Partisan self-interest is an important driver for people's support for the regulation of targeted political advertising.

The rapid emergence of online targeted political advertising has raised concerns over data privacy and what the government's response should be. This paper tested and confirmed the hypothesis that public attitudes toward stricter regulation of online targeted political advertising are partially motivated by partisan self-interest. We conducted an experiment using an online survey of 1549 Americans who identify as either Democrats or Republicans. Our findings show that Democrats and Republicans believe that online targeted political advertising benefits the opposing party. This belief is based on their conviction that their political opponents are more likely to be mobilized by online targeted political advertising than are supporters of their own party. We exogenously manipulated partisan self-interest considerations of a random subset of participants by truthfully informing them that, in the past, online targeted political advertising has benefited Republicans. Our findings show that Republicans informed about this had less favorable attitudes toward regulation than did their uninformed co-partisans. This suggests that Republicans' attitudes regarding stricter regulation are based not solely on concerns about privacy violations, but also, in part, are caused by beliefs about partisan advantage. The results imply that people are willing to accept violations of their privacy if their preferred party benefits from the use of online targeted political advertising.

Introduction

The growing popularity of social media platforms has dramatically changed the landscape of political campaigning [1]. Campaigns now increasingly target specific messages to narrow groups of voters on various digital platforms [2-4] (see S1 Text in S1 File for a discussion). Individual-level voter targeting based on data from state-wide voter registries merged with different kinds of public and commercial data has become a widespread practice since the implementation of the Help America Vote Act of 2002 [3, 5–7]. While early efforts of specific targeting largely relied on door-to-door canvassing, telephone calls, and direct mail, today, platforms like Facebook allow political actors to reach millions of users in a cost-effective way [3, 4, 8]. Thus, spending on online targeted political advertisements has tripled between the 2016 presidential election and the 2020 presidential election in the United States [9].

Observing these astounding levels of targeted political advertising taking place on social media, numerous stakeholders have taken a closer look at these practices and their potentially undesirable consequences for users and society [6, 10–14]. Moreover, private initiatives like ProPublica have raised awareness for online political targeting on Facebook by making political advertisements targeted to specific audiences publicly available [15]. In terms of regulation, there have been several attempts to tighten the election law. For example, in 2017, the Honest Ads Act was proposed, a regulation focused on more rigorous disclosure requirements [16]. Further, some states have introduced legislation that obligates platforms to disclose funding of online targeted political advertisements or to store the advertisements in databases [17]. The Microtargeted Political Ads Act, introduced in 2020, is even more extensive. It would ban online targeted political advertisements based on personal data on online platforms on the federal level [18]. As a reaction to these developments, Twitter, Google, and Facebook have already significantly altered their policies regarding targeted political advertisements [19-22].

A central concern raised by supporters of stricter regulation is a lack of protection and transparency regarding the use of personal data [10–12, 23, 24]. Regulation proponents argue that the sheer amount of available data, advanced predictive modeling, and increasingly sophisticated personalization techniques require new regulatory responses [12]. These demands are aligned with previous research that has established that people value the privacy of their data [25, 26] and that privacy concerns are an important factor in determining attitudes toward the regulation of targeted advertising in general [27] (see S2 Text in S1 File for a discussion). In particular, people seem especially worried about the use of their private data by political actors [28]. Thus, according to public opinion polling, most Americans oppose online targeted political advertisements and consider the use of personal data for online targeted political advertisements to be unacceptable [29, 30].

In this paper, we argue that attitudes toward the regulation of online targeted political advertising are driven not only by concerns over private data misuse. Online targeted political advertising has the potential to influence voting decisions and, as a result, elections [31, 32]. This has consequences for broader societal outcomes, affecting far more than individual data protection. We posit that people take these consequences into account when forming preferences regarding the regulation of online targeted political advertisements. Specifically, we propose that people’s preferences regarding the regulation are aligned with their partisan self-interest considerations that reflect one’s desire to ensure advantages for one’s preferred party and its agenda. Thus, if individuals believe that the opposing party may experience an advantage from targeted political advertisement, they could be motivated by partisan self-interest to support stricter regulation. Indeed, research on public attitudes about other aspects of the electoral process indicates that partisan self-interest is an important factor in people’s opinions [33, 34]. Attitudes on gerrymandering, voter ID laws, or same-day voter registration seem to be driven by partisan self-interest [34-36]. Hence, this study seeks to explore whether, in addition to privacy concerns, partisan self-interest is also an important determinant of people’s attitudes towards stricter regulation of online targeted political advertising.

To answer this research question, we ran an online survey experiment in the United States, working with a sample of Republican and Democratic participants. First, we established that people regard online targeted political advertisements as opposed to their partisan self-interest. We then manipulated partisan self-interest in our experiment to determine whether there is a causal link between respondents’ concerns for partisan advantage and their support for regulation.

Perceptions about partisan self-interest

It is not immediately obvious whether people perceive online targeted political advertisements to be in line with or opposed to their partisan self-interest. Political parties use targeted political advertisements mainly to mobilize their own partisans who are likely to turnout to vote [3, 37]. Therefore, people’s perceptions of whether this advertising benefits or harms their party depend on whether they believe that voters of their party versus voters of the opposing party are mobilized more strongly. In other words, to assess whether the advertisements are to their party’s advantage or not, Democrats must guess how Republicans react to mobilizing messages and vice versa. However, campaign messages delivered to targeted recipients remain mostly unavailable to others [15, 32]. Given this limited transparency [24, 31], and the paucity of information about the effects of online targeted political advertising on voters [38], it seems likely that people have difficulties arriving at accurate estimates on the actual impact of online targeted political advertising on others. Therefore, it is plausible that they could hold biased or unfounded beliefs about the issue.

A large body of research documents that people generally believe that others are more influenced by undesirable persuasive mass communication than they are themselves—a phenomenon known as the third-person effect [39, 40] (see S3 Text in S1 File for a discussion). The prevalence of the third-person effect has been documented across a variety of contexts, including press coverage [41] rap music [42], television violence [43, 44], direct-to-consumer advertising [45], media influence [46], social media [47], and, more recently, fake news [48]. Importantly, the third-person effect predicts that people not only believe that others are more influenced by undesirable mass communication (perceptual component), but that these people also take action to rectify the consequences of such persuasive messages (behavioral component) [49]. Frequently, this behavior involves the support for censorship of undesirable media content. For example, this can take the form of censorship endorsement or support for specific restrictions with regard to commercial advertisement [50], violent and misogynous rap lyrics [42], television violence [44] or election campaign news [51].

