It’s about the relevance, stupid!
A study about the effects of psychological targeting on open to experience voters Aldona Adamowska
11805358
Master’s Thesis
Graduate School of Communication Master’s programme Communication Science
Sanne Kruikemeier 12 July 2019
Abstract
This is one of the first studies that addresses the effects of psychological targeting in the context of political communication. The purpose of the study was twofold. First, it aimed to examine the effects of a message tailored to recipients’ personality on affective and behavioral responses towards the message and the source. Second, it investigated what role sponsorship disclosure and persuasion knowledge plays in this relation. The results provide evidence for the effectiveness of psychological targeting. Messages tailored to recipient’s personality profile positively influence their likeliness to engage in electronic word of mouth, but negatively feelings towards the source of the message, thus the political party. These contradictory findings suggest that the effects of psychological targeting in the context of politics are more nuanced than in the other fields. With regard to the role of sponsorship disclosure and persuasion knowledge, the results suggest that tailored messages with sponsorship disclosure do not trigger persuasion knowledge and lead to negative evaluations of the message and the source. These findings adhere to public debate about transparency around political advertising. They might provide relevant implications for political parties, politicians, political campaigners, policymakers, as well as for voters.
Keywords: Psychological targeting, personality, tailoring, sponsorship disclosure, political communication
It’s about the relevance, stupid!
The potential of psychological targeting in politics has become a very popular topic in the public discussion, even before this phenomenon had a chance to mature (Matz, 2018). The controversy around Cambridge Analytica – a company behind the 2016 Trump presidential and pro-Brexit Vote Leave campaign, casted a spotlight on psychological persuasion in its very early stage (Bodo, Helberger & de Vreese, 2017). Despite this scandal, it is very unlikely that
psychological targeting will disappear from politics anytime soon. This new technique can be defined as using personal data (psychographics) to create personality profiles of individuals and reach them with tailored messages (Matz, 2018). Although psychological targeting is often described in the media as a “new wonder weapon” or “dark art” that helped succeed both Trump and the pro-Brexit campaign, the evidence supporting this headline-grabbing claims remains unclear (Graves & Matz, 2018). Even though psychological persuasion might not be the main driver behind Trump or Brexit, it does not necessary mean that is does not work.
In fact, studies on psychological persuasion show that tailoring messages to peoples’ personality, as was in the case with Trump and the pro-Brexit campaign, makes them more effective. For example, a mobile phone advertisement that was congruent with personality traits of participants lead to more positive attitudes towards the advertisement and the brand (Hirsch, Kang & Bodenhausen, 2012). Similarly, matching the content of advertisements to peoples’ personality profiles influences their behaviour indicated as clicks on the add, app instalment or product purchase (Matz, Kosinski, Nave & Stillwell, 2017). While these findings provide evidence for the influence of personality targeting on attitudes and behaviours, their generalizability might be limited. What works for retail, not necessarily needs to works for
political campaigns, as voters’ behaviour is more complex and might also be influenced by by political ideologies and party identifications (Graves & Matz, 2018).
Although the effects of psychological targeting in the political domain are still uncertain, there is one thing for sure: psychological characteristics can be very accurately predicted from people’s digital footprints, like their social media profiles (Kosinski et al., 2013). Such digital data of millions of people, although harvested illegally, allowed Cambridge Analytica to create personality tailored “dark posts”: sponsored content that is only displayed for the specifically target users (Borgesius et al., 2018).
To prevent such a manipulation in the 2019 EU parliamentary elections, Facebook recently updated its policy on political ads. Among few changes, they introduced a new label, that shows who paid for an ad (Protecting Elections in the EU, 2019). According to Facebook, this will increase transparency and prevent abuse or foreign interference in the election – thus bring improvement on two main issues Facebook is often criticized for. However, it remains a question whether those changes have an effect on psychological microtargeting and political advertisement. A growing body of research suggests that sponsorship disclosures embedded in political context activate persuasion knowledge and in consequence leads to negative responses (Main, Dahl & Darke, 2007; Kruikemeier et al., 2016). Specifically, a disclosure will make people aware of the advertisement, which enhances their knowledge, which makes people more likely to cope with the message. However, disclosures are rarely recalled. Therefore, it is interesting to investigate whether the new labels lead to more awareness.
Taking into consideration all the issues presented above, the main aim of this study is to examine whether messages congruent with individuals’ personality profile are more effective. I also investigate the role of transparency, by examining whether a disclosure makes people more
aware that they are being persuaded, which in turn could affect the effectiveness of the targeted message. Thus, the research question of this study is:
RQ: To what extent does a targeted political message tailored to a recipients’ personality have an effect on the recipients’ affective and behavioural responses towards the message and its source, what role does a sponsorship disclosure and persuasion knowledge play in this relationship?
This research question is particularly important, as psychological targeting can have both positive and negative effects on the marketplace of ideas and a democracy in general. On the one hand, psychological persuasion in politics might be abused to manipulate voters. On the other hand, it might be used to stimulate voters’ political interest and participation.
Political micro-targeting
The advent of social media changes the way political campaigns look these days. Facebook, Twitter or YouTube not only introduced new ways of communication, but also facilitate gathering data of internet users (Dobber, Trilling & Helberger, 2017). Both the 2008 and 2012 campaigns of Barack Obama’s were successful examples of the use of data and brought attention to micro-targeting (Bodo et al., 2017). Although, targeting is nothing new in politics, the advances in technology significantly increase the amount of information that can be received from individuals. Therefore, the core of political micro-targeting (PMT) is the collection, processing and analysing of digital data (Borgesius et al., 2018). The data consist of various information such as demographics, lifestyles, habits or peoples’ online behaviour. This data is used to target voters with individually tailored ads (Borgesius et al., 2018). As a majority of political tactics, PMT originates from the commercial sector. However, the persuasion is not it’s only goal. PMT can be employed to “(dis)encourage political participation, including election turnout, (dis)encourage donations and contributions to candidates and campaigns, create energy and interest in a campaign, election and candidate, but also disinterest and apathy” (Bodo et al.,
2017). Although political campaigners are positive about the results of PMT, there is a rather limited amount of research examining its effect. Among few, the study of Liberini, Redoano, Russo, Cuevas & Cuevas (2018) demonstrates that PMT ads “increased Republican turnout by up to 10% among voter groups” in 2016 US election. Other research more broadly shows that
messages tailored in terms of moral foundations enhance sustainable behaviours of liberals and conservatives (Kidwell, Farmer & Hardesty, 2013).
