the
“neutral text”. Multiple sentences with populist ideas were added to this neutral text, which transformed the neutral text into a “populist text”. These populist ideas consisted of
anti-establishment rhetoric, simplistic language, and the exclusion of the out-group and promotion of the in-group, which were the citizens who have lived in the city for a long time. A stock-picture was added to the text in order to help the respondents to feel more engaged with Mark Timmers. The “condition video”, which consisted of two campaign videos, was created by using neutral campaign videoclips of Dutch cities. These videoclips were made by real Dutch political parties. Using familiar videoclips would hopefully help to make the campaign videos look ‘normal’ for Dutch citizens. Also, the videos were made more ‘familiar’ by using a talking voice in the videos, which explained how the city was doing. This is also what real Dutch political parties tend to do. The same voice was used for both videos. In one video; the “neutral video’, only neutral and positive elements of the city were mentioned. In the second video, the “negative video”, mainly negative elements were mentioned about the city and about the city-council.
It was decided that all respondents would first see the neutral text or the populist text, in order to get to know the fictive politician Mark Timmers. After this, the respondents would see the neutral campaign video or negative campaign video. Combining the text with a video was a good way to make respondents familiar with the politician without getting bored, and was also similar to real life, because in real life it is common to read a text in the newspaper about a politician and then later see a television commercial that is sponsored by the same politician.
Dependent variable: Voting behavior towards politicians
The voting behavior towards politicians was divided into two directions. Firstly the voting behavior towards the main subject of the text and supporter of the video, namely Mark Timmers, was measured.
This was done by asking respondents two questions, which were: “How likely is the chance that you would vote for Mark Timmers in the next city-elections?” and “How likely is the chance that you will vote for the party of Mark Timmers in the next city-elections”? Both were measured on a 0-100 scale, in which “0 %” meant the chance of the respondents voting on Mark or his party would be none, and “100 %” meant this chance would be the highest as possible.
These two questions together created the scale “Chance to vote for Mark Timmers”, which was reliable because the Cronbach’s alpha was higher than 0.70. The Cronbach’s alpha was 0.92, M = 38.46, SD = 24,16.
To create the scale “chance of voting towards the target of the text and video” the following two questions were asked: “What is the chance of you voting on the politicians in the city council?’ and ‘What is the chance of you voting towards the parties of these politicians?”. The same scale
from “0” to “100” as mentioned above, was used. This scale had a Cronbach’s alpha of 0.87, M = 46.41, SD = 20.67, which meant the scale was reliable.
Dependent variable: Attitude towards politicians
The attitude towards politicians was divided into two groups:1. The attitude towards Mark Timmers, and 2. The attitude towards the current politicians in the city council.
The attitude towards Mark Timmers was measured by using seven questions with answer categories on a seven-point Likert scales, ranging from 1 (fully disagree) to 7 (fully agree). These seven questions were: 1. “Mark Timmers is competent”, 2. “Mark Timmers shows great
leadership strength”, 3. “I can trust Mark Timmers”, 4. “Mark Timmers is friendly”, 5. “Mark
Timmers cares about someone like me”, 6. “Mark Timmers is part of the establishment”, 7.
“Mark Timmers represents the will of the people”.
After a factor analysis and reliability test were conducted, it showed that the reliability of the scale “attitude towards Mark Timmers” could be heightened from 0.81 to 0.84 by leaving question 6 out of the scale. This was therefore done.
A factor analysis was conducted for questions 1, 2, 3, 4, 5 and 7. These six questions computed into one item with an Eigenwaarde that was above 1.00, namely 3.13. This item explained for 55.22 the variance in “Attitude towards Mark Timmers. The Cronbach’s alpha was 0.84, M = 4,51, SD = 1.00, the scale was therefore reliable.
and the same seven questions as were used for analyzing the “attitude towards Mark Timmers”. Only this time, the seven questions were not about Mark Timmers, but about the current
politicians in the city council. Again, a factor-analysis and reliability test showed that the reliability of the scale could be heightened by leaving question six out of the analysis. This was done. These six questions computed into one item with an Eigenwaarde that was higher than 1.00. The Eigenvalue was 3.82, and this item explained for 63,62% the variance in the ‘Attitude towards the current politicians”. The Cronbach’s alpha was 0.88, M = 4.07, SD = 1.04, which showed the scale was reliable.
