Beneficial or Harmful? How (Mis)Fit of Targeted Political Advertising on Social Media Shapes Voter Perceptions

Abstract

Targeted political advertising (TPA) on social media builds on tailoring messages to (groups of) individuals’ characteristics based on user data. Questions have been raised about the impact of TPA on recipients and society. In this study, we focus on the fit of TPA, that is, the congruence between TPA and recipients’ preferences, and draw on congruity theories, social identity theory (SIT), and persuasion knowledge. In a two-wave panel study (N = 428) during a Viennese state election, we investigated the relationships between individuals’ perceived fit and misfit of TPA on perceptions about the manipulative intent as well as the benefits and harms for democracy. The findings showed that perceived fit of TPA at Time 1 decreased perceived manipulative intent and increased perceived benefits of TPA at Time 2. The perceived misfit of TPA at Time 1 did not influence individuals’ perceptions at Time 2, and perceptions about the harms of TPA to democracy stayed stable. Findings imply that political campaigners might benefit from targeting but raise questions about individuals’ defense mechanisms against the persuasive technique.

Every day an adult makes approximately 35,000 decisions (Hoomans 2015). While for certain decisions we are left on our own, settings exist in which technology fills in to select the content that might be relevant for us, for example, through algorithm-based recommender systems (e.g., Beam 2014) or targeted advertising on social media (e.g., Van den Broeck, Poels, and Walrave 2020). In the commercial sector, online behavioral advertising (OBA) takes advantage of today’s digital world. OBA involves two features: “(1) the monitoring or tracking of consumers’ online behavior and (2) use of the collected data to individually target ads” (Boerman, Kruikemeier, and Zuiderveen Borgesius 2017, p. 364). Hence, individual behavior on social media influences the algorithmic function of assigning individuals into categories (Thorson et al. 2019) to deliver content to chosen segments of social media users (Dommett 2019).

In recent years, OBA practices have also manifested in political campaigning. In online political advertising, “collecting information about people, and using that information to show them targeted political advertisements” is referred to as political microtargeting (Zuiderveen Borgesius et al. 2018, p. 82). To maximize the effectiveness of their digital campaigns, parties try to create a fit between the message receiver and the message itself, using targeted political advertising (TPA). TPA is “the direct transmission of stimuli, which are formed based on the preferences and characteristics of an individual” (Papakyriakopoulos et al. 2018, p. 2). Thus, TPA aims to send fitting—in other words, congruent—advertising to recipients (see Bleier and Eisenbeiss 2015; Papakyriakopoulos et al. 2018).

Political marketing differs from commercial marketing because, on one hand, voting decisions have large-scale and long-standing implications. On the other hand, “the political party or candidate is a complex intangible product,” which is difficult for voters to unpack (Lock and Harris 1996, p. 15). Even more than in commercial advertising, people may be interested in keeping their opinions private to avoid derivations that do not reflect their actual political views (Zuiderveen Borgesius 2015). When campaigners’ decisions are based on data derived from personal search behaviors or interests, voters can become vulnerable to discrimination, manipulation, polarization, or misinformation, as well as privacy infringements (Roemmele and Gibson 2020; Zuiderveen Borgesius 2015). In contrast, scholars also highlight the potential benefits of microtargeting practices connected to directly speaking to the needs of different voter groups and reengaging less interested parts of the electorate (Haenschen and Jennings 2019; Zuiderveen Borgesius et al. 2018). At the same time, we still know little about how voters perceive this form of advertising, how it affects them, and if they accept targeting as a campaigning tool.

To shed light on this issue, this study focuses on how perceptions of the fit and misfit of TPA shape individuals’ acceptance of the practice. Specifically building on the self-congruity theory (Sirgy 2018), schema incongruity processing theory (Mandler 1982), and social identity theory (SIT; Tajfel and Turner 1986), we investigate how exposure to TPA that aligns or misaligns with preferences affects citizens’ perceptions of manipulative intent, as well as the benefits and harms to democracy regarding TPA. Focusing on individuals’ perceptions of manipulative intent, we investigate an important dimension of the persuasion knowledge model by Friestad and Wright (1994). The development of persuasion knowledge is decisive because it helps recipients to “identify how, when, and why marketers try to influence them,” as well as how to react to this influence in an appropriate manner (Friestad and Wright 1994, p. 1). By investigating perceived benefits and harms to democracy, we contribute to understanding how voters perceive the implications of TPA (i.e., in terms of potential promises and threats).

The current study advances the field in two ways: First, to the best of our knowledge, no research exists that simultaneously investigates the perceived implications of TPA on both the individual level (i.e., the perceived manipulative intent) and a societal level (i.e., benefits and harms to democracy). Second, longitudinal research in this area is still scarce, especially with a focus on European campaigns. Therefore, our study gives insights into an unfolding digital campaign under conditions of the E.U. General Data Protection Regulation (GDPR), complementing current studies that come predominantly from a less regulated U.S. context. Article 8 of the Charter of Fundamental Rights of the European Union (2012) describes the right to protection of personal data that is granted to every E.U. citizen. This regulation and the GDPR (see E. U. General Data Protection Regulation 2016) state the duty of each company to inform and ask citizens for their consent before any personal data can be stored or used—for advertising purposes, for example. Users of websites or social media experience this regulation through messages that ask for permission to collect specific data or to use cookies. Because these regulations differ from regulations in the United States, it is essential to investigate the perception of TPA in Europe. To fill these gaps, we conducted a two-wave panel survey between August 2020 (Wave 1) and October 2020 (Wave 2) during the Viennese state election in Austria. In this panel survey, we relied on self-reported measures to investigate the relationships between individuals’ perceptions of the (mis)fit of TPA and their perceptions of individual- and societal-level implications.

Political Campaigning in the Digital Age

TPA stems from classical political campaigning with the significant difference that it takes large numbers of data points of potential voters into the equation to tailor political messages or advertising (e.g., Kreiss 2017; Zuiderveen Borgesius et al. 2018). Even though the separation of the electorate into groups has always been a component of campaigning to address voters most efficiently (König 2020), new technologies relying on algorithms and targeting are considered promising practices to grasp better the increasing heterogeneity of voters and their political stances (Pilditch and Madsen 2021). These significant changes might even have set the path for a new era of election campaigning, which scholars define as the data-driven or fourth era of campaigning (Roemmele and Gibson 2020). Roemmele and Gibson (2020) further describe four key shifts in the era of data-driven campaigning, namely, (1) reliance on digital technology and large amounts of data; (2) a “more developed and networked approach to voter communication”; (3) the production of “much more fine-grained and personality-based understanding of the persuadable electorate,” hence microtargeted messages in campaigns; and (4) “more internationalized” campaigns (p. 597). While microtargeting practices profit from a less restrictive legal framework and have been successful in the United States, they face stricter legal constraints in the European context due to privacy and data collection regulations (e.g., E. U. General Data Protection Regulation 2016; Papakyriakopoulos et al. 2018). Nevertheless, microtargeting practices are also gaining momentum in European campaigns (Bennett 2016; Dobber et al. 2017) because social media users’ data can be employed for targeting practices due to opt-in consent and users’ willingness to share a large amount of data on their preferences and interests (Papakyriakopoulos et al. 2018).