Psychological targeting
As many consider behavioural micro-targeting as the latest development (Dobber et al., 2017) in political campaigning, both the 2016 US presidential and Brexit vote campaign drew attention to something even more novel: psychological microtargeting. The underlying concept of psychological microtargeting is to tailor messages to individuals’ unique psychological
characteristics and motivation, such as personality traits (Matz et al. 2017). Personality research in political communication.
Personality traits can be defined as “individual differences in characteristics patterns of thinking, feeling and behaving” (Mondak, 2010). The most popular taxonomy for personality traits is the five-factor model (FFM) that categorizes them in five dimensions (Goldberg, 1993). The five broad personality traits are represented by the acronym OCEAN that stands for:
openness to experience (openness in short), conscientiousness, extraversion, agreeableness or neuroticism. Studies have found that these five traits are highly stable over time and space (Cawvey, Hayes, Canache & Mondak 2017), which leads to the general assumption that
personality can influence attitudes and behaviour. Therefore, many political scientists turn to the FFM model as an important source of variation in political attitude or behaviour (Bakker,
Klemmensen, Nørgaard & Schumacher, 2016). Although personality is a highly researched topic in the field of politics, the evidence on the effects of psychological targeting in politics is scarce.
Even though, there are some findings, mostly related to the “congruency principle”, that could explain possible effectiveness of personality tailored messages. For instance, the study of Kam & Simas (2012) shows that voters are more likely to select candidates whose traits match their own traits. With regard to the message, the researches revealed that messages with violent rhetoric lead to greater partisan polarization among people with more aggressive personalities than among less aggressive individuals (Kalmoe, Gubler & Wood, 2018).
Psychological targeting research in other fields.
Mainly two disciplines, commercial and health communication, examine the effects of psychological targeting. The Matz et al. (2017), conducted a study in the domain of consumer products, that until now, remains the most prominent research about psychological
microtargeting. The results of their three field experiments suggest that advertising matching individuals’ psychological characteristics significantly influence their behaviour. This was measured with the use of marketing metrics: click-through rates and conversion rate. Facebook users who saw ads congruent with their level of openness or extraversion were more likely to click on the ad, install the promoted app or purchase the promoted product than users who were shown incongruent ads. Other papers revealed that advertisement adapted to target each of the five personality traits were evaluated as more effective among participants who score high on the given trait (Hirsch et al., 2012). Matching computers advice style with peoples’ personality makes it also more effective, as they more likely to change their initial opinions in response to congruent advices (Moon, 2002). The evidence on the effectiveness of psychological tailoring, also comes from other papers that did not concentrate on personality traits per se, but for instance, on a gain and loss frame. Their findings suggest that when a message is matching a recipient’s motivational orientation, it is evaluated more positive. (Cesario, Higgins, Scholer, 2008; Higgins, 2000).
As prior studies have demonstrated some evidence for the effectiveness of psychological microtargeting, I want to examine whether a political message tailored to respondents’ level of openness will have a positive influence on its perceived effectiveness, respondents’ likeliness to engage in electronic word of mouth or their attitude (as it has in other disciplines). I also want to measure the effectiveness of psychological persuasion on key political outcomes: feelings towards the party and voting behaviour. Therefore, I expect that:
H1: A political message tailored to respondents’ personality will positively influence affective and behavioural responses compared to message not tailored to respondents’ personality.
As can be seen, personality is operationalized in my study as level of openness to experience – personality trait that characterize curious, open-minded individuals with lively imagination (Bekker et al., 2016). This is in line with previous studies on psychological
persuasion that also tailored messages in terms of separate traits (Matz et al., 2017). I specifically choose openness, as its frequently investigated in the context of psychological microtargeting. Moreover, open to experience citizens are more likely to switch parties, therefore might constitutes the perfect audience for persuasion (Bekker et al., 2016).
Persuasion Knowledge of voters
The evidence from previous studies suggests that voters have a rather negative attitude towards tailored messages if they know they target them on the basis of their information. While commercial messages in some cases are perceived as useful, this does not apply to messages that have political content (Boerman, Willemsen & van der Aa, 2017). A study of Turow, Capini & Draper (2012) found that 86% of Americans do not want political advertising tailored to their interests, and 64% of them say they will be less willing to vote for a candidate if they learn that she or he uses behavioural microtargeting. While examining this type of effects, the theory of persuasion knowledge (PK) might be relevant. Peoples knowledge and beliefs about issues
related to advertising, such as its motives, goal and tactics, can be defined as PK (Boerman, van Reijmersdal & Neijenset, 2012). Although established in the adulthood, PK starts to develop gradually from the early childhood. First, it just appears as differentiating promotional messages from other types of content. Later, it can be characterized by understanding marketing aims, techniques or deception, and in the end, it leads to “the ability to use cognitive defences against advertising” (Boerman et al., 2017). In this paper, PK is operationalised as the respondents’ recognition of the message as advertising, that is the most basic form of persuasion knowledge. Based on previous research, I presume that voters affective and behavioural responses will be affected by their level of persuasion knowledge.
Sponsorship Disclosure in political messages
However, persuasion knowledge must be triggered by a cue. On the basis of recent policy changes that Facebook announced, this study will focus on the use of disclosures (i.e., the cue that might people aware and thus triggers persuasion knowledge). When sending out targeted messages on Facebook, politicians and parties use post that are labelled “sponsored content”. These are promotional messages that look almost like the regular posts from a person’s network of friends (Boerman et al., 2017). To make consumers aware of the persuasive intent of the sponsored post, they include the label “Sponsored” at the top (this is the cue). When such disclosures are noticed, they enhance the activation of PK (Boerman & Kruikemeier 2016; Campbell & Kimani, 2000). Still, the resemblance of sponsored ads to the content that is posted by befriended contacts is so great, that it is difficult for a majority of social media users to recognize them (Kruikemeier et al. 2016, Wojdynski & Evans, 2016). However, this might change, as in the run up to the European Parliamentary Election in May 2019, Facebook restricted the way ads related to politics works on their platform (“Protecting Elections in the EU”, 2019). This was an answer to intense scrutiny from the European Commission that aims to
prevent foreign interference in the national election processes. One of the new rules requires the advertiser to add a “Paid for by” disclaimer to the label. According to Facebook, this will increase transparency, as users will be able to see who is paying for the add. Previous research established a negative relation between recognition message as advertisement and attitudes towards the message and its sender (Main et al., 2007). Moreover, some studies show that when the sponsorship disclosure is accompanied by additional information, this leads to an increase in PK (van Noort, Smith, Voorveld, 2013). Furthermore, a study of Weber, Dunaway & Johnson (2018) demonstrates that ads that are sponsored by a political candidate or known interest group are less persuasive. All these effects can be explained by the Elaboration Likelihood Model (ELM) that proposes two routes of processing a message: central and peripheral (Petty, Cacioppo, 1986). According to the ELM the presence sponsorship disclosure evokes the central route of processing, thus activates PK and leads to more careful and critical processing. When the disclosure is absent, people are more likely to use the peripherical route, as they do not realize that the message has a commercial purpose. I believe it is important to examine whether messages with new sponsorship disclosure on Facebook evokes biased processing and in consequence leads to more negative responses, while messages without the new label don’t.