Notably, past studies have shown that the strength of the third-person effect increases with social distance to the “other” [48, 52, 53]. This finding is important since high levels of polarization and mistrust between Democrats and Republicans in the United States suggest that the social distance between partisans is large [54-60]. Hence, the potential presence of the third-person effect, combined with a large social distance between the parties, suggests that both Republicans and Democrats may believe that opposing partisans are influenced by online targeted political advertising to a larger extent than are supporters of their party. Crucially, this means that the opposing party is perceived as experiencing an advantage from mobilizing messages directed at their respective electorate. Hence, we hypothesize that supporters of both parties will believe that online targeted political advertisements exert a greater influence on partisans of the opposing party than on their co-partisans (H1). This means that online targeted political advertisements are perceived to go against the partisan self-interest of both Republicans and Democrats.

In turn, the perception that opposing partisans are more easily influenced by online targeted political advertisement should result in the desire to impose stricter regulations on these advertisements, as this would mitigate the perceived mobilization advantage of the opposing party. Indeed, past literature on the third-person effect has established that people frequently favor stricter regulation if they believe others to be more influenced by different sorts of media messages [44, 50, 51]. As a result, it can be inferred that people who believe voters of the opposing party are more influenced by online targeted political advertisements than are voters of their own party will also support the regulation of this advertising, perceiving it to be in their partisan self-interest. Therefore, we further hypothesize that the magnitude of the perceived difference in the effect of online targeted political advertisements on opposing versus fellow partisans is associated with demand for regulation (H2). Correlating privacy concerns to the demand for regulation will also allow us to explore the importance of privacy concerns.

If the demand for regulation is causally explained by partisan self-interest, then altering partisan self-interest considerations of participants should also change their demand for regulation. To test whether it is partisan self-interest that motivates regulation demand, we therefore exogenously manipulated partisan self-interest by changing beliefs about the effect of online targeted political advertising for a randomly selected sample of participants. Informing participants that co-partisans are more susceptible to online targeted political advertisement than supporters of the opposing party should shift these participants’ support for regulation downward. Hence, we hypothesize that participants who are informed that supporters of their own party are more influenced by online targeted political advertising are less supportive of regulation than uninformed participants (H3).

This study contributes to the growing literature that links partisan self-interest considerations to attitudes towards election laws [33, 61–65], and adds to the understanding of causality in this relationship [34-36]. Further, our findings contribute to the literature on the third-person effect by providing the first evidence of its existence in the context of online targeted political advertisements. Additionally, we contribute to the scarce literature that supports a causal relationship between third person perceptions and behavior [66, 67]. While most studies in this domain only report correlational evidence, our experiment allows us make causal conclusions. Our results reveal the challenges posed by new technological advances in the political domain and the ensuing need for new regulation. We show that some partisans are willing to oppose regulation if they believe that online targeted political advertising benefits their preferred party, even at the expense of concerns about privacy violations. Our findings further show that attitudes toward regulation are partially driven by biased beliefs about the effect of online targeted political advertising on others, since participants from both parties believe that regulation is in their own partisan self-interest.

Structure of the study

This study is composed of a correlational and an experimental part. The correlational part provides evidence that participants believe that supporters of the opposing party are more influenced by online targeted political advertising than are supporters of their own party (H1). This way, we establish that supporters of both parties believe that these advertisements are not in line with their partisan self-interest and yield an advantage for the out-party. Importantly, we also show that beliefs about the effect of online targeted political advertising on supporters of the other party relative to supporters of one’s own party are positively correlated with a stronger demand for regulation (H2). As a consequence, support for stricter regulation is linked not only to concerns about individuals’ privacy, but also to participants’ beliefs about partisan self-interest. In the experimental part of the study, we manipulated partisan self-interest by truthfully informing a randomly selected sample of participants that the Republican party benefited more than the Democratic party from the use of online targeted political advertising in the 2016 presidential election. Thereby, we changed Republicans’ perceptions of partisan self-interest. Republican recipients of this information were less supportive of regulation than were their co-partisans who did not get this information (H3). This means that Republicans are less in favor of a regulation when they learn that this would go against their partisan self-interest. Thereby, we establish a causal link between considerations of partisan self-interest and people’s attitudes toward regulation.

Experimental design

We conducted an incentivized, between-subjects online survey experiment with a sample of adult Americans identifying either as Democrats or Republicans. The pre-registration of the study is available at the AEA RCT Registry AEARCTR-0005296. The study received an Institutional Review Board (IRB) approval from the Norwegian School of Economics. All participants gave informed consent and the data was collected anonymously. S1 Fig in S1 File provides an overview of the structure of the experiment.

There were three parts to this study, which were all completed within a single session. In the first part, we measured participants’ beliefs about the effect of online targeted political advertising on supporters of both the Republican and Democratic parties. In the second part, the experimental manipulation was conducted by informing a random subset of participants about the beneficial effects of online targeted political ads for Republicans. The main dependent variable, participants’ attitudes toward the regulation of online targeted political ads, was measured in the third part. Further, we measured posterior beliefs to check whether the treatment group had different beliefs than the control group, and measured respondents’ demographics along with a number of other control variables. The following describes measurements and procedures in detail.

To ensure that all participants had the same knowledge on the subject, in the first part of the study, participants were asked to read a text about online targeted political advertising that explained its technical aspects and its typical usage. We then asked participants to consider a hypothetical scenario in which both Republicans and Democrats competed in a close electoral race in which they spent equivalent sums on online targeted political advertising. We elicited participants’ beliefs about the extent to which they thought that they personally as well as Republicans and Democrats alike would be influenced by online targeted political advertising, using a five-point scale ranging from “not at all” to “to a very great extent.” This measurement corresponds to previous studies on the third-person effect [48, 53]. The order of the questions about Republicans and Democrats was randomized. This measure was used to establish whether participants thought that online targeted political advertisement was in line with their partisan self-interest or not. To address concerns that participants could potentially want to give negative answers about the opposing side while not necessarily believing that such answers had a basis in fact [68, 69], we emulated the approach of previous studies [70], and asked participants to commit to answering the questions to the best of their knowledge.

In the second part of the survey, we manipulated partisan self-interest by manipulating the treatment group’s beliefs about who benefits from online targeted political advertisements. To do so, participants were randomly placed in either the treatment or the control group. Participants in the treatment group were informed that controlling for the number of ads people saw, online targeted political advertising on Facebook significantly increased voter turnout for the Republicans in the 2016 presidential election, while having no effect on Democrats. With this wording, we ensured that participants did not look to different levels of campaign spending as a possible cause of the ads’ effects. The treatment text was based on results from a working paper that shows that targeted political advertisement on Facebook prior to the 2016 U.S. presidential election increased turnout among Republican, but not among Democratic voters [71].