In this regard, TPA has increasingly attracted scholarly attention to its effects as well as its democratic and societal implications. Effect studies have investigated recipients’ responses from various angles. One strain of research focused on how individuals respond to disclosures of sponsored/targeted advertising and how it influences recall or the persuasive impact (e.g., Boerman and Kruikemeier 2016; Kruikemeier, Sezgin, and Boerman 2016). Research by Boerman and Kruikemeier (2016) found that the recognition of advertising, that is, persuasion knowledge, reduced intentions to engage in online behavior, such as sharing a tweet; decreased source trustworthiness; led to lower source attitudes; and caused higher skepticism toward the tweet when the message was sent by a political party rather than a brand (Boerman and Kruikemeier 2016). Contrarily, research found that the critical processing and effect of TPA is also highly determined by whether the portrayed party matches an individual’s party preferences (Binder et al. 2022).

Another research strain focuses on the importance of perceptions of TPA. Experimental research revealed that the effectiveness of a message in terms of attitudes and behavioral intentions depends on perceived personalization rather than actual personalization (Li 2016). A panel study by Matthes et al. (2022) investigated the longitudinal relationships between perceived online political microtargeting and perceptions about trust in democracy and political interest. Their findings highlighted that perceived exposure to TPA can be simultaneously detrimental and useful, reducing individuals’ trust in democracy and increasing their political interest. These findings resonate with the overall debate about the potential benefits and threats of political targeting practices for individuals, society, and democracy (Barocas 2012; Zuiderveen Borgesius et al. 2018). Despite these initial insights, recipients’ perceptions of the impact of TPA on democracy remain understudied.

In a series of experimental studies, Li (2016) found evidence that a message must be perceived as personalized to yield positive effects, which might depend on the match between the message and the message receiver. Li (2016) further outlines that inaccurate information in a message, causing a mismatch, will enhance recipients’ feelings that they were not the intended target of the message. Conversely, messages containing accurate information, creating a match, are more easily perceived as personalized. This rationale points to another strain of research investigating the effects of (mis)matching TPA. Scholars have investigated the influence of varying forms or levels of targeting procedures (e.g., Binder et al. 2022; Zarouali et al. 2019) and the role of different congruency components, for example, personality traits (Krotzek 2019; Zarouali et al. 2020), identity (Holman, Schneider, and Pondel 2015), party preferences (Binder et al. 2022; Lavigne 2021), or issues (Endres 2020).

The Role of Fit and Misfit in Targeted Political Advertising

Targeted advertising positions itself within the key shifts of data-driven campaigning, profiting from the “fine-grained and personality-based understanding of the persuadable electorate” (Roemmele and Gibson 2020, p. 597). Hence, one aim of TPA is to create a match (i.e., a fit) between the advertised message and the characteristics of the message recipients (Bleier and Eisenbeiss 2015; Papakyriakopoulos et al. 2018). As a result, the fit constitutes a central feature and contributes to the effectiveness of an advertising appeal. A recent study has investigated the influence of psychometric profiling—such as the matching of messages with psychological traits—and found that individuals are more persuaded by TPA that was in line with their personality in terms of more positive political attitudes and voting intentions (Zarouali et al. 2020). This study also showed that when the personality trait matched the message frame—that is, introverted individuals matched with a fear-based ad and extroverted individuals matched with an enthusiasm-based ad—it increased individuals’ message elaboration, which then increased their voting intentions. While Krotzek (2019) similarly found that messages congruent with individuals’ personality traits led to more positive feelings toward a candidate, no evidence was found for an increase in the likelihood of voting. Moving beyond psychometric targeting, Holman, Schneider, and Pondel (2015) found in their experiment that targeting women with messages aimed at women (identity-based targeting) increased their propensity to vote for a candidate.

However, while the fit of targeted messages can be effective, people might still be confronted with misfitting TPA. In a U.S. survey, around 4 in 10 citizens indicated that their interests were not well reflected in the ads that target them (Auxier et al. 2019), which indicates that mistargeting might be a common experience in individuals’ media consumption. While algorithms are often perceived as highly sophisticated, algorithmic profiling can also result in various inaccurate categorizations based on changes in individuals’ preferences or wrong inferences (Büchi et al. 2021). In-depth interviews by Van den Broeck, Poels, and Walrave (2020) revealed that when targeted advertising did not fit, people reflected on the usage of their data and reported feelings of disappointment or frustration. These feelings stemmed from a perceived violation of the expected social contract that users enter with Facebook, which includes relevant—i.e., fitting—targeted advertising (Van den Broeck, Poels, and Walrave 2020). As a result, mistargeting might foster mechanisms of algorithmic disillusionment, referring to individuals’ realization that algorithms are “less powerful and useful but more fallible and inaccurate than previously thought” (Felzmann et al. 2019, p. 7; see also Büchi et al. 2021). Mistargeting recipients by addressing misfitting TPA might also come with considerable risks for political parties, because this tactic might backfire. Hersh and Schaffner (2013) found in their experimental study that citizens penalize politicians who target them based on false categorizations. Based on these findings, misfitting TPA might foster negative effects.

In the present study, we investigate how perceptions of fit and misfit might shape individuals’ subsequent evaluations of TPA on an individual and societal level. As such, we operationalize fit and misfit as separate concepts. Throughout a digital political campaign, different constellations of fit and misfit might occur. First, fit and misfit of TPA might happen simultaneously, with individuals receiving messages that are in line with their preferences as well as messages that come from parties or cover topics that misalign with their preferences. Second, it is possible that neither fit nor misfit of TPA might occur. In this situation, individuals might be confronted with predominantly random content in which they have little interest but which also does not contradict their preferences. Third, individuals might be exposed to TPA with high fit but not with misfit, or vice versa. Hence fit and misfit cannot be considered opposites but must be understood as separate constructs when considering the course of a campaign.