Hence, taking into consideration previous studies and the ELM theory, I propose a moderated mediation hypothesis (Fig 1):
H2: Sponsorship disclosure moderates the effect of tailored message (vs. non-tailored message) on affective and behavioural responses through the activation of persuasion knowledge: The mediated relationship of the tailored message on respondents’ responses will be more negative when the message includes sponsorship disclosure (vs. excluding sponsorship disclosure).
To a set of five variables that operationalize the concept of affective and behavioural responses I add the “source trustworthiness”. In prior studies trustworthiness was defined as “the
perceived credibility and honesty of communicator” (Kruikemeier et al., 2016). It’s assumed that when the message is perceived as commercial in nature, it’s perceived as less trustworthy (Main et al., 2007). Therefore, I expect that sponsorship disclosure will reduce the credibility of the political party.
Figure 1. Conceptual model of moderated mediation, in which a tailored message (vs. non-tailored message) positively affect respondents’ affective and behavioral responses (H1) and tailored message (vs. non-tailored message) generate different levels of persuasion knowledge: increase in persuasion knowledge is expected to be more pronounced for messages with
sponsorship disclosure (vs. no sponsorship disclosure), which consequently affects responses to the message and its source (H3).
Methods Sample
A total of 257 participants were recruited for the research, and the study was completed by 210 of them (completion rate 82%). Participants were selected using a convenience sample with the combination of snowball sampling. As an incentive, participants were offered a chance to win one out of five Netflix cards. In order to meet the inclusion criteria, participants needed to
persuasion knowledge
tailored vs. non tailored message
persuasion knowledge
affective and behavioural responses sponsorship disclosure vs. no sponsorship disclosure
speak polish and be aged 18 and older. One participant reported age as 16, therefore she was excluded from the sample. During the closer examination of the data I identified few outliers, but I decided not to remove it from the analysis as it could not be determined that those observations are erroneous. The final sample consisted of 81% (N = 209) of all participants. The mean age of participants was 33 years, (SDage = 11.63, range 18 - 85). There were more females (56%) than
males (42.1%) and 4 participants identified themselves as “non-binary”. My sample was uniformly based on the country of origin, 96.7% of participants were Polish. The majority of participants (52.2%) held a master degree, over a quarter (25.4%) held a bachelor degree and 22.5% completed a lower level of education. It is interesting to note, that the final sample resembles the average Facebook user (Facebook by the numbers, 2019).
Design
To test my hypotheses, a 2 (a type of the message: tailored message vs. not tailored message) x 2 (disclosure: sponsorship disclosure vs. no sponsorship disclosure) between-subjects design was employed. The study was in the form of an online-survey-embedded experiment. I chose experiment as my research design because it allowed me to rule out alternative
explanations of causal relationship. By employing random assignment, I cancelled out the variation of participants pre-existing attributes and ensured there are no other reasons for the potential effect. This maximized internal validity of my research. Participants were randomly assigned to four conditions, and exposed to either a regular Facebook post tailored to recipients high on openness (n = 54), the same post but labelled as sponsored (n = 53), the Facebook post tailored to recipients low on openness (n = 52) or the same post but labelled as sponsored (n = 50).
The stimulus material consisted of a fictious Facebook post that was manipulated in terms of openness to experience, one of the personality traits of the Five Factor Model and sponsorship disclosure, a label that allows identifying the source of sponsored material. Each participant viewed a screenshot of the post sent by a polish political party Spring (Wiosna). I intentionally chose a party over a single candidate, as the study was conducted in the parliamentary election year in Poland. The Spring party was selected, because it is the third political power in the
country. As polarization – the vast gap between two major political parties is a defining feature of Polish politics today, it was beneficial to choose a party that do not have any affiliation with the biggest ones. The post was based on authentical posts that were published by the Spring party on Facebook during the last political elections in Poland, the 2019 European elections. The post was tailored to recipients’ level of openness. Post targeting high openness highlighted creativity, innovation and diversity, while low openness familiarity, traditionalism and convention (McCrae & John, 1997). A few lines of text were inspired by previous studies on psychological persuasion (Hirsch et al., 2012; Matz et al., 2017) as well as research about personality (Pennebaker & King, 1999; Schoonvelde et al., 2019; Schwartz et al., 2013; Park et al., 2014). For instance, the high openness condition made use of complex language, as well as longer words referring to art forms such as music or movies (e.g. “You will also have free access to a collection of books, music or movies from all over the world”). While the low openness condition included shorter words related to family or home (e.g. “Kids and youngsters will also be able to get familiar with the values, traditions and customs present in our country for generations”; see Appendix A for details). The text used in the final posts was revealed out of fourteen messages tested in the pilot test. The accompanying picture was also tailored to recipients’ level of openness and was inspired by previous research. For instance, a recent study found that “open-minded people favour images with rather cold colours over images with warm colours and images that feature people and
faces” (Matz et al. 2019). In line with this finding, a photo in the condition targeting open individuals depicted objects in a blue and black wavelet, while the picture in the condition targeting closed individuals feature people and have orange and brown tones. Moving to
sponsorship disclosure, in two out of four conditions, the post includes a label “Sponsored. Paid for by Wiosna Roberta Biedronia” that appeared underneath the name. This is identical to disclosures that are nowadays posted on Facebook (Protecting Elections in the EU, 2019). In the other two conditions, the post did not contain sponsorship disclosure and resemble a regular Facebook post. Despite those differences, I tried to make the stimuli material as similar as possible. The profile picture, the main slogan, the design of the photos, the number of words in the message as well as likes, shares and comments are identical across conditions.