Control variables: political knowledge, education level, age and gender
By asking four questions regarding the Dutch politics, the amount of political knowledge was measured (M = 2.70, SD = 1.02). Each question had four possible answers of which only one was correct. The questions were:
Question 1: ‘How many politicians are part of the “Tweede Kamer” ?’
Question 2: ‘Which branches make up the legislative branch (Wetgevende macht) in the
Netherlands?’
Question 3: ‘Which parties are in the current coalition?’ Question 4: ‘How often are the elections for the city council?’
When a participant answered a question correct, the respondent received one point. When the question was answered wrongly then the respondent received no points at all.
= 118) were placed in the “high political knowledge-group”. All respondents who had 0, 1 or 2 questions correct (N = 66) were placed in the “low political knowledge-group”.
The level of education of all respondents (M = 5.89, SD = 1.62) was measured by 9 categories, which ranged from 1. “Basisonderwijs” to 9. Professional school degree and/or doctorate degree (Prof., PHD, EdD). Of all respondents, most completed a bachelor’s degree (N = 59), second to most have completed “VWO” (N = 42), and third to most completed the HBO (N = 33).
The age of all respondents was between 18 and 63, (M = 24.18, SD = 6.86) and was measured by one question: “How old are you?”
The gender (M = 1.51, SD = 0.50), was measured with one question: “What is your gender?”, which had the answer categories: 1. “Male”, 2. “Female”, 3. “Other, non-binair”, 4. “I prefer not to tell”. All respondents filled in number 1 or 2.
Randomization
The experiment consisted of four conditions. Respondents were placed randomly in these conditions by the database Qualtrics. The graph below shows these different conditions:
To investigate if these four conditions had the same characteristics, a one-way ANOVA was conducted for the variables: “age”, “level of education” and “political knowledge”. Also, Chi2-tests were conducted for the variables: “gender” and “amount of citizens”.
Regarding the “age” (F( 3,183) = .630, p = .597), “level of education” (F(3,183) = .570, p = .635), “political knowledge” (F(3,183) = 2.031, p = .111), “gender” (X2(3) = 1.871, p = .600) and “amount of citizens” (X2(3) = 1.871, p = .600) of the respondents we see that the groups show the same characteristics because all p - values are not significant.
Therefore, the randomization of all respondents into the four conditions was done successfully.
Manipulation control
Five questions were added to the survey to find out if the manipulations had worked. The
respondents could answer the five question with answer categories on a seven-point Likert scales, ranging from 1 (fully disagree) to 7 (fully agree).
For all questions a one-way ANOVA was conducted. For questions 1-4 the independent variable was: “Condition Video”, for question 5, the independent variable was: “Condition text”
Regarding the tone of the video, two questions were asked: Question 1. “The campaign video had a neutral tone” (F(1,183) = 160.620, p < .001, for the neutral video: M = 5.12, SD = 1.34, for the negative video: M = 2.57, SD = 1.38). Question 2. “The campaign video had an overall positive tone” (F(1,183) = 45.449, p < .001, for the neutral video: M = 3.71, SD = 1.60, for the negative video: M = 2.26, SD = 1.30).
Regarding the blame attribute, two questions were asked: Question 3. “The blame attribute in the campaign video was directed towards the current politicians in the city council” (F(1,183) =
102.071, p < .001, for the neutral video: M = 4.16, SD = 1.53, for the negative video: M = 5.99,
SD = .91). Question 4. “There was no blame attribute in the campaign video” (F(1,183) =
115.738, p < .001, for the neutral video: M = 4.16, SD = 1.70, for the negative video: M = 1.79,
SD = 1.29).
Regarding the amount of populism of the text, one question was asked: Question 5. “How populistic did you find Mark Timmers in the text?” (F(1,182) = 1.943, p = .165, for the neutral video: M = 4.88, SD = 1.43, for the negative video: M = 5.17, SD = 1.33).
These results make clear that the video manipulation, regarding the negative tone and the direction of the blame attribute, were successful because all p – values were significant. The manipulation regarding the amount of populism of the text was unsuccessful because p > 0.05. This can be explained because more respondents thought the “neutral text” was populistic than expected. Another explanation could be that the term: “populism” is difficult for respondents who are not familiar with the meaning of it.
This result will make the possible moderation effects of “influence of populism” weaker.