These processes become especially relevant in a multiparty system, such as the one in Austria. First, the numbers of possible fit and misfit options are likely to increase with a bigger number of political parties in the run for electoral positions. Second, it is likely that while some parties might evoke strong feelings of fit or misfit, others might be considered irrelevant or random content. Considering fit and misfit as conceptually the same would therefore not comprehensively represent the real-life phenomenon of TPA in a multiparty system.

To account for these processes, we opted to assess (mis)fit perceptions in a summary manner and specifically focus on two fit components, namely the match between TPA and recipients’ party preferences and their issue interests. Political preferences or partisanship and issues constitute important aspects of electoral campaigns (Hillygus and Shields 2009). Individuals’ partisan preferences inform parties’ decisions on whom to target as they primarily aim to reach voters likely to support their candidates (Barocas 2012). Targeting potential supporters was found to render partisan identification stronger and to lead to electoral support (Lavigne 2021). Hence, the political party and individuals’ partisan attachments are an essential part of individuals’ self-concept (see Green, Palmquist, and Schickler 2004). The congruence between the party and the recipient is, therefore, integral to our concept. In our study, we built on Binder et al.’s (2022) conceptual definition of political fit, that is, the congruence between participants’ party preference and the party as a message sender of TPA. We then translated it to a different methodological context, namely, an unforced longitudinal setting, and extended the concept of political fit by incorporating the issue component and additionally measuring the misfit of TPA. Overall, issues matter in campaigning because parties might be evaluated based on their issue engagement (Meyer and Wagner 2016) or their issue ownership (Budge and Farlie 1983; Petrocik 1996). Issue ownership theory describes how political parties are expected to focus on issues they own (Walgrave et al. 2015) and can influence voters in terms of preferring parties owning issues that the voter perceives as important (Lefevere, Tresch, and Walgrave 2015). This highlights the interconnectedness between political parties, issue stances, and voters’ preferences. Also, in the targeting context issues matter in terms of issue-based targeting, Endres (2020) found that targeting voters with messages from the opposing party, that is, Democrats being targeted by the Republican Party, was associated with less support for Barack Obama when the issue was congruent between the targeted voter and Mitt Romney.

The Influence of Fit and Misfit on Perceived Manipulative Intent

The fit or misfit of TPA can influence how recipients react to an advertising appeal and how it shapes their evaluations and reactions (Binder et al. 2022; Krotzek 2019; Zarouali et al. 2020). To evaluate and deal with advertising appropriately, individuals need to activate knowledge structures (Friestad and Wright 1994). The persuasion knowledge model (PKM) describes these processes and focuses on various dimensions (Boerman et al. 2018). One dimension of particular importance in the targeting context is manipulative intent (e.g., Binder et al. 2022). Perceived manipulative intent can be defined as “consumer inferences that the advertiser is attempting to persuade by inappropriate, unfair, or manipulative means” (Campbell 1995, p. 228). According to PKM, individuals incorporate such normative ideas about the appropriateness of persuasive tactics into their evaluation of ads and advertisers (Friestad and Wright 1994).

To theoretically understand the psychological mechanisms behind the influence of (mis)fitting TPA on the perceived manipulative intent, we draw on self-congruity theory. This theory establishes the basis for understanding how fit can impact individuals’ attitudes and behavioral intentions (Sirgy 2018). It is suggested that people generally prefer consistency with self-concept (Aaker 1999). Self-concept can consist of “more than one type of self-perspective,” or self-images (Sirgy 1985, p. 195), and can be extrapolated to different contexts, such as commercial advertising (e.g., brand-user image, Sirgy 2018) or politics (i.e., political self-concept). Political self-concept can be defined based on an individual’s self-perception of being a “political person” (Lane et al. 2019, p. 51), based, for example, on party partisanship (Knobloch-Westerwick and Meng 2011) or political interest (Lane et al. 2019). Accordingly, we examine the political self-concept regarding partisanship and issue interests in the present study. Based on self-congruity theory, research found that advertising in line with one’s self-concept has positive effects on an individual’s attitudes toward a product or buying intentions (Hong and Zinkhan 1995; Sirgy 2018). Likewise, when emotional targeting based on individuals’ moods matched the product category, consumers reacted more positively (Garaus, Wagner, and Rainer 2021).

Similarly, schema incongruity processing theory, proposed by Mandler (1982), offers another framework to understand the different effects related to congruity (i.e., fit) and incongruity (i.e., misfit). Schemas are “representations of experience that guide action, perception, and thought” (Mandler 1982, p. 3). According to Mandler (1982), individuals evaluate newly encountered information against existing schemas. When information is congruent with existing schemas, it is perceived as more familiar and acceptable (Halkias and Kokkinaki 2013). As a result, it is processed more easily, does not require further cognitive arousal, and can lead to positive evaluations (Halkias and Kokkinaki 2011; Mandler 1982). To be able to engage in perceived manipulative intent evaluations, we argue that individuals need to pay attention to the information and need cognitive arousal, which again is stimulated more by misfitting TPA than by fitting TPA. Thus, we hypothesize:

H1: The perceived fit of TPA at Time 1 reduces the perceived manipulative intent at Time 2.

Contrarily, we assume different mechanisms for perceived misfitting TPA, also based on schema incongruity processing theory (Mandler 1982). When individuals encounter incongruent information, it functions as an “interruption of expectations and predictions” (Mandler 1982, p. 21). Hence, people pay more attention to solving the incongruency, which then leads to increased cognitive arousal and can influence evaluations (Halkias and Kokkinaki 2011, 2013; Mandler 1982). Research has found that misfitting content could encourage recipients to critically evaluate advertising, which can further initiate evaluations about how appropriate an ad is (Germelmann et al. 2020; Maheswaran and Chaiken 1991). When advertising strategies increase individuals’ processing, it might activate their persuasion knowledge along with their “inappropriateness judgments,” which may induce “negative processing consequences, such as inferences of manipulative intent” (Campbell 1995, p. 228). In this vein, Campbell (1995) found that when the fit of a borrowed interest appeal in advertising was low, it led to lower personal benefits and increased perceptions of manipulative intent. Likewise, political advertising not in line with individuals’ opinions—cross-cutting advertising—can be argued to relate to perceptions of manipulative intent (Matthes and Marquart 2015). Based on the outlined theoretical and empirical insights, we hypothesize as follows:

H2: The perceived misfit of TPA at Time 1 increases the perceived manipulative intent at Time 2.