Procedure
Shortly after the pilot test, participants were recruited over a week timeframe in June 2019 by the use of social media, text messages and word of mouth. Most of them received short
messages with the link to the online survey that invited them to participate in the study. Participants could access the survey independently from their own environment and devices which increased the ecological validity of the experiment. Every participant starts the survey with an introduction page, that informed them about their anonymity, voluntary participation and possibility to withdraw their answers at a later point in time. It also presents the research as a general study about assessment political messages on Facebook. To process further, participants had to check the box indicating their informed consent. On the beginning of the survey, all of them had to answer on a few questions related to Facebook usage, political orientation as well as a personality profile. After completing the pre-questionnaire, participants were randomly
assigned to one out of four experimental conditions. They were instructed to look at the Facebook post as it appeared on their own newsfeed and to read it carefully, in order to answer the
following questions. Then, a Facebook post was shown. Participants could look at it as long as possible before continuing with the questionnaire, but at least for 20 seconds, to ensure its completion (Vraga & Bode, 2017). The post-test starts with questions about likeliness to engage in EWOM, followed by the attitude towards the post as well as the source. Next, the other
dependent variable was measured such as feelings towards Spring or voting intention. They were followed by questions on the perceived trustworthiness of the source, persuasion knowledge, recall and recognition of the sponsorship disclosure. This order of the questions ensured that attitudinal and behavioural answers were not primed by questions that insinuated the promotional nature of the Facebook post (Boerman et al., 2017). At the end, participants answer on a few general demographic questions and could give their e-mail address to take part in Netflix gift cards lottery. Afterwards, participants proceed to the final screen, where they were thanked and debriefed. They learned that the post they saw was fake, created exclusively for the needs of my study and any political party or pooling firm was involved in their creation. They were also instructed not to share the aim of the research with others, until the end of data collection. On average, participants took around 15 minutes to complete the survey.
Measures
Personality was assessed using two batteries. Openness to experience was measured with the Polish adaptations of Goldberg’s IPIP-BFM-50 questionnaire for measuring the five
personality traits in a lexical tradition (Strus et al. 2014). Its shorter version, the MINI-IPIP (Donnellan et al. 2006) assessed other psychological traits. Both inventories proved to be a valid and reliable measure of personality, regardless of different research conditions (online vs. offline) or different research arrangements. They are also more immune to type S and M errors than brief measures of personality traits (Bakker & Lelkes, 2018). The items have a form of short sentences, e.g. “have a rich vocabulary” (See Appendix B) and participants give their response on a
five-point Likert scale, where the anchors are 1 (very inaccurate) to 5 (very accurate). 10 items measuring openness were collapsed into a scale by averaging and had a Cronbach’s alpha of 0.87 (eigenvalue = 3.85, explained variance = 38.85%, M = 3.76, SD = 0.59). Openness was also recoded with the use of median split into a categorical variable with the mean M = 1.54 (SD = 0.50) and 96 participants in low openness group and 113 in high openness group. Scales for extraversion (M = 3.34, SD = 0.96), agreeableness (M = 3.93, SD = 0.66), conscientiousness (M = 3.36, SD = 0.87) and neuroticism (M = 3.06, SD = 0.84) were created by averaging 4 items assessing each personality trait.
The dependent variable, thus attitudinal and behavioural responses regarding the message and its source, Spring party was operationalized with five measures.
Perceived effectiveness (PE) of the Facebook post was measured with a six-item scale adopted from previous research to fit the political context (Davis, Uhrig, Bann, Rupert & Fraze, 2011). Participants were asked to indicate whether they agreed or disagreed with the following statements: “this post was worth remembering”, “this post grabbed my attention”, “this post does not speak to me” (reverse coded)”, “this post was convincing” , “I liked this post overall”, “ this post was interesting”, “I do not like this post (reverse coded)”. Those items were measured on seven-point Likert response scale (1 = strongly disagree, 7 = strongly agree) and averaged together to form a measure of PE (eigenvalue = 3.94, explained variance 65.66%; Cronbach's α = 0.89; M = 4.06, SD = 1.39).
The likeliness to engage in electronic Word of Mouth (eWOM) was assessed using five items: “I would write comment under the post”, “I would share the post with friends”, “I would visit the Facebook page of Spring”, “I would like the post of Spring on Facebook” and “ I would like the Facebook page of Spring” from Kruikemeier et al. (2016) study. These items were measured on a seven-point scale (1 = strongly disagree, 7 = strongly agree) and factor analysis
showed a one-factor solution for the scale (eigenvalue = 3.02, explained variance = 60.46%, Cronbach’s α = 0.84; M = 3.07, SD = 1.28).
Attitude towards source was derived from Boerman et al. (2017) study, and it consists of three seven-point semantic differential scales: bad/good, negative/positive and dislike/like that measured a respondent’s overall attitude towards the Spring party. I used the mean score of the items as a measure of source attitude (eigenvalue = 2.70, explained variance = 90.04%;
Cronbach's α = 0.95; M = 4.65, SD = 1.70).
Feelings towards the source were measured by asking participants to rate the Spring on a thermometer scale, whereas rating between 0 and 50 degrees indicate cold, unfavourable feelings and ratings between 50 and 100 degrees warm, favourable feelings towards the party (Alwin, 1992). The mean was calculated (M = 49.78, SD = 28.20).
The Voting intention was measured with the question “Assuming that a presidential election would happen tomorrow, how likely would you be to vote for Spring based on the post you saw?” based on previous studies (Aird, Ecker, Swire, Berinsky & Lewandowsky (2018). Responses were provided on an eleven-point scale, ranging from 0 (very unlikely) to 100 (very likely) and the mean was M = 35.19 (SD = 30.67).
The perceived trustworthiness of the source of the post was assessed with five seven-point semantic differential scales: undependable/dependable, dishonest/honest,
unreliable/reliable, insincere/sincere, and untrustworthy/trustworthy (Kruikemeier et al., 2016) The mean score of them was calculated (eigenvalue = 4.27, explained variance 85.29%; Cronbach's α = 0.96; M = 4.78, SD = 1.04).
Conceptual Persuasion Knowledge (PK) is a mediator that was measured by asking participants to indicate whether they agreed with 5 statements: ‘‘the post of Spring feels like an ad’’, ‘‘the post promotes Spring’’, ‘‘Spring paid to post this message’’, ‘‘the post of Spring is an
ad’’ and ‘‘the post is sponsored by Spring’’ based on Kruikemeier et al. (2016) study. The responses were recorded on a seven-point scale (1 = strongly disagree, 7 = strongly agree). Again, I calculated a mean score (eigenvalue = 2.96, explained variance = 59.14%, Cronbach’s α = 0.81, M = 4.95, SD = 1.15).
Two measures served as exposure check. The memory and recognition of sponsorship disclosure were measured in a similar manner as in a Boerman et al. (2012) study. I asked participants whether they recalled the label that informed them that the post was sponsored (0 = no, 1 = yes) and I also present them with four disclosures and asked them which of the four labels they had seen (0 = wrong disclosure, ‘none of the above’ or ‘I did not see disclosure’, 1 = correct disclosure). 82% of the respondents that were exposed to a label did not remember it and 97 % could not identify it correctly. Both measures are moderately correlated r (209) = 0.50, p < .001).