In the final part of the study, we measured all participants’ attitudes towards regulation of online targeted political advertising on a four item, seven-point Likert scale (1 = strongly disagree to 7 = strongly agree), adapted from [48]. The items were: (i) Targeted political advertising should be banned; (ii) I support legislation that requires targeted online political advertising to be clearly marked as targeted; (iii) More regulation is needed when it comes to targeted online political advertising; and (iv) The government is already doing enough to regulate targeted online political advertising (reverse coded). The order of these responses was randomized. We incentivized honest answers by informing participants that their responses would be sent to the United States Congress in an aggregated and anonymous form [72], stressing that there was no deception in the study.

To determine whether the information treatment succeeded in manipulating beliefs about the effects of online targeted political advertising of participants in the treatment group, all subjects were then asked to make an estimation of the number of interactions (likes, shares, comments) that social media campaigns on Facebook of both Republicans and Democrats received relative to each other prior to the midterm elections in 2018. This enabled us to ascertain whether participants generalized from the treatment information about the 2016 Presidential election and applied it to other elections, and, hence, whether the treatment altered participants’ perceptions. We offered a monetary incentive for participants to answer the question to the best of their knowledge [73]. Participants giving the correct answer received a bonus of $1 [69]. In order to control for the possibility that the intervention influenced not only beliefs about online targeted political advertising’s persuasiveness, but also about other problematic aspects of such advertising, we also measured whether participants thought the advertising was: (i) socially desirable; (ii) harmful to society (reverse coded); (iii) beneficial to cultural values; and (iv) unfavorable to societal norms (reverse coded) on a ten-point scale.

To assess the level of privacy concerns, we presented participants with a four item, seven-point Likert scale (1 = strongly disagree, 7 = strongly agree) questionnaire (adapted from [74]) in which we asked participants whether they were concerned that their data was: (i) collected and stored by third-parties; (ii) shared with third-parties; (iii) used to display targeted advertising to them; and (iv) used for commercial purposes. The order of the items was randomized. We further included a fifth item as an attention check to ensure that participants carefully read the items. In accordance with our pre-analysis plan, participants who failed this attention check and another attention check were not included in the final sample.

We further collected data for political attitudes in terms of political engagement, subjective political knowledge, participants’ level of social and economic conservatism [75], a feelings thermometer towards both the Republican and the Democratic parties [55], and participants’ perceived political efficacy [76]. The demographic control variables included age, gender, ethnicity, education, income, household size, use time on the internet, use of an ad-blocker and social media usage.

Sample characteristics

We collected the data for this survey between January 15, 2020 and January 24, 2020. We collaborated with the survey company Dynata to recruit our participants. For that purpose, we used Dynata’s political panel to recruit Republicans and Democrats, as Dynata collaborates with L2. Therefore, we were able to recruit Democrats and Republicans for whom party affiliation was partially verified by their actual voting behavior and partially derived from other known attributes about the participants. In the study, we asked participants for partisanship to further ensure that only Democrats and Republicans participated. That enabled us to avoid recruiting Independents for our study. In total, we recruited a sample of 1549 American participants with quotas on age, gender, region and party affiliation. The distribution of age, gender and region broadly followed the overall distribution in the general population. In accordance with our pre-analysis plan, the fastest 3% of respondents were removed from the sample to increase data quality. On average, participants were 47.49 (SD = 16.48) years old. Of the sample, 50.55% were female and 25.05% were non-white. The participants were better educated than the overall population of the United States. S1 Table in S1 File provides an overview of the characteristics of our sample.

Among the participants, 777 identified as Republicans and 772 as Democrats. Given the nature of the experimental design, Independents were not included in the study. We randomly assigned the participants to either the treatment group (755 participants: 369 Democrats, 386 Republicans) or the control group (794 participants: 403 Democrats, 391 Republicans). Treatment assignment was balanced taking into consideration observable characteristics and pre-treatment beliefs (S2 Table in S1 File).

Results

This section presents the results of the study. First, we will present evidence supporting the hypothesis that supporters of both parties believe that supporters of the opposing party are influenced to a larger extent by online targeted political advertising than are supporters of their own party (H1). This implies that they believe that the use of online targeted political advertising undermines their partisan self-interest. We will then present correlational results regarding the link between these beliefs, privacy concerns and support for stricter regulation (H2). Last, we will present our findings about the causal role of beliefs about the effects of online targeted political advertising on attitudes towards regulation (H3). The analysis was performed using Stata SE 16.0. The data, full instructions for participants, analysis code and variable coding are available at 10.17605/OSF.IO/QM7DZ.

Beliefs about the differential effect of online targeted political advertising on opposing versus fellow partisans

This section reports results for Hypothesis 1. Fig 1 shows the participants’ beliefs about the extent to which online targeted political advertising influences Republicans and Democrats. We found that Republicans believed that Democrats (μ = 3.20, SD = 1.18) were significantly more influenced than Republicans (μ = 2.83, SD = 1.10, Wilcoxon-signed-rank-test, z = -8.67, p < 0.001, r = 0.41). In contrast, Democrats stated that they believed that Republicans (μ = 3.41, SD = 1.17) were more influenced than were Democrats (μ = 2.94, SD = 1.02, Wilcoxon-signed-rank-test, z = -11.336, p < 0.001, r = 0.31). This result supports our Hypothesis 1 that claimed that Republicans as well as Democrats expressed the belief that supporters of the opposing party are more influenced by online targeted political advertisement than are supporters of their own party. This finding is consistent with previous literature on the third-person effect [39, 40]. The magnitude of this perceived difference in the effect of online targeted political advertisement on opposing party supporters relative to supporters of their own party is not significantly different between Republicans and Democrats (two-sided Welch t-test, t(1540) = 1.61, Cohen’s-d = 0.08, p = 0.11). Believing that voters of the opposing party are influenced to a larger extent by these advertisements than voters of one’s own party, hence being more easily persuaded to vote for their respective parties, indicates that both Republican as well as Democratic think the other party gains more votes by using these ads. This implies that participants perceive such advertising as harmful to their own party, undermining their partisan self-interest. Further, we found that participants believed that online targeted political advertising had a smaller influence on themselves (μ = 2.39, SD = 1.21) than on others. The perceived desirability of these advertisements was slightly lower than medium (μ = 4.66, SD = 2.01, measured on a ten-point scale).

Fig 1

Beliefs about the effect of online targeted political advertising.

Participants’ beliefs about the effect of online targeted political advertising on Democrats and Republicans. Beliefs are measured on a five-point Likert scale (1 = “not at all”, 5 = “to a very great extent”). The bars show 95% confidence intervals. * p < 0.05, ** p < 0.01, *** p < 0.001.