The Influence of Fit and Misfit on Perceived Benefits and Harms for Democracy

Going beyond individuals’ perceptions about the appropriateness of TPA, it is also relevant to address how this advertising practice is evaluated regarding societal implications. In this context, positive and negative effects on democracy are widely discussed. Regarding potential benefits, it is argued that targeting practices could lead to higher degrees of mobilization of voters, especially of those more active on social media (Zuiderveen Borgesius et al. 2018) and younger voters (Haenschen and Jennings 2019). In addition, targeting practices might also increase voters’ interests or be perceived as beneficial because of receiving more relevant content (Van den Broeck, Poels, and Walrave 2020; Zuiderveen Borgesius et al. 2018).

At the same time, concerns are raised about TPA’s potential drawbacks. For example, such advertising techniques can potentially exclude certain voter groups (Barocas 2012) and enhance polarization (Roemmele and Gibson 2020). Furthermore, TPA can influence the public debate (see Jamieson 2013) by changing individuals’ information-seeking habits, for example, in the form of chilling effects, which imply that people refrain from engaging in online activities or looking for information online (Dobber et al. 2019).

While these possible benefits and harms for democracy are discussed among scholars, little is known about how individuals perceive the impact of TPA on democracy overall. To fill this gap, we focus on individuals’ perceptions of the democratic implications of TPA. By doing so, we consider benefits and harms not as parallel but as separate concepts. We operationalized perceived benefits in terms of a more engaged and interested electorate, more diverse election campaigns, and an increase in electoral turnout (see Zuiderveen Borgesius et al. 2018). Perceptions of harms were operationalized on a more abstract and fundamental level, based on an overarching assessment of whether TPA is perceived as harmful to free democratic elections and democracy. Whether individuals perceive benefits or harms might be driven by the extent to which the fit or misfit is perceived. This relationship can be explained based on the assumptions of SIT.

According to SIT (Tajfel and Turner 1986), belonging to one group implies favoring and elevating the ingroup and harboring resentment and stricter feelings toward the outgroup (Huddy and Bankert 2017; West and Iyengar 2020). Research has shown that partisanship can be defined as a social identity (Huddy and Bankert 2017) and an important component of individuals’ self-concepts, especially during election campaigns (West and Iyengar 2020). Therefore, SIT provides a suitable framework to explain why recipients might react differently to messages stemming from their own party compared to those from an opposing party. Based on the assumptions of SIT, we argue that when individuals perceive the fit of TPA, it might lead to more positive evaluations of the targeting strategy for democracy at a later point in the campaign. This mechanism is based on ingroup dynamics, related to elevating one’s own ingroup while disdaining the outgroup (Huddy and Bankert 2017; West and Iyengar 2020). In other words, TPA from the preferred political party and covering topics that are personally interesting might be perceived as less harmful. This means, driven by party heuristics, fitting TPA increases perceived benefits and reduces perceived harms. Hence, we hypothesize:

H3: Perceived fit of TPA at Time 1 (a) increases perceived benefits and (b) decreases perceived harms for democracy at Time 2.

As suggested by SIT (Tajfel and Turner 1986), outgroups opposing one’s ingroup can be perceived negatively, possibly inducing feelings of opposition (West and Iyengar 2020). When individuals receive misfitting TPA, which implies that it came from an opposing party and is not in line with their interests, it might lead to negative evaluations of the parties’ actions. In our context, it would imply that misfitting TPA increases such negative evaluations, highlighting the perceived harms of TPA rather than the perceived benefits for democracy over the course of the election campaign. We therefore hypothesize:

H4: Perceived misfit of TPA at Time 1 (a) decreases perceived benefits and (b) increases perceived harms for democracy at Time 2.

Method

Because targeted political messages and ads are more likely to be received and noticed during an election campaign (Dobber et al. 2019), we conducted our two-wave panel study during the campaign period of the Viennese state election in 2020. Panel surveys allow accounting for the dynamic character of election campaigns because the change in the responses of the same individuals can be investigated (Bartels 2006). Accounting for preexisting attitudes or opinions is one of the advantages of panel surveys because prior attitudes can influence the “amount and direction of across-time change” (Taris 2000, p. 56). Omitting such variables can cause potential bias in estimates (Bartels 2006). Furthermore, panel surveys constitute a stricter test of relationships between different variables compared to cross-sectional survey data, as they allow researchers to model the correct temporal order of relationships (Bartels 2006; Burkholder and Harlow 2003), which allows a hint at directionality (e.g., Bartels 2006). While experimental research would be needed to evaluate causality, panel studies profit from high levels of external validity because they are embedded in real life and lie outside forced-exposure settings.

The data collection of our two-wave panel survey was administered by Dynata, a professional polling company, between August 2020 (Wave 1) and October 2020 (Wave 2), until Election Day on October 11, 2020. Similar to previous research (e.g., Dimitrova et al. 2014), we left approximately six weeks of time difference between the waves, which is deemed suitable to see campaign effects. This study was part of a more extensive panel survey on the topic of political communication during the Viennese state election. The data from the same sample were used in another article on TPA (Stubenvoll et al. 2022) and other unrelated political topics (e.g., Nanz and Matthes 2022). However, independent and dependent variables are original to this study. To participate in the study, participants had to give their consent and needed to be eligible to vote in the election, for which Austrian citizenship and having main residence in Vienna were required. Furthermore, participants had to use social media to some extent in both waves. In addition, we applied a cutoff point of less than 10 minutes for the 25-minute survey in Wave 1 to exclude participants who took too little time, which raises the risk that instructions and questions were not read carefully. This is a conservative approach relying on response exclusion based on response time to reduce potential biases (Leiner 2019).

Sample

Based on these criteria, our sample consisted of 428 participants in both waves (N_W1 = 718 respondents, N_W2 = 428 respondents). The calculated retention rate of the completed cases between both waves was 59.61%. We applied quotas based on age (18 to 65 years, M = 43.32, SD = 13.16) and gender (54% female, 46% male). The educational backgrounds were heterogeneous among participants (25.9% lower, 46.5% medium, 27.6% higher education). No significant differences were found for gender (χ² (1) = .47, p = .495) or education (χ² (6) = 3.45, p = .745) of participants who participated only in Wave 1 and participants who participated in both waves. However, we found significant differences regarding participants’ age. Respondents who participated in Waves 1 and 2 were significantly older (M = 43.32, SD = 13.16) than respondents participating in Wave 1 only (M = 39.14, SD = 13.44, t (716) = −4.13, p < .001). We received ethical approval from the Institutional Review Board (IRB) of the Department of Communication at the University of Vienna (approval ID 20200722_017).