Several control variables were included to make sure that the effect of the post openness and the disclosure are not caused by other differences between the experimental groups. Political orientation was measured using a scale developed by Czarnek, Dragon, Szwed & Wojciszke, 2017) to fit the Polish context. The scale includes subscales that resemble social (9 items, e.g. “conceived life should be subject to absolute law protection”; a = 0.91, eigenvalue = 5.46, explained variance = 60.68%, M = 3.09, SD = 1.41) and economic (10 items, e.g. “economy should be centrally planned by the state”; a = .81, eigenvalue = 3.78, explained variance = 37.81%, M = 3.85, SD = 1.01) ideology on the left-right dimension. Participants indicate the extent they agree with the statements on the seven-point response scale, ranging from 1 (Strongly disagree) to 7 (Strongly agree), with higher scores indicated more conservative preferences. Party affiliation was assessed by asking participants to indicate how close do they feel to the political party on an eleven-point answer scale (0 = not at all close, 10 = very close) based on
previous research (Dalton, 2016) (M = 49.78, SD = 28.20). I also controlled for party familiarity by asking respondents whether they knew Spring or whether they visited and liked the Facebook page of Spring before participating in the study (no = 0, yes = 1) based on Kruikemeier et al. (2016). Majority of participants (88%) knew Spring, 27.3% visited and 9.1% liked their page. Furthermore, a Facebook account ownership and the frequency of Facebook use was assessed by asking participants if they currently have a Facebook account (no = 0, yes = 1) and how often they check their accounts (1 = never, 2 = less than once a month, 3 = weekly 2–5 times, 4 = once a week, 5 = once a day, 6 = 2–5 times a day, 7 = 6–10 times a day, 8 = 11–15 times a day, 9 = 16 or more times a day) as in Boerman et al. (2017) study. Majority of the participants (82.8%) had an account on this social media platform, while most of them (21.5%) reported checking their Facebook account two to five times a day. Lastly, a few backgrounds variables were included such as gender, age, nationality and educational level.
Results Randomization
To check for any differences between the experimental groups, I ran one-way ANOVAs and Chi-square tests. The results showed that experimental groups did not significantly differ with regards to age F (3,205) = 0.12, p = .948; gender, χ² (6, N = 209) = 5.47, p = .485; level of education, χ² (6, N = 209) = 4.26, p = .642; social F (3,205) = 0.46, p = .709 and economic
political ideology F (3,205) = 0.87, p = .460; having a Facebook account χ² (3, N = 209) = 0.21, p = .975; the frequency of Facebook use F (3,169) = 0.99, p = .398; extraversion F (3,205) = 1.63, p = .184; consciousness F (3,205) = 1.49, p = .218; neuroticism F (3,205) = 1.63, p = .183; party affiliation F (3,205) = 2.01, p = .114 and visiting χ² (3, N = 209) = 6.11, p = .106 or liking
Facebook page of Spring, χ² (3, N = 209) = 1.46, p = .692. Despite the random assignment familiarity with the party (knowing Spring) χ² (3, N = 209) = 8.97, p = .030 and agreeableness F
(3,205) = 3.44, p = .018 did differ between groups, and were therefore included as covariates in the analyses.
Exposure and Manipulation Check
As mentioned in the Method section, an important result of my experiment is that many participants that were exposed to the sponsorship disclosure did not recall seeing it. A Pearson’s chi-square revealed no significant difference between condition with and without sponsorship disclosure in terms of recall of the label χ² (1, N = 209) = 1.54, p = .251. The exposure check showed that only 19 participants recalled the batch, while 84 not when the sponsorship disclosure was present, and 13 respondents reported seeing the label, while 93 not when it was absent. When it comes to the recognition of the sponsorship disclosure, only 3 out of 103 participants in the “labelled” condition chose a correct disclosure. Any out of 106 respondents in “non-labelled” condition recognizes the correct batch. Therefore, the results indicate no significant difference between experimental groups χ² (1, N = 209) = 3.13, p = .118 in terms of disclosure recognition.
Hypotheses Testing
Hypothesis 1 predicts that a message tailored to respondents’ personality will positively influence affective and behavioural responses compared to message not tailored to respondents’ personality. In order to test this hypothesis, I conducted a one-way ANOVA with the type of message (tailored message vs. non-tailored message) as independent variable and perceived effectiveness, eWOM, source attitude, source feelings and voting intention as dependent variables. The analysis of variance showed a statistically significant effect of the message
congruency with recipients level of openness on likelihood to engage in electronic word of mouth F (1, 207) = 4.21, p = .041, η2 = .11 and a statistically significant but small effect on feelings towards the message source F (1, 207) = 4.27, p = .040, η2 = .02. Participants who received tailored message were significantly more likely to engage in eWOM (M = 3.25, SD = 1.29) than
participants who received non-tailored message (M = 2.89, SD = 1.25). In contrary, respondents that were presented with the mismatching message had more warm and favorable feelings towards the source (M = 53.74, SD = 25.62) than respondents presented with the matching message (M = 45.72, SD = 30.22). It should be noted that the assumption of equal variances in the population has been violated in regards to source feelings, Levene's F (1, 207) = 4.301, p = .039, thus Welch robust test of equality of means was run. Message congruency with participants level of openness explained 11% of the variance in the level of likeliness to engage in eWOM and 2% of the variance in the level of feelings towards the source, thus Spring. Additional analyses showed that significant effect on eWOM was mainly driven by the participants high on openness, who were significantly more likely to engage in eWOM when they were presented with tailored message F (1, 111) = 13.831, p < .001, η2 = .11. There was no significant difference in eWOM among people low in openness from congruent or incongruent condition F (1, 94) = 0.680, p = .412. Oppositely, the significant effect on feelings towards the source was driven by the participants who score low on openness, while they felt warmer and more favorable towards Spring when they have seen non-tailored message F (1, 94) = 8.39, p = .005, η2 = .08. No significant difference in feelings was found among participants who score high on openness in congruent or incongruent condition F (1, 111) = 0.09, p = .770. The results of ANOVA revealed also no significant effect of psychologically tailored message on perceived effectiveness F (1, 207) = 1.36, p = .245, attitude F (1, 207) = 0.31, p = .580 and voting intention F (1, 207) = 1.67, p = .197. Participants presented with the matching messages did not perceive message as more effective, did not have significantly more positive attitude towards the source, neither they were significantly more likely to vote for Spring than participants who saw mismatching messages. These results only partly support H1. Message tailored to respondents’ personality had a positive effect, but only on respondents’ likeliness to engage in eWOM. Conversely to my assumption,
messages matching respondents’ level of openness negatively influences their feelings towards the party. Other attitudinal and behavioural responses of participants were not affected by message congruency.