Exploring the size of the belief gap between the perceived effect that online targeted political advertising has on supporters of the other versus one’s own party shows that it is correlated to a number of different attitudes that participants hold (see Table 1). We found that affective and ideological polarization, perceived desirability of the advertising, and high subjective political knowledge are significant predictors of this gap. Participants holding a more negative view of the opposing party as measured on a feelings thermometer (i.e. affective polarization) showed a larger belief gap (β = 0.170, p < 0.001). We also found that the level of conservatism for Republicans and liberalism for Democrats (i.e. ideological polarization) as measured on a scale for social and economic conservatism [75] positively predicted the size of the belief gap (β = 0.112 p < 0.001). Participants who saw the advertising as more socially and culturally desirable reported a significantly smaller gap in beliefs between their own party and the other party (β = -0.149, p < 0.001). Taken together, these results suggest that people’s belief that supporters of the opposing party are more influenced than supporters of their own party by online targeted political advertising is linked to a negative perception of the opposition and a more general dislike of online targeted political advertising. This conclusion accords with previous literature on the third-person effect that suggests that people’s belief about the influence of media messages on others relative to themselves correlates with the social distance to the other and a negative perception of the message [48, 52, 53]. Moreover, participants who self-reported a high level of political knowledge reported a larger gap between their own party and the other party (β = 0.133, p = 0.04).

Table 1

Regression of determinants predicting the size of the difference between the perceived effect of online targeted political advertisement on the other party versus one’s own party.

	Effect on other minus effect on own party
	Coef.	Robust SE	p-value	95% CI
Affective polarization	0.170	0.032	<0.001	0.107, 0.232
Ideological polarization	0.112	0.031	<0.001	0.051, 0.174
Desirability of advertising	-0.149	0.017	<0.001	-0.183, -0.116
High political knowledge	0.133	0.065	0.040	0.006, 0.261
Use of internet in hours	0.003	0.005	0.642	-0.008, 0.132
Use of ad-block	0.018	0.029	0.530	-0.039, 0.076
User of social media	0.000	0.085	0.995	-0.166, 0.167
Attitude towards government regulation	-0.013	0.016	0.424	-0.046, 0.019
External efficacy	-0.002	0.001	0.119	-0.004, 0.001
Politically active	0.078	0.060	0.193	-0.039, 0.197
Constant	0.808	0.197	<0.001	0.422, 1.193
Demographics	Yes
Observations	1464
R2	0.148

Note: The table reports the results for an OLS-regression with the difference between how much participants thought online targeted political advertising influences voters of the other party minus how much they thought it influences voters of their own party as dependent variable. The dependent variable is standardized. Affective and ideological polarization are standardized. User of social media is a dummy for the use of social media, use of ad-block is a dummy for ad-block use. Political engagement is a dummy variable for being politically active within the last year, political knowledge is a dummy for above median knowledge. Demographics include age, gender, income, education, ethnicity, and household size. S3 Table in S1 File provides an overview of all variables in the regression.

The relationship between beliefs about voters’ susceptibility to online targeted political advertisement and support for its regulation

In this section, overall demand for regulation as well as results on Hypothesis 2, namely the association between beliefs about the effect of online targeted political advertisement and support for its regulation are presented.

When analyzing both the treatment and the control group together, participants were slightly in favor of regulation of online targeted political advertisement on average (μ = 4.82, SD = 1.18, Cronbach’s-α = 0.67). Overall, 70% of all participants supported stricter regulation. In the control condition, support for stricter government regulation was higher among participants who identified as Democrats (μ = 5.06, SD = 1.10) compared to Republicans (μ = 4.59, SD = 1.21, two-sided Welch t-test, t(782) = 5.79, Cohen’s-d = 0.41, p < 0.001). We further found that, on average, participants in both conditions were concerned about the use of their personal data in online targeted political advertising (μ = 5.63, SD = 1.25, Cronbach’s-α = 0.90). This concern was not significantly different (two sided Welch t-test, t(1529) = 0.10, Cohen’s-d = 0.05, p = 0.31) between Democrats (μ = 5.67, SD = 1.26) and Republicans (μ = 5.60, SD = 1.25). S10 and S11 Figs in S1 File show the distributions of support for regulation and privacy concerns.

To investigate the relationship between participants’ beliefs about the influence of online targeted political advertisement and support for its regulation an OLS-regression was performed, analyzing participants in the control group only (see Table 2). This was done in order shed light on the association between these variables without incorporating the effect of the treatment manipulation. Our results show that among control group participants, beliefs about the influence of online targeted political advertisement on voters of the other party relative to its perceived influence on voters of their own party is a significant predictor of support for stricter regulation of such advertisement (belief other party—own party, β = 0.124, p < 0.001). This confirms Hypothesis 2, which claimed that the more people think that members of the out-party are more susceptible to online targeted political advertising than members of their in-party, the more they are in favor of regulating such advertisement. Furthermore, privacy concerns were significant predictors of participants’ support for regulation in the control condition (β = 0.257, p < 0.001). We find no significant link between participants’ beliefs about the effect that online targeted political advertising has on themselves and their support for stricter regulation (belief about the effect on self, β = 0.052, p = 0.19).

Table 2

Regression of determinants predicting the willingness to support stricter regulation of online targeted political advertising, control group.

	Support for regulation
	Coef.	Robust SE	p-value	95% CI
Belief other party—own party	0.124	0.035	<0.001	0.055, 0.193
Belief about the effect on oneself	0.052	0.039	0.187	-0.025, 0.129
Privacy concerns	0.257	0.045	<0.001	0.169, 0.344
Observations	754
R2	0.125
Demographics	Yes
Social Media use	Yes
Political Engagement	Yes

Note: The regression only includes participants in the control group who answered all questions of the survey. The table reports results from an OLS-regression with respondents’ support for stricter regulation of online targeted political advertisement as the dependent variable. Belief other party—own party is defined as the difference in participants’ beliefs about the effect that online targeted political advertising has on the other party minus its effect on the own party. Belief about self is people’s belief about the effect that online targeted political advertising has on themselves. Privacy concerns are measured on a seven-point Likert scale (1 = strongly disagree, 7 = strongly agree). All variables were standardized. Demographic information includes age, education, income, household size, gender, and ethnicity. Social media use includes whether the participant uses social media, the time they spent online in general (in hours), and and the use of an ad-blocker. Political engagement measures include a variable for being politically active within the last year, external political efficacy, political knowledge, and attitudes towards government regulation in general. S11 Table in S1 File provides an overview of all variables in the regression.

To further investigate our findings, we also ran an OLS-regression using participants’ beliefs about the effect of online targeted political advertising on the opposing party and their beliefs about the effect on their own party as individual independent variables. The more participants thought that members of the opposing party would be influenced by online targeted political advertisement, the more they supported stricter regulation of such ads (β = 0.169, p < 0.001). Their belief about the effect of these ads on voters of their own party was negatively correlated to support for regulation, but not significantly (β = -0.043, p = 0.26) (see S4 Table in S1 File).