Analyses
The analyses that were conducted to find out if hypotheses 1a, 1b, 2a and 2b were right were one-way ANOVA analyses. To find out if hypotheses 3a, 3b, 4a and 4b were right, the conduction of two-way ANOVA analyses were done.
Hypotheses 1 and 2
To find the answers to hypotheses 1a,1b, 2a and 2b, one-way ANOVA analyses were conducted. The results of these one-way ANOVA tests and Levene’s F test can be found below in table 3. In this table, DV1 is the voting behavior towards the target (H1a), DV2 is the voting behavior towards the sponsor (H1b), DV3 is the positive attitude towards the sponsor (H2a) and DV4 is the positive attitude towards the sponsor (H2b).
Table 3. Results found for hypotheses 1a, 1b, 2a and 2b.
DV1 DV2 DV3 DV4 Neutral N 82 82 82 82 M 47.98 40.93 4.26 4.59 SD 21.90 23.68 1.02 0.90 Negative N 102 102 102 102 M 45.14 36.48 3.92 4.46 SD 19.65 24.47 1.02 1.07 Difference M 2.84 4.45 .34 .13 F(df1,df2) Levene's test .857 (1,182) .977 (1,182) .052 (1,182) 3.837 (1,182) p .554 .324 .819 .052 F(df1,df2) ANOVA .857 (1,183) 1.553 (1,183) 5.121 (1,183) 1.553 (1,183) p .356 .214 .025 .214
Table 3 shows that the mean of the neutral campaign video group for DV1 (M = 47.98) is higher than the mean of the negative campaign video group for DV1 (M = 45.14). The difference (Mdiff = 2.84) is however not significant (F(1,183) = .857, p = .356), which results in Hypothesis 1a
being rejected. Therefore, the conclusion is that seeing a negative campaign instead of a neutral campaign does not statistically change the voting behavior of Dutch citizens towards the target of the campaign video.
This is not the only non-significant result, because the ANOVA analyses results for DV2 and DV4 are also not significant, and therefore hypotheses 1b and 2b are rejected. This means that when a group of Dutch respondents see a negative campaign instead of a neutral campaign their voting behavior towards the sponsor of the campaign video is not statistically different
(hypothesis 1b), and their positive attitude towards the sponsor of the campaign video is not statistically different (hypothesis 2b).
These results are not significant, which means that we cannot distinguish the differences in the sample from chance. However, these results do indicate that there are two interesting
observations to be made. Observation 1 is that when Dutch respondents see a negative campaign video instead of a neutral campaign video, they have a lower voting behavior towards the target of the video (DV1), and a lower positive attitude towards the sponsor of the video (DV4), which seems to suggest that there is some truth in hypotheses 1a and 2b. Observation 2 is that the results also suggest that Dutch respondents who see a negative campaign video instead of a negative campaign video will have a lower voting behavior (DV2) towards the sponsor of the video, which indicates that there is some truth in the opposite conclusion of what hypotheses 1b suggested.
Table 3 shows that the ANOVA analysis result for DV3 (H2a) is significant, which results in hypothesis 2a being confirmed. This means that Dutch respondents who see a negative campaign
video instead of a neutral campaign video will have a statistically less positive attitude towards the target of the campaign video.
Hypothesis 3
Hypotheses 3a and 3b were analyzed by conducting two-way ANOVA analyses. The results can be found below in table 4. In this table, DV1 is the voting behavior towards the target (H3a) and DV2 is the voting behavior towards the sponsor (H3b) of the campaign video.
Table 4 shows that the ANOVA analyses results are not significant, which results in hypotheses Table 4: Results found for hypotheses 3a and 3b
DV1 DV2
Neutral Populist Neutral Populist
Neutral N 35 47 35 47 M 48.69 47.46 37.59 43.43 SD 23.69 20.72 24.07 23.33 Negative N 43 59 43 59 M 45.44 44.92 37.24 35.92 SD 18.79 20.41 26.32 23.24 Difference M 3.25 2.54 .35 7.51 F(df1,df2) Levene's test .558 (3,180) 1.123 (3,180) p .643 .341 F(df1,df2) ANOVA .013 (1,184) .977 (1,184) p .910 .324
3a and 3b being rejected. Therefore, reading a populist text does not statistically moderate the possible main effects of seeing a campaign video on the voting behavior towards the target and sponsor of a campaign video.