Hypothesis 2 forecasts that the effect of tailored message on affective and behavioural responses through the activation of persuasion knowledge is moderated by the sponsorship disclosure: The mediated relationship of the tailored message on respondents’ responses will be more negative when the message includes sponsorship disclosure. In order to test H3, I conducted a moderated mediation analysis with SPSS macro PROCESS (Model 7) using 5000 bootstrap samples to estimate 95% bias-corrected bootstrap confidence intervals (BCIs) for estimating indirect effects. The mediation analyses were run sixfold, separately for each dependent variable including effectiveness, eWOM, attitude, feelings, trustworthiness and voting intention. The independent variable was the type of the message (tailored vs. non tailored message), disclosure (present vs. absent) was moderator and persuasion knowledge were mediator. In line with the one-way ANOVA the results showed significant direct effect of tailored message on eWOM (b = 0.42, p = .016) but contractionary not on feelings towards the source (b = -7.46, p = .056). The result revealed a non-significant interaction effect between tailored message and presence of disclosure via persuasion knowledge on respondents’ responses (b = -0.19, p = 0.560). The index of moderated mediation was not significant for each of the dependent variable (see table 1). Therefore, moderated mediation did not occur. There was no conditional indirect effect of the participants’ exposure to tailored message and presence of disclosure on participants’ affective and behavioural responses via persuasion knowledge (see figure 2,3,4,5,6 and 7). These results indicate that persuasion knowledge does not mediate the effect of the tailored message on respondents’ affective and behavioural responses and this effect does not differ among
included a disclosure, that did not lead to recognition of the post as advertising, and in turn did not lead to more negative responses towards the message and the source in comparison when the message did not include disclosure. H3 thus has to be rejected.
Table 1
Indexes of moderated mediation
95% BCI
Outcome variable B BSE LL UL
disclosure vs. no disclosure Effectiveness 0.04 0.08 -0.10 0.22 eWOM 0.04 0.08 -0.10 0.21 Attitude 0.04 0.09 -0.13 0.23 Trustworthiness 0.06 0.11 -0.16 0.29 Feelings 0.34 0.90 -1.25 2.52 Voting intentions 0.07 0.70 -1.28 1.75
Note. N = 209; B, b-coefficient; BCI, bootstrap confidence; LL, lower limit; UL, upper limit; BSE, bootstrap standard error
Figure 2. Moderated mediation effect between the tailored message and presence of the disclosure via persuasion knowledge on perceived effectiveness of the message.
Figure 3. Moderated mediation effect between the tailored message and presence of the disclosure via persuasion knowledge on likeliness to engage in electronic word of mouth.
b = 0.32, p =.157 Indirect effect interaction b = -0.19, p =.560 Direct effect, b = 0.28, p = .146 I b = -0.23, p =.007 tailored message persuasion knowledge perceived effectiveness Disclosure, b = -0.03, 95% BCa CI [-0.14, 0.08] disclosure No disclosure, b = -0.07, 95% BCa CI [-0.21, 0.02] b = 0.32, p =.157 Indirect effect interaction b = -0.19, p =.560 Direct effect, b = 0.42, p = .016 I b = -0.22, p =.003 tailored message persuasion knowledge eWOM Disclosure, b = -0.03, 95% BCa CI [-0.14, 0.08] disclosure No disclosure, b = -0.07, 95% BCa CI [-0.20, 0.03]
Figure 4. Moderated mediation effect between the tailored message and presence of the disclosure via persuasion knowledge on attitude towards source.
Figure 5. Moderated mediation effect between the tailored message and presence of the disclosure via persuasion knowledge on perceived source trustworthiness.
b = 0.32, p =.157 Indirect effect interaction b = -0.19, p =.560 Direct effect, b = 0.20, p = .401 I b = -0.24, p =.021 tailored message persuasion knowledge source attitude Disclosure, b = -0.03, 95% BCa CI [-0.18, 0.08] disclosure No disclosure, b = -0.08, 95% BCa CI [-0.24, 0.03] b = 0.32, p =.157 Indirect effect interaction b = -0.19, p =.560 Direct effect, b = 0.33, p = .139 I b = -0.34, p =.001 tailored message persuasion knowledge source trustworthiness Disclosure, b = -0.05, 95% BCa CI [-0.23, 0.10] disclosure No disclosure, b = -0.11, 95% BCa CI [-0.29, 0.04]
Figure 6. Moderated mediation effect between the tailored message and presence of the disclosure via persuasion knowledge on feelings towards the source.
Figure 7. Moderated mediation effect between the tailored message and presence of the disclosure via persuasion knowledge on intention to vote.
Discussion
This is one of the first studies that address the effects of psychological targeting in the context of political communication. The purpose of the study was twofold. First, it aimed to
b = 0.32, p =.157 Indirect effect interaction b = -0.19, p =.560 Direct effect, b = -7.46, p = .056 I b = -1.79, p =.291 tailored message persuasion knowledge feelings No disclosure, b = -0.57, 95% BCa CI [-2.72, 0.61] disclosure Disclosure, b = -0.24, 95% BCa CI [-1.99, 0.74] b = 0.32, p =.766 Indirect effect interaction b = -0.19, p =.560 Direct effect, b = -5.25, p = .221 I b = -0.37, p =.843 tailored message persuasion knowledge intention to vote No disclosure, b = -0.12, 95% BCa CI [-1.84, 1.26] disclosure Disclosure, b = -0.05, 95% BCa CI [-1.15, 0.97]
examine the effects of messages tailored to recipients’ personality on affective and behavioral responses towards the message and the source. Second, it investigated the role of sponsorship disclosure and persuasion knowledge in this relationship.
With regard to my first aim, the research provides evidence for the effectiveness of psychological targeting. (Matz et al., 2016; Hirsch et al. 2012). Messages tailored the
respondents’ level of openness leads to higher likeliness to engage in word of mouth and more negative feelings towards the source. These paradoxical findings suggest that the effects of psychological targeting in the context of politics are more nuanced than in the fields of consumer goods or health communication. The contradictions of my results might be explained by the fact that there are possibly other factors affecting the relationship between tailored messages and the respondents’ responses. An explanation might be also found in the claim, that in general, it is more difficult to influence political feelings and voting (Graves & Matz, 2018). As these core political responses might be affected by political ideologies or identities (Motta & Fowler, 2016). Some of the studies on psychological persuasion highlight its promising role in fighting political apathy of voters (Dobber et al., 2017; Bodo et al., 2017). Therefore, my findings might indicate that political microtargeting is more effective in stimulating people online participation in politics, rather than manipulating their political attitudes and behaviors. More research is needed to support this claim. It is also worth to mention that post-hoc analyses revealed that open to experience participants were more likely to engage in eWOM and closed to experience
participants have less favorable feelings towards the party when they receive messages congruent with their personality. This finding might be explained by the study, that shows that openness to experience is associated with broad “persuasibility” (Geber, Huber, Doherry, Dowling
&Panagopoulos, 2012). Therefore, it might be suggested that our results indicate that open to experience people are easier to persuade, also by means of psychological microtargeting.