The causal effect of beliefs about voters’ susceptibility to online targeted political advertisement and support for its regulation

This section reports the experimental results, which were predicted by Hypothesis 3. In the treatment condition, we manipulated partisan self-interest by informing a randomly selected subgroup of Republicans and Democrats that the Republican party benefited more from the use of online targeted political advertising in the 2016 presidential election than did Democrats. To determine whether this information shifted respondents’ support for stricter regulation of online targeted political advertising, we compared levels of support for regulation between Democrats and Republicans in the treatment and the control group. With Republicans, we found significantly lower support for stricter regulation of online targeted political advertising in the treatment than in the control group (two-sided Welch t-test, t(776) = 2.08, Cohen’s d = 0.15, p = 0.04). This confirms Hypothesis 3, showing that Republicans who learn about the advantageous effects of online targeted political advertising for their party are less in favor of regulation than their uninformed co-partisans. This means that Republicans are less in favor of a regulation that goes against their partisan self-interest. These effects remained qualitatively the same when examining only participants who wanted their opinions to be considered by Congress (98.7% of the sample) and participants who expressed trust in the information that they had received about the effect of online targeted political advertisement (85.7% of the treatment group), although in the latter case, the effect became insignificant for Republicans (S5 and S6 Tables in S1 File).

For Democrats, we found no difference in levels of support for regulation between the treatment and the control group (two-sided Welch t-test, t(759) = -0.55, Cohen’s d = 0.04, p = 0.58). This result is in accordance with the finding that Democrats believed Republicans are more influenced by political online advertising than members of their own party. Therefore, the information we gave them corresponded with their pre-existing beliefs, and should not alter their regulation demand. S15 and S16 Figs in S1 File show the distribution of answers for Democrats and Republicans in the treatment and the control groups.

To make sure the shift in demand for regulation resulted from a shift in beliefs about the extent to which online targeted political advertisement influenced voters of each party, we tested whether the treatment group had different beliefs regarding Republicans’ susceptibility to online targeted political advertisements. This was done by measuring beliefs about the effect of online targeted political advertising on Republicans and Democrats a second time after the treatment information, this time in the form of beliefs over interactions on social media posts. Fig 2 shows the effect that the treatment information had on beliefs about social media interactions in the 2018 midterm election. We found that in this incentivized question, Republicans and Democrats in the control condition reported beliefs that were qualitatively similar to the first measure of their stated beliefs. Uninformed Republicans believed that Democrats received more interactions in the run-up to the 2018 midterm elections while uninformed Democrats believed that Republicans received more interactions. Responses to this question and to the pre-treatment measure of the effects of online targeted political advertising on Republicans and Democrats are well correlated (r = 0.24, p < 0.001, see S12 Fig in S1 File). However, results differed for respondents in the treatment condition: Republicans who received the treatment information about online targeted political advertisements helping them in the 2016 Presidential election reported that they believed that Republicans received more interactions in 2018. This result represents a significant divergence in beliefs between informed and uninformed Republicans that corresponds to the information that they received, χ2(13.04, N = 771), p = 0.04. For Democrats, though, no shift in their beliefs about the 2018 midterm elections was detected, χ2(6, N = 764), p = 0.65. This further corroborates the finding that the treatment information altered partisan self-interest considerations and caused Republicans to demand less regulation due to a shift in beliefs about the advantageous effect of online targeted political advertisement for their party. Moreover, this result shows that the treatment information reinforced Democrats’ pre-existing beliefs about Republicans being more susceptible to online targeted political advertisement. Therefore, their regulation demand is not altered by our treatment.

Fig 2

Beliefs about social media engagement in the 2018 midterm elections.

Participants’ beliefs about the ratio of interactions in the run-up to the 2018 Midterm election. This was measured on a scale that ran from “Democrats three times more than Republicans” to “Republicans three times more than Democrats” with “Equal” as the mid-point.

To preclude the possibility that the information about the effect of online targeted political advertising changed participants’ perception of how desirable such advertising is or participants’ privacy concerns, we tested for significant differences in these measures between the treatment and the control group. We found that, in general, participants viewed the desirability of using online targeted political advertising as slightly lower than medium (μ = 4.66, SD = 2.01). Comparing the ratings of the desirability of online targeted political advertising for Republicans in the treatment (μ = 4.75, SD = 2.00) and the control group (μ = 4.85, SD = 2.00), we found no statistically significant difference (two-sided Welch t-test, t(769) = 0.69, Cohen’s d = 0.05, p = 0.49). The same result was found for Democrats in the treatment (μ = 4.42, SD = 2.03) and the control group (μ = 4.61,SD = 2.00 two-sided Welch t-test, t(755) = 1.28, Cohen’s d = 0.09, p = 0.20). We also found no significant differences in privacy concerns between the treatment and the control groups, for both Democrats (two-sided Welch t-test, t(759) = 0.10, Cohen’s d = 0.05, p = 0.46) and Republicans (two-sided Welch t-test, t(768) = -0.70, Cohen’s d = -0.05, p = 0.49). Hence, the treatment manipulation altered Republicans perceptions about how much other voters are influenced by online targeted political advertising, while neither affecting perceived desirability nor privacy concerns.

Exploratory data analysis reveals that the effect of the information on Republicans was heterogeneous between different levels of conservatism (see S17 and S18 Figs in S1 File). We found that for those Republicans scoring below the median in social and economic conservatism, the information that their party benefited from the use of online targeted political advertisement did not significantly change their support for regulation compared to the same group who did not receive this information (two-sided Welch t-test, t(403) = 0.16, Cohen’s d = 0.02, p = 0.87). The support for stricter regulation of online targeted political advertisement among Republicans scoring at or above the median in economic and social conservatism differed significantly between the treatment and the control groups (two-sided Welch t-test, t(373) = 2.59, Cohen’s d = 0.27, p = 0.01). This effect cannot be attributed to initial differences in the support for regulation in the baseline condition between above median and below median conservative Republicans (two-sided Welch t-test, t(347) = 0.47, Cohen’s d = 0.05, p = 0.64).

Discussion

The present study reveals that both Republican and Democratic participants in our sample believe that supporters of the opposing party are influenced by online targeted political advertising to a greater extent than are supporters of their own party. In the context of our study, this implies that both Democrats and Republicans assume that the opposing party benefits more from such advertisement, as they think their members are more easily persuaded by being exposed to online targeted political advertisement. Hence, both Democrats and Republicans regard online targeted political advertisement as disadvantaging their partisan self-interest. We further found that both this belief and people’s concern over data privacy significantly predict people’s support for policies limiting the use of such ads. This indicates that people are in favor of regulation since they seek to protect their private data, but also due to an urge to mitigate the perceived advantageous effect of the advertisements for the opposing party. These results therefore provide evidence that the support for stricter regulation of online targeted political advertising is partially motivated by partisan self-interest. Causal evidence that substantiates these findings was provided in the experimental part of the study. The results of the experimental manipulation of partisan self-interest show that Republicans who were informed about the beneficial effects of online targeted political advertisement for their own party reported lower support for regulation than did Republicans in the control group. Therefore, we are able to show that the perception bias is causally linked to Republicans’ support for stricter government regulation. This suggests that participants might make a trade-off in favor of partisan self-interest and contrary to concerns about the violation of data privacy. We found that this effect is not present with all Republican participants, but is concentrated among those with the highest levels of conservatism. This finding concords with the idea that people trade-off personal costs, such as privacy concerns, against partisan self-interest. As more conservative Republicans gain more from an electoral advantage for their party, they are more willing to accept violations of privacy if these violations provide their preferred party with a benefit in an election.