Measures

All participants received a short explanation of what TPA on social media is, stating that TPA refers to ads on social media that are specifically tailored to certain groups of people, which can be based on different data, for example, demographic data, interests, or click behavior (similar definition to Facebook 2022). Thus, the provided definition highlights that TPA uses individuals’ data to deliver messages to specific segments of voters, which is a key feature of TPA practices (Dommett 2019). If not indicated differently, we applied seven-point Likert scales ranging from 1 (Do not agree at all) to 7 (Totally agree). For correlations among the main variables, see Supplemental Online Appendix C, Table S1. The data can be found at https://osf.io/msp58/.

Table 1. Results of the structural equation model controlling for autoregressive relationships.

Predictors	Dependent Variables
	Perceived Manipulative Intent (W2)	Perceived Benefits for Democracy (W2)	Perceived Harms for Democracy (W2)
	b (SE)	b (SE)	b (SE)
Perceived fit	−0.15 (0.06)*	0.14 (0.06)*	0.01 (0.06)
Perceived misfit	0.11 (0.07)	−0.04 (0.06)	0.10 (0.07)
Age	0.01 (0.01)	−0.01 (0.01)*	0.00 (0.01)
Gender (female = 1)	0.09 (0.16)	0.23 (0.12)	0.20 (0.16)
High education	−0.11 (0.22)	0.09 (0.17)	0.37 (0.22)
Medium education	−0.29 (0.19)	0.03 (0.15)	0.06 (0.18)
Political ideology (W2)	0.01 (0.03)	0.06 (0.03)	0.02 (0.03)
OBA knowledge (W2)	0.06 (0.05)	−0.14 (0.04)**	−0.06 (0.05)
Political interest	0.08 (0.08)	0.09 (0.05)	−0.08 (0.08)
Political distrust	0.11 (0.07)	−0.04 (0.06)	0.26 (0.07)***
Privacy concerns	0.16 (0.07)*	−0.04 (0.05)	0.14 (0.07)
Perceived manipulative intent (W1)	0.38 (0.07)***	—	—
Perceived benefits for democracy (W1)	—	0.39 (0.07)***	—
Perceived harms for democracy (W1)	—	—	0.32 (0.08)***
R²	0.34	0.36	0.31

Note. N = 428. OBA = online behavioral advertising. Robust maximum likelihood estimation was applied. Robust fit indices are reported χ² (617) = 857.83, p < .001, CFI = .97, TLI = .96, RMSEA = .03, SRMR = .05. The control variables age, gender, education, political interest, political distrust, and privacy concerns were measured in Wave 1 (W1); OBA knowledge and political ideology in Wave 2 (W2).

*p < .05; **p < .01; ***p < .001.

Independent Variables

To measure the extent to which participants perceived TPA encountered on social media during the election campaign to be fitting or misfitting, we applied the measures of perceived fit (M_W1 = 3.18, SD_W1 = 1.53, Cronbach’s α_W1 = .92) and perceived misfit (M_W1 = 3.41, SD_W1 = 1.41, Cronbach’s α_W1 = .80) of TPA. The fit and misfit constructs were self-developed, and each comprised four items: two about political (mis)fit and two about issue (mis)fit. Participants had to evaluate perceived fit in the following way: When they saw TPA on social media, to what extent (1) it corresponded to their general party preference; (2) it came from a party that they tend to like; (3) it addressed issues that interest them; and (4) it addressed issues that reflect their interests. The perceived misfit was gauged by asking: When participants saw TPA on social media, to what extent (1) it contradicted their general party preference; (2) it came from a party that they tend to dislike; (3) it addressed issues that do not interest them; and (4) it addressed issues that are contrary to their interests.

Drawing on the issue ownership theory (e.g., Bélanger and Meguid 2008), we accounted for the political and topical level of fit and misfit perceptions. We gauged individuals’ perceptions in a summary manner to account for the (co)existence of the perceived fit and the perceived misfit of TPA throughout the campaigning phase and to acknowledge the dynamics of a multiparty system. Notably, the study was conducted in a multiparty context, and individuals had to remember which ads they had seen in the past. Thus, it would have taken considerable memory effort on the participants’ side to memorize ads they had seen by each party to then calculate the fit based on individual party and issue preferences. Our measurement of perceived (mis)fit accounts for this because participants could give a more intuitive overall assessment of perceived political and issue (mis)fit of TPA.

To confirm the theorized dimensional structure of our measures, we conducted a confirmatory factor analysis (CFA). The CFA confirmed the two-dimensional structure, indicating that fit and misfit constituted two separate factors, showing an acceptable model fit based on robust model fit indices, χ² (19) = 44.02, p = .001, comparative fit index (CFI) = .98, Tucker–Lewis index (TLI) = .96, root mean square error of approximation (RMSEA) = .07, standardized root mean squared residual (SRMR) = .05. We tested the two-dimensional solution against a one-factorial model, which further indicated that the two-factorial structure is reasonable (Δ χ² (1) = 105.41, p < .001).

Dependent Variables

We assessed perceived manipulative intent using three items based on Campbell (1995) to match our study and the panel survey context (M_W1 = 4.45, SD_W1 = 1.81, M_W2 = 4.55, SD_W2 = 1.76, Cronbach’s α_W1 = .90, α_W2 = .91). Participants were asked to rate the extent to which TPA on social media (1) tries to manipulate them; (2) tries to push them in a certain direction; and (3) tries to manipulate them in a way that they do not like. Perceived benefits for democracy were assessed with four items (M_W1 = 3.13, SD_W1 = 1.42, M_W2 = 3.17, SD_W2 = 1.32, Cronbach’s α_W1 = .90, α_W2 = .87). Participants had to evaluate the extent to which they agreed that TPA on social media was good because (1) it makes voters more willing to vote; (2) it gets voters more engaged; (3) parties have to cater to the needs of very different target groups; and (4) it allows parties to address the concerns of different segments of the population. Perceived harms for democracy were measured using two items (M_W1 = 3.84, SD_W1 = 1.71, M_W2 = 3.92, SD_W2 = 1.67, Spearman–Brown coefficient ρ_W1 = .87, ρ_W2 = .85). Participants had to evaluate the degree to which they agreed that TPA on social media (1) harms free, democratic elections; and (2) endangers democracy. Our scales of perceived benefits and harms of TPA for democracy were self-constructed but inspired by arguments in the field (Zuiderveen Borgesius et al. 2018).