With respect to the second aim, the present study demonstrates that tailored messages with a sponsorship disclosure do not trigger persuasion knowledge and lead to more negative evaluations of the message and its source. These findings contradict with my hypothesis and the results from previous studies (Kruikemeier et al. 2016; Boerman & Kruikemeier, 2015; Boerman et al., 2017). They also defy the idea that sponsorship disclosures in the message inform the audience about its persuasive purpose of the message and consequently leads to negative assessments. These results might be explained by the fact that a new disclosure introduced on Facebook is not more effective in informing the audience about promotional content than its older version. This is in line with studies that found that sponsorship disclosure is difficult to notice, especially when is placed on the top of the message (Wojdynski & Evans, 2015). When a disclosure is not recalled, it doesn’t lead to negative responses towards the message and its source (Boerman et al. 2017). As 84% of participants in my study who were exposed to sponsorship disclosure did not remember it, my results might support this claim. Therefore, the results in respect to sponsorship disclosure and persuasion knowledge should be reviewed with caution. Limitations and implications for further research
This study has limitations that might provide interesting openings for future research. First of all, I focused only on a single personality trait: openness to experience. Future research can examine other personality traits, personality variables (ex. need for cognition, regulatory focus) or personality profiles (build out of a combination of certain traits) that might prove to be more effective. Secondly, I assessed participants’ openness to experience using IPIP-BFM-50 questionnaire. Other research might check if similar effects occur if personality is measured with different scales, for instance, commercial NEO PI-R, where openness constitutes of more factors. Furthermore, as self-report questionnaires might be affected by response biases (Podsakoff & Organ, 1986), thus further research can increase the reliability and ecological validity of this
study by assessing personality of participants from their digital footprints (Matz et. Al, 2016). Thirdly, it is possible that my stimuli material did not reflect both high-openness and
low-openness conditions, as it was only pre-tested on a relatively small group of participants (N=22). Moreover, the stimuli material was based on a growing amount of studies connecting personality traits with certain types of images or language concepts, which sometimes show contractionary results. Therefore, I propose a literature review as one of the future studies to zoom in the most relevant correlations. In regards to stimuli material, it might be interesting to differential
conditions in terms of emotional valance. As my stimuli were neutral, it would be interesting to see if negative messages, for instance about failures of competition, are more effective in “suppressing” the vote. It is claimed that such messages were used in the 2016 US election. Fourthly, an important limitation of my research is the fact that 84% of all participants did not notice sponsorship disclosure. This significantly decreased the power in my study and
consequently the ability to detect the results. A future study with a bigger sample size would be recommended to overcome this obstruction. Lastly, while selecting participants I did not use probability sampling method. Moreover, the results of this study are only limited to one country (Poland) and online psychological targeting performed on Facebook. As this study is only a first step towards assessing the effectiveness of psychological targeting in politics, future studies might extend generalizability of this research by conducting it in different geographical locations, across different (social media) or using samples randomly selected from population.
This study is a first step in what appears to be a promising line of research on the effectiveness of psychological targeting in politics. Although my findings have few limitations, they provide a shred of evidence for the effectiveness of messages congruent with personality profiles of the voters. The study offers a few contributions. It adheres to personality and politics literature that until now did not do not examine psychological microtargeting per se. Moreover,
the study expands research on psychological microtargeting, showing what role sponsorship disclosures and persuasion knowledge might have. These particular findings adhere to public debate about transparency around political advertising. Moreover, the research has implications for those interested in using psychological targeting in the political field, thus political parties, politicians, political campaigners but also for policymakers, as well as for voters.
References
Aird, M. J., Ecker, U. K., Swire, B., Berinsky, A. J., & Lewandowsky, S. (2018). Does truth matter to voters? The effects of correcting political misinformation in an Australian sample. Royal Society open science, 5(12), 180593.
Alwin, D. F. (1992). Information transmission in the survey interview: Number of response categories and the reliability of attitude measurement. Sociological methodology, 83-118. Bakker, B. N., Klemmensen, R., Nørgaard, A. S., & Schumacher, G. (2016). Stay loyal or exit the party? How openness to experience and extroversion explain vote switching. Political Psychology, 37(3), 419-429.
Bakker, B. N., & Lelkes, Y. (2018). Selling ourselves short? How abbreviated measures of personality change the way we think about personality and politics. The Journal of Politics, 80(4), 1311-1325.
Boerman, S. C., & Kruikemeier, S. (2016). Consumer responses to promoted tweets sent by brands and political parties. Computers in human behavior, 65, 285-294.
Boerman, S. C., Willemsen, L. M., & Van Der Aa, E. P. (2017). “This post is sponsored”: Effects of sponsorship disclosure on persuasion knowledge and electronic word of mouth in the context of Facebook. Journal of Interactive Marketing, 38, 82-92.
Boerman, S. C., Van Reijmersdal, E. A., & Neijens, P. C. (2012). Sponsorship disclosure: Effects of duration on persuasion knowledge and brand responses. Journal of
Communication, 62(6), 1047-1064.
Bodó, B., Helberger, N., & de Vreese, C. H. (2017). Political micro-targeting: a Manchurian candidate or just a dark horse? Internet Policy Review, 6(4).
Campbell, M. C., Mohr, G. S., & Verlegh, P. W. (2013). Can disclosures lead consumers to resist covert persuasion? The important roles of disclosure timing and type of response. Journal of Consumer Psychology, 23(4), 483-495.
Cawvey, M., Hayes, M., Canache, D., & Mondak, J. J. (2017). Personality and Political Behavior. In Oxford Research Encyclopedia of Politics.
Cesario, J., Higgins, E. T., Scholer, A. A. (2008). Regulatory fit and persuasion: Basic principles and remaining questions. Social & Personality Psychology Compass, 2, 444–463.
Czarnek, G., Dragon, P., Szwed, P., & Wojciszke, B. (2017). Kwestionariusz przekonań politycznych: własności psychometryczne. Psychologia Społeczna, 12(2 (41)).
Dalton, R. J. (2016). Party identification and its implications. In Oxford Research Encyclopedia of Politics.