Although not significant, the results indicate regarding hypothesis 3a, that reading a populist text does have some moderating power. As is shown below in graph 1, when a populist text is read, the voting behavior for the target will be lower. Also, when a populist text and negative video are present, the voting behavior for the target will be the lowest. This indicates that there is some truth in hypothesis 3a.
Graph 1: Hypothesis 3a results: Moderating effect of text condition on the voting behavior towards the
target of the campaign video.
Although not significant, the results for hypothesis 3b, as shown below in graph 2, indicate that seeing a populist text and negative campaign video does not score highest, as hypothesis 3b suggested. It actually scores lowest on voting behavior towards the sponsor. Surprisingly, the populist text and neutral campaign video scores highest. Also, when someone reads a neutral text
it does not really matter what campaign video is seen; the voting behavior is relatively similar. These results indicate that there is some truth in the opposite of what hypothesis 3b suggested.
Graph 2: Hypothesis 3b results: Moderating effect of text condition on the voting behavior towards the
sponsor of the campaign video.
Also, an interesting finding is that these conclusions hint towards the direction that Dutch citizens will remain an overall higher voting behavior towards the target of the campaign video instead of the sponsor of the campaign video, no matter the text condition or video condition.
Hypothesis 4
Hypotheses 4a and 4b were also analyzed by conducting two-way ANOVA analyses. The results can be found in table 5 below. In this table, DV1 is the positive attitude towards the target (H4a) and DV2 is the positive attitude towards the sponsor (H4b) of the campaign video.
Table 5 shows that the ANOVA analyses results are not significant, which results in hypotheses 4a and 4b being rejected. Therefore, reading a populist text does not statistically moderate the possible main effects of seeing a campaign video on the positive attitude towards the target and sponsor of a campaign video.
Although not significant, regarding hypothesis 4a, the results suggest that reading a populist text instead of a neutral text does have some moderating power. As is shown below in graph 3, the Dutch respondents have the lowest positive attitude for the target of the video when they read a Table 5: Results found for hypotheses 4a and 4b
DV1 DV2
Neutral Populist Neutral Populist
Neutral N 35 47 35 47 M 4.23 4.28 4.62 4.56 SD 1.09 .99 .89 .92 Negative N 43 59 43 59 M 3.99 3.87 4.53, 4.41 SD 1.10 .97 1.05 1.08 Difference M .24 .41 .09 .15 F(df1,df2) Levene's test .267 (3,180) 1.243 (3,180) p .849 .295 F(df1,df2) ANOVA .308 (1,184) .031 (1,184) p .579 .860
populist text and see a negative campaign video, which hints that there could be truth in hypothesis 4a.
Graph 3: Hypothesis 4a results: Moderating effect of text condition on the positive attitude towards the
target of the campaign video.
Although not significant, regarding hypothesis 4b, the results suggest that reading a populist text instead of a neutral text does have some moderating power. As is shown below in graph 4, the Dutch respondents have the lowest positive attitude for the sponsor of the video when they read a populist text and see a negative campaign video. This hints that there is some truth in hypothesis 4b.
Also, these results suggest that for both the populist and neutral text, seeing a negative video will decrease the positive attitude towards the sponsor. This decrease is stronger for reading the populist text than for reading the neutral text.
Graph 4: Moderating effect of text condition on the positive attitude towards the sponsor of the campaign
video.
Also, an interesting finding is that these conclusions hint towards the direction that Dutch citizens will remain an overall higher positive attitude towards the sponsor of the campaign video instead of the target of the campaign, no matter the text condition or video condition.
Conclusion and Discussion
This research intended to reveal if seeing a negative campaign could lead to a different voting behavior and different positive attitude towards the target and sponsor of a campaign video. Furthermore, it was researched if “populism”, could moderate the possible results of the main effects. In this research a total of 211 respondents participated, of which 184 results could be used.
The results for hypothesis 1a,1b,2b,3a,3b, 4a and 4b were not significant. Therefore, we cannot distinguish the differences in the samples from chance. However, the result for hypothesis 2a was significant, which confirms hypothesis 2a. This means that Dutch citizens who see a negative
campaign video instead of a neutral campaign video, will get a statistically lower positive attitude towards the target of the campaign video. This is an interesting finding, because it contradicts the findings of Min (2004) and Frimer and Skitka (2018). Frimer and Skitka (2018, page 863) found out that incivility harms the sponsors reputation and also “does little little to harm to the target of
incivility”. Therefore, the results in this study show opposition to their “Montagu Principle”.