Control Variables

In Wave 1, we controlled for participants’ age, gender, and education. We measured participants’ political interest on a scale from 1 (Not interested at all) to 7 (Very interested) based on Reiter and Matthes (2021; M = 5.28, SD = 1.47, Spearman–Brown coefficient ρ = .82) and their political distrust (Craig, Niemi, and Silver 1990; M = 4.20, SD = 1.54, Spearman–Brown coefficient ρ = .81). We controlled for participants’ privacy concerns based on Hsu and Lin (2018) as well as Mani and Chouk (2017) and adapted these to the social media context (M = 4.82, SD = 1.69, Cronbach’s α = .89). In Wave 2, we assessed participants’ political ideology, ranging from 0 (Left) to 10 (Right; M = 4.52, SD = 2.29) and measured individuals’ knowledge about OBA based on a scale by Smit, Van Noort, and Voorveld (2014). Participants indicated which knowledge items were right or wrong but could also choose the answer option I do not know. Correct items were summed up to a scale ranging from 0 to 6 (M = 4.33, SD = 1.44).

We included the mentioned control variables because of their relevance to our model. Political dispositions like political distrust (Chan 1997; Hetherington and Globetti 2002) and political interest (Lovejoy, Cheng, and Riffe 2010; Nelson et al. 2021) can affect individuals’ attention and processing. Privacy concerns might influence individuals’ perceptions of the (mis)fit of TPA. For example, an increased alert state of possible privacy infringement might also increase the awareness that personal data were used (e.g., van Doorn and Hoekstra 2013). At the same time, privacy concerns could also affect the evaluative outcomes of TPA (e.g., Dobber et al. 2019). Furthermore, we theorized that OBA knowledge might affect individuals’ recognition of targeted advertising and their overall evaluation (Friestad and Wright 1994). We included these variables as control variables to account for potential spurious relationships and to minimize the risks of omitted variable bias (Kline 2011). An overview of the measures is shown in Supplemental Online Appendix A.

Data Analysis

We applied structural equation modeling (SEM) with the lavaan package in R (Rosseel 2012). We used SEM to account for the measurement error of latent variables (Kline 2011). First, we tested the accuracy of our measurement model with a cutoff value of λ > .40 for weak factor loadings. Second, we tested our hypotheses with the full structural model. We opted for the robust maximum likelihood estimation because a QQ-plot showed that the data violated the assumption of multivariate normality and reported robust model fit indices. As model fit indices, we applied the comparative fit index (CFI; ≥ .95), the Tucker–Lewis index (TLI; ≥ .95), the root mean square error of approximation (RMSEA; ≤ .05), and the standardized root mean square residual (SRMR; ≤ .05) indices (Byrne 2016). Error terms of identical items of Waves 1 and 2 were allowed to covary.

Results

Measurement Model

Prior to hypotheses testing, we conducted a CFA of our measurement model. The model fit indices indicated good model fit, χ² (431) = 540.04, p < .001, CFI = .99, TLI = .98, RMSEA = .03, SRMR = .03. Factor loadings ranged from λ = .61 to .92 (see Supplemental Online Appendix B). All items were used for the final structural model based on these results.

In addition, we tested measurement invariance and compared the configural model (Model 1), a model with constrained factor loadings (Model 2), and constrained intercepts (Model 3). We did not find significant differences between Models 1 and 2 (Δ χ² (6) = 5.26, p = .511) and Models 2 and 3 (Δ χ² (6) = 3.42, p = .754). Hence, we established scalar and metric measurement invariance.

Hypotheses Testing

In the full structural model, we tested our hypotheses over two time points while controlling for autoregressive relationships between the dependent variables. Model fit indices of the full structural model indicated a good fit, χ² (617) = 857.83, p < .001, CFI = .97, TLI = .96, RMSEA = .03, SRMR = .05. The model explained 34% of the variance in perceived manipulative intent, 36% of the variance in benefits, and 31% of the variance in harms in Wave 2.

First, we examined the relationships between perceived fit and misfit of TPA and individual-level implications, that is, perceived manipulative intent. Our findings revealed that the perceived fit at Time 1 significantly predicted the perceived manipulative intent at Time 2 (b = −0.15, SE = 0.06, β = −.14, p = .012), confirming hypothesis 1. Perceived misfit at Time 1 did not significantly impact perceived manipulative intent at Time 2 (b = 0.11, SE = 0.07, β = .09, p = .133). Therefore, hypothesis 2 was rejected.

Second, we tested the relationships between perceived fit and misfit of TPA and perceived benefits and harms for democracy. We found a significant positive relationship between the perceived fit at Time 1 and perceived benefits for democracy at Time 2 (b = 0.14, SE = 0.06, β = .16, p = .029), which confirms our hypothesis 3(a). However, against our assumptions in hypothesis 3(b), the perceived fit of TPA at Time 1 did not significantly decrease perceived harms for democracy at Time 2 (b = 0.01, SE = 0.06, β = .01, p = .842). Our results revealed no significant relationships between the perceived misfit at Time 1 and individuals’ perceptions about the benefits (b = −0.04, SE = 0.06, β = −.04, p = .449) or harms (b = 0.10, SE = 0.07, β = .09, p = .147) of TPA for democracy at Time 2. Hence, we rejected hypotheses 4(a) and 4(b).

In addition, we found that some control variables revealed significant relationships. Individuals who reported higher privacy concerns at Time 1 also indicated stronger perceived manipulative intent of TPA at Time 2 (b = 0.16, SE = 0.07, β = .13, p = .023).

Regarding perceived benefits for democracy, our results indicated that older participants perceived fewer benefits of TPA for democracy at Time 2 (b = −0.01, SE = 0.01, β = −.12, p = .014). Further, OBA knowledge negatively predicted perceived benefits (b = −0.14, SE = 0.04, β = −.16, p = .001). In terms of perceived harms for democracy, we found that individuals with higher political distrust at Time 1 perceived more harm of TPA for democracy at Time 2 (b = 0.26, SE = 0.07, β = .24, p < .001). The SEM model is presented in Figure 1 and findings are shown in Table 1.

Figure 1. Model showing the relationships between the perceived fit and misfit of targeted political advertising (TPA) on perceived manipulative intent and perceived benefits and harms for democracy. N = 428. The paths were estimated applying robust maximum likelihood estimation. Robust fit indices are reported: χ² (617) = 857.83, p < .001, CFI = .97, TLI = .96, RMSEA = .03, SRMR = .05. Political interest, political distrust, political ideology, age, gender, education, online behavioral advertising (OBA) knowledge, and privacy concerns served as control variables. Dotted lines show autoregressive relationships. W1 = Wave 1; W2 = Wave 2. *p < .05; **p < .01; ***p < .001.