Davis, K. C., Uhrig, J., Bann, C., Rupert, D., & Fraze, J. (2011). Exploring African American women's perceptions of a social marketing campaign to promote HIV testing. Social Marketing Quarterly, 17(3), 39-60
Dobber, T., Trilling, D., Helberger, N., & de Vreese, C. H. (2017). Two crates of beer and 40 pizzas: the adoption of innovative political behavioural targeting techniques. Internet Policy Review, 6(4).
Donnellan, M. B., Oswald, F. L., Baird, B. M., & Lucas, R. E. (2006). The mini-IPIP scales: tiny-yet-effective measures of the Big Five factors of personality. Psychological
assessment, 18(2), 192.
Facebook by numbers: Stats, Demographics & Fun Facts (2019, January 6). Omnicore agency, Retrieved from: https://www.omnicoreagency.com/
Gerber, A. S., Huber, G. A., Doherty, D., Dowling, C. M., & Panagopoulos, C. (2013). Big five personality traits and responses to persuasive appeals: Results from voter turnout
experiments. Political Behavior, 35(4), 687-728.
Graves C., Matz S., (2018 May 02). What Marketers Should Know About Personality-Based Marketing. Harvard Business Review. Retrieved from: https://hbr.org/
Hirsh, J. B., Kang, S. K., & Bodenhausen, G. V. (2012). Personalized persuasion: Tailoring persuasive appeals to recipients’ personality traits. Psychological science, 23(6), 578-581. Kalmoe, N., Gubler, J.R., & Wood, D.A. (2018). Toward conflict or compromise? How violent
Kam, C., & Simas, E. (2012) Risk attitudes, candidate characteristics, and vote choice. Public Opinion Quarterly, 76 (4), 747-760.
Kosinski, M., Stillwell, D., & Graepel, T. (2013). Private traits and attributes are predictable from digital records of human behavior. Proceedings of the National Academy of
Sciences, 110(15), 5802-5805.
Kruikemeier, S., Sezgin, M., & Boerman, S. C. (2016). Political microtargeting: Relationship between personalized advertising on facebook and voters' responses. Cyberpsychology, Behavior, and Social Networking, 19(6), 367-372.
Liberini, F., Redoano, M., Russo, A., Cuevas, A., & Cuevas, R. (2018). Politics in the Facebook Era. Evidence from the 2016 US Presidential Elections. CAGE Online Working Paper Series, (389).
Park, G., Schwartz, H. A., Eichstaedt, J. C., Kern, M. L., Kosinski, M., Stillwell, D. J., & Seligman, M. E. (2015). Automatic personality assessment through social media language. Journal of personality and social psychology, 108(6), 934.
Pennebaker, J. W., & King, L. A. (1999). Linguistic styles: Language use as an individual difference. Journal of personality and social psychology, 77(6), 1296.
Petty, R. E., & Cacioppo, J. T. (1986). The elaboration likelihood model of persuasion. In Communication and persuasion (pp. 1-24). Springer, New York, NY
Podsakoff, P. M., & Organ, D. W. (1986). Self-reports in organizational research: Problems and prospects. Journal of management, 12(4), 531-544.
Richard A., (2019, May 16). Protecting Elections in the EU. Facebook Newsroom, Retrieved from: https://newsroom.fb.com/
Schoonvelde, M., Brosius, A., Schumacher, G., & Bakker, B. N. (2019). Liberals lecture, conservatives communicate: Analyzing complexity and ideology in 381,609 political speeches. PloS one, 14(2), e0208450.
Schwartz, H. A., Eichstaedt, J. C., Kern, M. L., Dziurzynski, L., Ramones, S. M., Agrawal, M., & Ungar, L. H. (2013). Personality, gender, and age in the language of social media: The open-vocabulary approach. PloS one, 8(9), e73791.
Strus, W., Cieciuch, J., & Rowiński, T. (2017). Polska adaptacja kwestionariusza IPIP-BFM-50 do pomiaru pięciu cech osobowości w ujęciu leksykalnym. Roczniki Psychologiczne/Annals of Psychology, 17(2), 327-346.
Main, K. J., Dahl, D. W., & Darke, P. R. (2007). Deliberative and automatic bases of suspicion: Empirical evidence of the sinister attribution error. Journal of Consumer Psychology, 17(1), 59-69.
Matz, S. (2018, July 20). Psychological microtargeting could actually save politics. The Guardian, Retrieved from: https://www.wired.co.uk/
Matz, S. C., Segalin, C., Stillwell, D., Müller, S. R., & Bos, M. W. (2019). Predicting the personal appeal of marketing images using computational methods. Journal of Consumer Psychology, 29(3), 370-390.
Matz, S. C., Kosinski, M., Nave, G., Stillwell, D. J. (2017). Psychological targeting as an effective approach to digital mass persuasion. Proceedings of the National Academy of Sciences, 2017 10966.
McCrae, R. R., & Costa Jr, P. T. (1997). Personality trait structure as a human universal. American psychologist, 52(5), 509.
Mondak, J. J. (2010). Personality and the foundations of political behavior. Cambridge University Press.
Moon, Y. (2002). Personalization and personality: Some effects of customizing message style based on consumer personality. Journal of Consumer Psychology, 12(4), 313-325. Turow, Joseph, Michael X.D. Carpini, and Nora Draper (2012), “Americans Roundly Reject
Weber, C., Dunaway, J., & Johnson, T. (2012). It’s all in the name: Source cue ambiguity and the persuasive appeal of campaign ads. Political Behavior, 34(3), 561-584.
Wojdynski, B. W., & Evans, N. J. (2016). Going native: Effects of disclosure position and language on the recognition and evaluation of online native advertising. Journal of Advertising, 45(2), 157-168.
Wong, J., (2018 March 19). 'It might work too well': the dark art of political advertising online. The Guardian, Retrieved from: https://www.theguardian.com/
Van Noort, Guda, Edith G. Smit, and Hilde A. Voorveld (2013), “The Online Behavioural Advertising Icon: Two User Studies,” in Advances in Advertising Research. Vol. 4: The Changing Roles of Advertising, Sara Rosengren, Micael Dahlén, and Shintaro Okazaki, eds., Wiesbaden, Germany: Springer Fachmedien Wiesbaden, 365–78.
Zuiderveen Borgesius, F., Möller, J., Kruikemeier, S., Ó Fathaigh, R., Irion, K., Dobber, T., & de Vreese, C. H. (2018). Online political microtargeting: Promises and threats for
Appendix Appendix A
Stimuli material
Condition: High openness/No sponsorship disclosure Slogan: Free access to culture!
Condition: High openness/Sponsorship disclosure Slogan: Free access to culture!
Condition: Low openness/ No sponsorship disclosure Slogan: Free access to culture!
Condition: Low openness/ Sponsorship disclosure Slogan: Free access to culture!
Appendix B