These results contribute to the findings of previous research examining motivations behind attitudes toward election laws. While multiple studies provide evidence that partisan self-interest considerations are associated with attitudes towards the election law, many of these studies are cross-sectional and, hence, do not establish causality [33, 61–64, 77]. Several recent studies use experimental designs to measure causal effects of partisan self-interest on attitudes toward the election law [34-36]. We follow this line of research to provide more experimental evidence of the causal relationship between partisans’ self-interest considerations and their opinions towards the election law. Our findings support the idea that the broader public favors regulation based on their partisan self-interest, and supports laws that contribute to the electoral success of their preferred party.

Our findings further add to an emerging body of literature that shows that some people are willing to make trade-offs between established democratic norms and partisan self-interest [78-81]. According to our results, participants holding the strongest policy views have the greatest reaction to the information that online targeted political advertising benefits their party. This finding accords with previous findings that people are willing to accept the undermining of democratic or moral principles if it benefits their political goals. For example, previous research has documented that partisans are less willing to take corrective measures against politically biased messages if they benefit their party [82].

Similarly, in our case, people’s attitudes towards the regulation of online targeted political advertising are partially driven by the desire to set rules that benefit people’s preferred party, even if they view online targeted political advertising as harmful to societal norms. This behavior might be perceived as a threat to perceptions of the fairness of elections, which could then undermine peoples’ support for a electoral system that relies on a shared understanding of democratic norms [83-88]. We show that the rise of new technologies could potentially contribute to perceptions of “democratic backsliding” [78], as people might be willing to use the newly-required technologies to pursue their partisan self-interest. We further show that beliefs about the impact that new technologies have on the electoral process are crucial to our understanding of public attitudes towards them. This finding contributes to a wider body of literature that investigates how potentially erroneous beliefs that people hold drive their behavior [89-95].

This study reveals that it is difficult to understand public preferences for certain policy measures without understanding the beliefs that people hold about key variables that are affected by these policies. Preferences for regulation of online targeted political advertising are currently driven in part by third-person perceptions, leading to biased beliefs about their effect. This situation could lead to potentially sub-optimal policy decisions, as politicians might follow public preferences that are driven by biased beliefs. Our findings underscore the necessity of providing the public with truthful information about the effect of online targeted political advertising. We show that support for stricter regulation among Republicans would be significantly lower if they were correctly informed about the effect that it had on the 2016 Presidential election, because they underestimate the positive effect that online targeted political advertising might have had or will have on their own party.

Previous research on the third-person effect found evidence for a gap between the perceived effect of persuasive mass communication on the self and on others [39, 40]. Furthermore, correlational research supports the hypothesis that this gap motivates people in performing mitigating actions against the negative consequences of such persuasive communication [49]. Our study adds to this literature in three ways. First, this study is the first to show that a perceptual gap exists in the context of targeted online political advertising. Second, this study establishes a causal link between the perceptual gap described by the third-person effect and a behavioral measure for support for government regulation. Thereby, we add to the scarce previous studies that show a causal relationship between third person perceptions and behavior [66, 67]. By manipulating the perception gap of Republicans in our information treatment downward, and by showing that this decreases their support of the mitigating action, we were able to show causality between perception and behavior. Third, our results also add to previous studies reporting that the third-person perception increases with social distance, or between in-groups and out-groups [48, 52, 53]. To the best of our knowledge, this is the first study to show that the gap between Democrats and Republicans in their perceptions of the influence of undesirable mass communication is strongly linked to affective as well as ideological polarization, and it is the first study to measure this outcome with an unincentivized and an incentivized measure.

Our results have several limitations. We were unable to show similar causal results for Democratic supporters. We found a strong correlation between the beliefs that Democrats report about the effect that online targeted political advertising has on Republicans and their support for stricter government regulation, but cannot claim causality for this group. Due to our incentivization of our outcome measure, we needed to truthfully inform participants that we were not using deception in this study, and we were therefore unable to manipulate Democrats’ beliefs in a way that was equivalent to that used with Republicans. At the time of designing the study, no scientific evidence was available to support the claim that Democrats benefited more from online targeted political advertisement in some election. Future research should address this shortcoming and examine possible partisan self-interest motives of Democrats in this context more closely. We have not used Independents as a control group due to the mixed political leanings of Independents. Only a small percentage of Independents do not favor either the Democrats or the Republicans [96]. Hence, their attitudes would likely depend on the composition of the sample. In addition, our treatment manipulation stated that online targeted political advertisements “significantly increased the number of votes for the Republican party, but not for the Democratic party. Hence, online targeted political ads influenced Republican voters, but did not influence Democratic voters.” This information is based on results from a study on the effect of online targeted political advertisements on Facebook during the 2016 U.S. presidential campaign on voter turnout and candidate choice [71]. These results were the only results about the influence of online targeted political advertisements on voters available to us at the time of designing the experiment. While the results of that research show that overall, these ads did not impact voting behavior of Democrats, the phrasing that Democrats were not influenced at all, as stated in our treatment, might not resemble actual beliefs of Democrats or Republicans. Hence, while we are able to show that partisan self-interest considerations motivated a shift in regulation demand for Republicans, the size of that effect might be smaller outside of this study’s setting. After our survey had been in the field, the authors of the 2018 working paper published a newer version, in which some of the earlier results were revised. In the new version, the authors conclude that online targeted political advertisements increased turnout for the Republican party, and decreased turnout for the Democratic party [97]. Our treatment text does not incorporate these latest results. Further, the main measure of interest, participants’ support for stricter government regulation, indicates relatively low-scale reliability (Cronbach’s α = 0.67). However, exploratory results show that a reduced scale (excluding the fourth item) has higher reliability (Cronbach’s α = 0.75) and that all of our main results are robust to the reduced scale (see S7 and S8 Tables in S1 File). Moreover, to address concerns that participants might be principally opposed to the idea of banning targeted advertising, we excluded the first item that asks for support for a ban. The results of this analysis are available in S9 and S10 Tables in S1 File. These findings replicate the effect we report in the results section of this paper. Another limitation is that even though participants read an explanatory text about the functionalities and use cases of online targeted political advertising in the beginning of the survey, we could not rule out the possibility that participants had diverging levels of knowledge of the topic. While this does not limit the interpretation of our treatment manipulation, it could be an important predictor of participants’ estimates of the influence of online targeted political advertisements on others. Future studies should address this by measuring participants’ knowledge after they have read an explanatory text.