In an additional analysis, we tested whether reciprocal relationships existed within our model. Our analysis revealed that perceived benefits at Time 1 significantly predicted perceived fit of TPA at Time 2 (b = 0.13, SE = 0.06, β = .12, p = .041). No other reciprocal relationships between our main independent and dependent variables were found. The full results of our additional analysis are shown in Supplemental Online Appendix C, Table S2.

Discussion

The present study investigated the relationship between individuals’ perceptions about the fit and misfit of TPA and their perceptions of the manipulative intent, benefits, and harms of TPA for democracy throughout an election campaign.

Our findings suggest that TPA can be a successful advertising strategy for political campaigners when individuals perceive an ad to be fitting, that is, when there is a match between the advertising message and party preferences and interests. While the bigger picture indicates no negative consequences for political parties, our findings on the individual level suggest more nuanced consequences. In line with our expectation, the perceived fit of TPA at Time 1 reduced individuals’ perceptions about the manipulative intent at Time 2. This result has two implications. On one hand, it shows that individuals might not perceive TPA as something negative when it fits. This resonates with insights from a qualitative study, highlighting that people expect targeting to happen on social media as part of the social contract with the platform (Van den Broeck, Poels, and Walrave 2020). Hence, one could argue that individuals who perceive that TPA fits might not question the manipulative intent behind an ad, because what they receive might satisfy their expectations. On the other hand, lower levels of perceived manipulative intent might also imply reduced critical elaborations of the advertising or the strategy. Critical reflections, however, are necessary to evaluate the manipulative intent of advertising (Campbell 1995), which constitutes an important dimension of PKM (Friestad and Wright 1994). Without the activation of critical thinking processes, which can make the manipulative intent of a strategy overt, people might not be able to activate their defense mechanisms. Such defense mechanisms, however, are important for individuals to protect themselves against unwanted messages (Fransen, Smit, and Verlegh 2015).

On a more societal level, we investigated the perceived democratic implications of TPA for democracy. Our findings show a relationship between the perceived fit at Time 1 and perceived benefits for democracy at Time 2. Two explanations might apply. On one hand, our findings are in line with the assumptions of SIT (Tajfel and Turner 1986). Our results indicate that when TPA comes from one’s ingroup, in other words, the preferred party, individuals strive to keep a positive image of their party’s actions, further protecting their partisanship identity (e.g., West and Iyengar 2020). On the other hand, individuals might perceive fitting TPA as positive and beneficial. Scholars have argued that possible benefits of TPA for democracy could be related to more relevant advertising or more mobilization of the electorate (Zuiderveen Borgesius et al. 2018). In our study, however, we did not account for these individual-level benefits of TPA. Hence, future research should investigate these relationships and include aspects such as personal usefulness. In this regard, it would also be of interest to investigate how fit and misfit perceptions are related to the privacy calculus, that is, benefits associated with more self-disclosure and privacy risks associated with less self-disclosure (see Bol et al. 2018). Based on the rationale of the privacy calculus model, users engage in trade-off reflections, weighing up potential benefits and risks connected to disclosing personal information online (Trepte et al. 2017; see also Laufer and Wolfe 1977). Individuals’ willingness to disclose their personal information online might have important implications on the preciseness or fit of algorithm-based advertising formats.

Our additional analysis of reciprocal relationships provides insights into possible positive spiraling mechanisms. The more individuals perceived that TPA is beneficial for democracy, the more they perceived the fit of TPA at a later point in the election campaign. However, panel surveys employing more waves are necessary to provide stronger evidence for such a spiraling mechanism.

Contrary to the perceived fit, we did not find any significant relationships between the perceived misfit and individual-level or societal-level implications. Previous research on algorithmic disillusionment helps put our null findings into perspective. Algorithmic disillusionment describes how individuals believe that algorithms in reality might not be as capable and sophisticated as they expect them to be (Felzmann et al. 2019). Such disillusionment might explain why the misfit did not yield any consequences. In this sense, when individuals experience a misfit, they might become more aware of the inaccuracy of targeting strategies and might not perceive them as manipulative or harmful. Future research would need to empirically investigate these processes in depth.

Our null findings regarding the perceived misfit highlight the complexity of this concept. While we expected overall negative consequences of the misfit of TPA on individuals’ perceptions, previous research has also pointed to possible benefits of misfit. Endres (2020) showed that an ideological misfit (i.e., when Republicans targeted Democrats in the 2012 U.S. presidential election) led to stronger mobilization among Democrats when the issue was congruent. At the same time, processes such as linked fate could influence the perceived misfit of TPA (see Hersh and Schaffner 2013). In other words, when individuals receive content targeted to a group they do not belong to, they might still “feel an affinity to [the] group that they are not a part of” (Hersh and Schaffner 2013, p. 527). When such a linked fate is established, the misfit might not have the expected negative effects.

Future research should also further look into different forms of fit and misfit. Our study empirically showed that the political and issue facets formed the concept of (mis)fit. The interrelatedness between political and thematic aspects can be explained based on the issue ownership theory (Bélanger and Meguid 2008). However, as Endres (2020) showed, issue-based targeting could override party preferences. Nevertheless, the findings of Endres (2020) are based on the U.S. context. Therefore, research in the European context is warranted due to different party structures and weaker lines of partisanship (e.g., Huddy and Bankert 2017; Zuiderveen Borgesius et al. 2018). In addition, to develop an in-depth understanding, content analytical evidence is needed to investigate which issues are used for TPA.

Regarding perceived harms, our findings suggest that perceived harms for democracy remain stable and are not influenced by the fit or the misfit of TPA. One possible explanation is rooted in the third-person effect, which describes that individuals perceive the influence of persuasive communication to be stronger for others than for themselves (Davison 1983). In the commercial context, Ham and Nelson (2016) showed that subjective persuasion knowledge—in other words, knowledge about how persuasion functions—was positively associated with third-person perceptions (TPPs). Furthermore, they revealed that harm (in terms of privacy infringement) and benefit (in terms of relevance) assessments of OBA functioned as mediators. More precisely, Ham and Nelson (2016) found a negative relationship between perceived benefits of OBA and TPPs and a positive relationship between perceived harms and TPPs. Similarly, Baum, Meissner, and Krasnova (2021) found not only that individuals thought that TPA would influence others more than themselves but also that the supporters of the opposing party (either Democrats or Republicans) were more influenced by TPA than the supporters of their own party. Similar mechanisms might apply to our context. Hence, evaluations of the harms of TPA for democracy might not depend on what one receives but on the extent to which one thinks such advertising would influence others. However, we did not account for such mechanisms, which opens avenues for future research.