Although not significant, the results of this research suggest that there could be some truth in hypotheses 1a, 3a and 4a. Dutch citizens that see a negative campaign video instead of a neutral campaign video, are less likely to vote on the target of the campaign video (H1a). Also, when populist elements are combined with a negative or neutral video, Dutch citizens are less likely to vote for the target (H3a) and will have a lower positive attitude towards the target (H4a). The voting behavior and positive attitude towards the target are the lowest when the combination of populist text and negative video is present.
The results of the analysis of hypotheses 2b and 4b suggest that there is also some truth in what the hypotheses suggested. When Dutch citizens see negative campaign video’s instead of neutral campaign video’s, they a lower positive attitude (H2b) towards the sponsor of the video. Seeing a populist text and negative video results in the lowest positive attitude towards the sponsor (H4b).
The results of the analysis of hypotheses 1b and 3b suggest that there is also some truth in the opposite conclusion of what the hypotheses suggested. Dutch citizens who see a negative video have a lower voting behavior (H1b) towards the sponsor, and Dutch citizens who see a text with more populist elements combined with seeing a negative campaign video score lowest on voting
behavior towards the sponsor (H3b). Surprisingly, the combination of populist text and neutral campaign video scores highest on voting behavior towards the sponsor.
Because the results towards the voting behavior of the target and sponsor are similar to the findings regarding the positive attitude, it is suggested that the results in this research are not just results of ‘chance’. A bigger sample would have more power and would most likely lead to statistical results for the hypotheses.
Politicians can take the following learning points from this research:
Firstly, when the politicians’ goal is decreasing the voting behavior or positive attitude of Dutch citizens towards the political target, then the politician should use negative campaign videos combined with populist texts. Secondly, as was expected, this research shows that there is a backlash for sponsors of negative campaigns. This backlash is highest when populist elements are used in texts of the sponsor combined with a negative campaign video. Thirdly, this research finds that there is distinction between ‘positive attitudes’ and ‘voting behavior’ towards
politicians, which is similar to the results in the articles of Frimer and Skitka (2018) and Carraro and Castelli (2010) in which sponsors of negative campaigns had a perceived less warmth (attitude) but received more positivity in terms of capability. In our research however, the voting behavior is always higher towards the targets of the campaign video, and the positive attitude is always higher towards the sponsor of the campaign videos, no matter the condition. One of the reasons for this can be that Dutch citizens are just more likely to vote for ‘incumbent politicians then for ‘candidate politicians’. To find out more regarding the influence of being new in the political arena, other scholars can decide to look into the possible moderating effects of being a candidate politician instead of being an incumbent politician.
These results indicate that politicians in the Netherlands should only publicly sponsor negative campaigns when they want to gain social approval over their opponents. Regarding the voting behavior, the politician should not be linked to sponsoring campaign videos against the targets, otherwise the sponsor will receive less votes than the target of the videos. Therefore, the best thing to do is to have the populist text and negative video published by another source. If the goal of the politician is to receive the greatest number of votes possible, he or she should publish a text with populist elements combined with publishing a neutral campaign video. Although the overall voting behavior will be higher for the target than the sponsor, this research gives reason to think that publishing multiple populist texts and neutral campaign videos will at some point result in a higher voting behavior towards the sponsor than towards the target. However, to follow up on this thought, it would be best if scholars conduct a longitudinal research to research this.
Fourthly, the results of this experiment indicate that voting behavior can be changed by negative campaigning, but that the concept of “voting towards politicians” in the human brain is not directly an outcome of their attitude towards the politicians. Therefore, the politician that would like to use negative campaigning should always focus on what the goal is; is it to change the positive attitude of citizens or is it to change the voting behavior?
Limitations and future research
This research also had some limitations. Firstly, the sample size was not so big, and with only 184 respondents that are divided into four conditions with different manipulations, one group consisted of ‘only’ 35 respondents, which is not a high number. Therefore, the external validation is not as high as was hoped for. Also, surprisingly, most respondents who read the neutral text said that the text was a populist text, which was not intended by the researcher. Because of this, probably the results of possible moderating effect of populism have become lower.