This paper develops a new experimental paradigm to study people’s attitudes towards technological change which has an influence on elections. We show that support for or opposition to the regulation of new technology that has implications for the political process is driven by potentially biased beliefs about how the use of this technology affects political outcomes for one’s preferred party. Therefore, our findings add to a growing policy debate and underscore the necessity of making the effects of online targeted political advertising transparent and of truthfully informing the public about the effects of the new technology so that the public can fully and knowledgeably realize their true attitudes. We believe that more research is necessary to fully understand the public’s attitude towards these innovations, especially regarding beliefs about the spread and effect of false information and divisive messages. Further, our result indicating that people consider the broader societal effects of targeted political advertising might have implications for certain aspects of targeted commercial advertising. We would encourage future research to investigate whether similar mechanisms would motivate people to oppose, for example, the use of targeted advertising to promote socially undesirable consumption, such as smoking, drinking or other unhealthy behavior.

(PDF)

Click here for additional data file.

16 Dec 2020

PONE-D-20-33004

Partisan self-interest is an important driver for people's support for the regulation of targeted political advertising

PLOS ONE

Dear Dr. Baum,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

Both reviewers and I see potential in your manuscript, but the reviewers have raised several concerns. I will not reiterate the points, as the reviewers have provided detailed comments. Please address all reviewer comments in a revision.

Please submit your revised manuscript by Jan 30 2021 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

We look forward to receiving your revised manuscript.

Kind regards,

Natalie J. Shook

Academic Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. During our internal checks, the in-house editorial staff noted that you conducted research or obtained samples in another country. Please check the relevant national regulations and laws applying to foreign researchers and state whether you obtained the required permits and approvals. Please address this in your ethics statement in both the manuscript and submission information.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Partly

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: I Don't Know

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: Authors present a timely and important study, and show that partisan self-interest is an important factor in support for regulation of political targeting. I want to applaud the transparency efforts of the authors. I am also impressed by the authors’ rigor. However, I do have some major concerns that I will now list below.

1. I was surprised the authors took the US as a case to study attitudes toward regulation of targeted political advertising. I understand that the US is a very polarized nation, so in that sense the US is a natural case. However, online political advertising is political speech and thus protected by the first amendment. As such, any effort to regulate targeted online advertising will be challenging at best. Looking at the four items that together measure attitude toward regulation, especially the first item: “Targeted political online advertising should be banned” seems to be measuring non-attitudes. I would invite the authors to construct a new scale with only items 2, 3, and 4 and subsequently run the relevant analyses once more. This scale will better capture US citizens’ attitudes toward regulation.

2. I think that knowledge about political targeting could be an important moderator for partisans’ estimates of the influence of targeted advertising on others. Why did the authors not include this variable and would the authors agree that this omission limits the interpretability of the findings?

3. I am concerned about the phrasing of the treatment. Authors phrased the treatment as follows: “[…] The researchers found that the targeted political ads significantly increased the number of votes for the Republican party, but not for the Democratic party. Hence, targeted political ads influenced Republican voters, but did not influence Democratic voters.”

First, this treatment text is not ‘truthfully’ informing participants about the effects in 2016. Essentially this text states that zero democratic voters were influenced? This can’t be the case. A small consequence Is that participants were, in fact, deceived.

Second, and more importantly, why phrase the treatment in such an extreme, zero-sum, way? Why not state that targeting was thought to have benefitted republicans more than democrats? This treatment now is very artificial, unrealistic even. Which is a shame because there was a perfectly natural way to do it. As a result I wonder what the experimental effect really means? Of course republicans are not going to oppose something that gives them, and only them, an advantage. Such an advantage is worth setting aside, say, privacy concerns because it gives them back electoral victory and thus important policies. But in reality, the republican advantage, in 2016, was much more modest than described in the treatment. In a more realistic scenario republican partisan self-interest might have played a (much) smaller role.

4. Would it not be better to include independents as a control group? Or to compare the scores to a ‘balanced condition’, where both democrats as well as republicans benefit about equally from targeting?

5. Authors did never explicitly measure partisan self-interest. This makes it difficult to show the hypothesized mechanism. Why did they not measure this key concept?

6. On p. 11 the authors state that perceived desirability of online targeted ads is low. But on a ten-point scale, perceived desirability is 4.66, sd = 2.01. I would not call this low.

I believe this study makes an important contribution to the field of online political advertising research, especially when issues 1 and 3 are addressed.

Reviewer #2: Manuscript PONE-D-20-33004 evaluates if Americans’ support for hypothetical regulations of microtargeting change when subjects are informed that microtargeting benefits Republicans while having no effect on Democrats. Results are not surprising based on the many published studies that find support for various (proposed) election laws shifts when partisans are informed that the proposed law disproportionately helps/hurt their party. In this experiment, the effects were only significant for Republicans – though, this is not clearly represented in the current wording of the abstract. I have a few concerns about the experimental components of the manuscript, as well as the article’s framings. Order is consistent with the manuscript’s layout and does not indicate priority.

Microtargeting in the United States is very different than microtargeting in Europe due to the accessibility of publicly available electronic lists of registered voters. States are required by the Help America Vote Act of 2002 (HAVA) to compile and maintain these lists that include basic information (e.g. voter history, party registration, and race/ethnicity). This information is far more valuable for campaigns than data they can purchase or acquire elsewhere. The authors do not seem to understand the distinctions between microtargeting in the United States and elsewhere. And they fail to cite any scholars of US microtargeting (e.g. Hersch, Hillygus, Endres, Nickerson, Rogers, Panagopoulos).

The study’s motivation feels like a straw man argument. The authors note on page 3, “a heated public debate calling for stricter regulations has accompanied the emergence of such ads.” However, none of these citations reference a “heated public debate” about regulating microtargeting in the United States, and I am not aware of a debate or proposed legislation in any US state or nationally. The merits of microtargeting are more frequently debated outside the US than inside. After all, the United States Congress paved the way for political microtargeting through their HAVA legislation (see Hersh 2015). The authors appear to be conflating debates about political misinformation, dark money in politics, and false statements on social media with political microtargeting. Further, the authors are somewhat cavalier with their references throughout the manuscript.