In addition, it is important to note that our conceptualization of perceived harms grasped only an overall abstract assessment. However, many different potential threats of microtargeting practices exist, which are discussed in the literature, such as exclusion or polarization of voters, the spread of misinformation, exposure to siloed information, or potential chilling effects. Hence, we cannot make any assumptions about the relationship between (mis)fit and more concrete perceived harms of TPA, which remains to be investigated in follow-up studies.

While the concepts of perceived fit and misfit give us valuable insights into how individuals’ perceptions about the implications of TPA can be shaped, future research should also investigate other relevant predictors, including individuals’ knowledge (e.g., OBA) or political dispositions (e.g., interest or distrust), which could explain perceptions of individual-level and societal-level implications of TPA.

Limitations

The present study has some limitations. First, we relied on self-reported measures. Memory biases might have influenced individuals’ reported perceptions because some ads might be better memorized than others. Because we cannot know which content individuals saw, it reduces the generalizability of our findings to all targeted political ads. A systematic content analysis would be necessary to shed more light on the actual content to which a person is exposed. Individuals’ memory biases could be addressed methodologically with mobile experience sampling studies. This method allows for measuring individuals’ perceptions immediately after exposure to the content (Otto et al. 2021). Second, longitudinal panel data have advantages over cross-sectional data, as prior levels of attitudes can be taken into account and the temporal order can be established with the different measurement points (e.g., Bartels 2006; Taris 2000). While temporal order is one criterion of causality, we can only hint at directional relationships in this article. Experimental research is needed to investigate the strict causal direction of our established relationships. Third, while our study provides insights into the perceptions of TPA during a state election in Austria, it is necessary to replicate our study in different country contexts and election campaigns. Connected to this, panel studies with more waves would be needed to cover more phases of an election campaign.

Fourth, the measures used in this study come with certain limitations: The concepts of perceived fit and misfit were self-developed. Extensive analysis of the factorial structure of our measures confirmed fit and misfit to be two separate concepts, and reliability proved to be high. To further validate our measures, future research should integrate these measures in different contexts. Because the focus of our study was to evaluate participants’ overall fit and misfit perceptions, we did not ask participants to evaluate one specific targeted political ad regarding a concrete political party or the content. While this might have exceeded the cognitive burden on participants in a panel study, methodological approaches, such as mobile experience sampling methods, could cover this research interest more accurately. Such a method would allow participants to evaluate targeted political ads at the very moment of exposure. This would increase external validity due to the unforced, real-world setting of exposure and strengthen the preciseness of concrete (mis)fit perceptions of TPA.

In this study, we assessed the (mis)fit in terms of political and issue components, which are important aspects of an election campaign. Other components or a combination should also be considered, for example, geographical or identity-based targeting. This also calls for developing a scale that fully reflects the concepts of fit and misfit with all necessary components in the context of TPA.

Regarding our measures of perceived benefits and harms of TPA for democracy, we used self-developed measures that were inspired by work on the democratic implications of TPA (Zuiderveen Borgesius et al. 2018). We accounted for overall perceptions about the benefits and harms of TPA for democracy. Nevertheless, our measures of perceived benefits are more specific (e.g., engagement of voters) compared to more abstract perceived harms (e.g., danger to democracy). These measures do not allow us to compare the two concepts. Therefore, we cannot conclude whether benefits or harms (also in a more specific manner) play a more influential role. This, however, would be an interesting focus for future research, which could also investigate the development of a specific scale of perceived benefits and harms for democracy.

Theoretical and Practical Implications

This study investigated the perceived influence of the fit and misfit of TPA on an individual and societal level. For the first time, we empirically showed that perceived fit of TPA at Time 1 reduced the perceived manipulative intent and increased perceived benefits for democracy at Time 2. On a theoretical level, our findings contribute to our conceptual understanding of individuals’ perceptions of the fit and misfit of TPA as being conceptually different. Second, our findings show that congruency mechanisms stemming from self-congruity theory and schema incongruity processing theory (Mandler 1982; Sirgy 2018) might also apply to the context of political marketing. In addition, we showed that when recipients perceive TPA to be fitting, they are less likely to perceive manipulative intent behind the persuasive strategy. In other words, successfully tailored TPA might hinder the activation of evaluative dimensions of persuasion knowledge (see Friestad and Wright 1994). When the presented TPA was successfully tailored to individuals’ party preferences (i.e., political fit) and issue interests (i.e., issue fit), it was associated with less suspicion about manipulative intent and more positive evaluations of democratic implications.

Our findings have relevant implications for practitioners and voters. On one hand, our findings imply that successful targeting might not be disadvantageous for political parties, especially when communicating to the ingroup, as individuals’ levels of perceived manipulative intent were low when ads were targeted successfully in terms of fit. On the other hand, this has important implications for recipients. Individuals’ critical evaluation of TPA might be limited, especially when fit is perceived as high, which manifests in fewer reflections about the appropriateness of an ad or an advertising strategy (i.e., perceived manipulative intent) and a stronger prevalence of perceived benefits for democracy related to TPA. Losing out on critical evaluations of advertising appeals could also imply that individuals cannot activate their defense mechanisms or coping strategies (see Friestad and Wright 1994). Hence, individuals might be unable to equip themselves against manipulative attempts of invasive advertising strategies, like targeting. Therefore, regulations and regular evaluations of TPA gain importance. Another way to counteract such influences could be initiatives and campaigns enhancing the awareness of new digital campaigning strategies. At the same time, our study indicates that individuals who encounter TPA that fits their party preferences or issue interests perceive TPA to be beneficial to democracy. This again raises questions about individuals’ capabilities of activating defense mechanisms against persuasive techniques aiming to create a fit with individuals’ self-concept. As a result, individuals might potentially underestimate the negative societal consequences of TPA when they are successfully targeted by ads that fit their party preferences and issue interests. This highlights the importance of increasing individuals’ awareness of targeting practices in political campaigning.

Declaration of Conflicting Interests

The authors have no conflicts of interest to disclose.

Additional information

Notes on contributors

Melanie Hirsch

Melanie Hirsch (MSc, University of Vienna) is a doctoral candidate, Department of Communication, University of Vienna.

Marlis Stubenvoll

Marlis Stubenvoll (MS, Aarhus University/University of Amsterdam) is a doctoral candidate, Department of Communication, University of Vienna.

Alice Binder

Alice Binder (PhD, University of Vienna) is a senior scientist, Department of Communication, University of Vienna.

Jörg Matthes

Jörg Matthes (PhD, University of Zurich) is a professor of communication science, Department of Communication, University of Vienna.