In this Master thesis, the focus was on negative campaigns that consisted of policy attacks and character attacks. Fellow students could do more intensive research into the possible differences in outcomes when using character-attack based campaigns or policy-attack based campaigns. They might find results that give more insights into the experiment of this Master thesis and could help politicians as well with campaigning. For example, using only policy-attacks could have decreased the backlash which was found in the results of this research. Also, more elements that might decrease a backlash effect could be researched by scholars. When more elements are found, next to populist elements in texts, politicians could use these elements when they publish negative campaigns in the future. Furthermore, when researches will focus again on using populist elements in campaigns, they should make sure that respondents do not find texts populistic when there are supposed to be neutral. Lastly, this research sheds light on negative campaigning in a multi-party electoral system and indicates that the results will be different than using negative campaigning in a two-party system. Scholars can do more intensive research into this by for example doing one research in multiple countries with multi-party electoral systems. This will hopefully give more information regarding the usage of negative campaigning in differences between electoral systems.
Literature list
Boeije, H., Hart, H., & Hox, J. (2005). Onderzoeksmethoden. Boom onderwijs.
Brooks, D., & Geer, J. (2007). Beyond negativity: The effects of incivility on the electorate. American
Journal of Political Science. 51, 1–16. doi: 10.1111/j.1540-5907.2007.00233.x.
Carraro, L., & Castelli, L. (2010). The implicit and explicit effects of negative political campaigns: Is the source really blamed? Political Psychology, 31(4), 617–645. doi: 10.1111/j.14679221.2010. 00771.x
special issue. Information, Communication & Society, 20(9), 12791292, doi: 10.1080/1369118X. 2017.1328525
Ernst, N., Blassnig, S., Engesser, S., Büchel, F., & Esser, F. (2019). Populists prefer social media over talk shows: An analysis of populist messages and stylistic elements across six countries. Social
Media+ Society, 5(1). doi.org/10.1177/2056305118823358
Fridkin, K.L., & Kenney, P. (2011). Variability in citizens’ reactions to different types of negative campaigns. American Journal of Political Science, 55, 307-325. doi:10.1111/j.1540-5907.2010. 00494.x
Frimer, J. A., & Skitka, L. J. (2018). The Montagu Principle: Incivility decreases politicians’ public approval, even with their political base. Journal of Personality and Social Psychology, 115(5), 845. doi:10.1037/pspi0000140
Hopmann, D. N., Vliegenthart, R., & Maier, J. (2018). The effects of tone, focus, and incivility in election debates. Journal of Elections, Public Opinion and Parties, 28(3), 283–306. doi:10.1080/17457289. 2017.1394310
Jagers, J., & Walgrave, S. (2007). Populism as political communication style: An empirical study of political parties' discourse in Belgium. European Journal of Political Research, 46(3), 319– 345. doi:10.1111/j.1475-6765. 2006.00690.x
Lau, R. R., & Pomper, G. M. (2001) Negative campaigning by US Senate candidates. Party Politics 7(1), 69–87. doi:10.1177/135406880.1007001004
Lau, R.R., Sigelman, L., & Rovner, I.B. (2007). The effects of negative political campaigns: A meta‐ analytic reassessment. Journal of Politics, 69, 1176–1209. doi:10.1111/j.1468-2508.2007.00618.x Martin, P.S. (2004). Inside the black box of negative campaign effects: Three reasons why negative
campaigns mobilize. Political Psychology, 25(4), 545–562. doi:10.1111/j.1467-9221.2004.00386.x
Min, Y. (2004). News coverage of negative political campaigns: an experiment of negative campaign effects on turnout and candidate preference. Harvard International Journal of Press/Politics, 9(4), 95–111. doi: 10.1177/1081180X04271861
Moffitt, B., & Tormey, S. (2014). Rethinking populism: Politics, mediatization and political style.
Political Studies, 62(2), 381–397. doi: 10.1111/1467-9248.12032
Mudde, C. (2004). The populist zeitgeist. Government and Opposition, 39(4), 542–564. doi: 10.1111/j.1477- 7053.2004.00135.x
Nai, A. (2018). Going negative, worldwide: Towards a general understanding of determinants and targets of negative campaigning. Government and Opposition, 1–26. doi:10.1017/gov.2018.32
Trump. Personality and Individual Differences, 121, 80–83. doi:10.1016/j-paid.2017.09.20 Weinschenk, A., & Panagopoulos, C. (2014). Personality, negativity, and political participation. Journal