I do not consider the treatment, as described in the text, to be truthful. Though, the exact treatment language is not quoted in the text. The authors note, “we truthfully informed a randomly selected sample of participants that the Republican party benefited more than the Democratic party from the use of targeted political advertising in the 2016 presidential election” (page 6). I am not aware of any legitimate, peer reviewed publication that support this claim. It is possible, that Republicans benefitted more than the Democrats in terms of vote choice / persuasion, fundraising, (de)mobilization, but one CANNOT simply assume that the party of the winning candidate benefitted more than the party of the losing candidate. Later in the paper the treatment is described slightly differently as, “Participants in the treatment group were informed that controlling for the number of ads people saw, targeted political advertising on Facebook significantly increased voter turnout for the Republicans in the 2016 presidential election, while having no effect on Democrats ” (Page 8). The working paper cited to support this claim does not support it. In fact, it argues that targeted advertising had a negative effect on the turnout of Democrats.

More information is needed about the survey. Was there any quota sampling? If so, which variables were used? What was the time interval between each survey wave? What was the attrition rate? Some aspects/concepts are incorrectly described (e.g. a response scale from "not at all" to "to a very great extent" is NOT a Likert scale). Do you have a citations for the description of L2 as “the largest voting tracking service in the United States.” How did L2 and/or Dynata identify Democrats and Republican? Is this based on self-reports from an earlier survey? Party registration (not all states have party registration)? Votes in primary elections (most registered voters do not vote in primaries)?

The "manipulation check", “all subjects were then asked to make an estimation of the number of interactions (likes, shares, comments) that social media campaigns on Facebook of both Republicans and Democrats received relative to each other prior to the midterm elections in 2018” does not appear to directly evaluate whether the treatment group read and/or processed the information in the treatment.

Finally, all control variables should be collected pre-treatment (See Gerber and Green 2012; Montgomery, Nyhan, Torres 2016; and many others), but it appears many were collected in wave 3 after the treatment was delivered (wave 2).

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

1 Mar 2021

Dear Editor,

Dear Reviewers,

We would like to thank you all for your very constructive and helpful feedback on our submitted manuscript and for your general interest in the topic. We highly appreciate that you have taken the time and effort to provide us with these constructive and insightful suggestions to improve the quality of our manuscript. Based on your suggestions we revised our manuscript and tried to incorporate all of the points raised. As suggested, we especially focused on rewriting the introduction to incorporate more research by scholars of targeted political advertisement and to revise our main argument. Based on your comments, we now also explain in more detail the phrasing of our treatment text and how we manipulated partisan self-interest. Further, we excluded control variables that were measured after the treatment from our analysis. In addition to this, we have also worked on strengthening the theory behind our hypotheses and making our reasoning more evident. Please find our detailed responses to your comments in the following section.

Comments from Reviewer 1

Authors present a timely and important study, and show that partisan self-interest is an important factor in support for regulation of political targeting. I want to applaud the transparency efforts of the authors. I am also impressed by the authors’ rigor. However, I do have some major concerns that I will now list below.

Response: Dear Reviewer, we would like to thank you for the appreciation of our submitted manuscript and thank you once more for taking the time and effort to provide these very constructive and insightful suggestions. We hope that we addressed them appropriately in our revised manuscript.

1. I was surprised the authors took the US as a case to study attitudes toward regulation of targeted political advertising. I understand that the US is a very polarized nation, so in that sense the US is a natural case. However, online political advertising is political speech and thus protected by the first amendment. As such, any effort to regulate targeted online advertising will be challenging at best. Looking at the four items that together measure attitude toward regulation, especially the first item: “Targeted political online advertising should be banned” seems to be measuring non-attitudes. I would invite the authors to construct a new scale with only items 2, 3, and 4 and subsequently run the relevant analyses once more. This scale will better capture US citizens’ attitudes toward regulation.

Response: We are grateful for this perspective. To address this concern, we re-ran our analysis and added the results to the paper. Please find the updated analysis in tables S9 and S10 in the appendix. Here, we find that the main effect persists and seems to get a little bigger with this specification. We still, however, report the full scale in the main body of the paper with reference to our pre-analysis plan. We also hope to follow the more ambitious legislative projects like the bill introduced by Rep. Anna Eshoo (1), which would ban online targeted political advertisement based on personal data on a federal level, in our research design to better understand how Americans think about a more radical approach to regulation.

2. I think that knowledge about political targeting could be an important moderator for partisans’ estimates of the influence of targeted advertising on others. Why did the authors not include this variable and would the authors agree that this omission limits the interpretability of the findings?

Response: Dear Reviewer, thank you very much for pointing this out. We agree that knowledge about political targeting could be an important predictor of participants’ estimates of the influence of online targeted political advertising on others. In order to ensure that all participants had the same level of knowledge about political targeting we provided an explanatory text to participants in the beginning of the survey. After reading the text, participants had to indicate that they had read and understood the text. The text read as follows:

„Please read the following information carefully. Targeted advertising is the practice of monitoring people’s online behavior and using the collected information to show people individually targeted advertisements. Online behavioral data can include web browsing data, search histories, media consumption data (e.g., videos watched), app use data, purchases, click-through responses to ads, and communication content, such as what people post on social networking sites. This online data is often combined with demographic data like age, gender and location. Political parties also use targeted advertising, for example, before presidential elections. Targeted political advertisement involves creating messages targeted at narrow categories of voters based on data analysis gathered from individuals' demographic characteristics and their online behavior. This enables political campaigns to send very specific messages to certain groups of potential voters. These messages are selected to be the most appealing to this group. Political actors use targeted advertising, for example, to reach voters who are likely to vote for them with messages that will influence them.”

With this text we intended to describe the basic mechanism behind targeted online advertisements, the data that is used for it, and potential applications in political campaigns. However, we agree that an additional measure of participants knowledge about targeted political advertisements would strengthen this assumption even more. Generally, the interpretation of our treatment manipulation should not be affected by any remaining differences in knowledge about targeted advertising, since participants were randomly assigned to the treatment or the control group after reading the text. We have added the following text to address this issue to the discussion section (p. 23, l. 601).:

“Another limitation is that even though participants read an explanatory text about the functionalities and use cases of online targeted political advertising in the beginning of the survey, we could not rule out the possibility that participants had diverging levels of knowledge of the topic. While this does not limit the interpretation of our treatment manipulation, it could be an important predictor of participants' estimates of the influence of online targeted political advertising on others. Future studies should address this by measuring participants' knowledge after they have read an explanatory text.“

3. I am concerned about the phrasing of the treatment. Authors phrased the treatment as follows: “[…] The researchers found that the targeted political ads significantly increased the number of votes for the Republican party, but not for the Democratic party. Hence, targeted political ads influenced Republican voters, but did not influence Democratic voters.” First, this treatment text is not ‘truthfully’ informing participants about the effects in 2016. Essentially this text states that zero democratic voters were influenced? This can’t be the case. A small consequence Is that participants were, in fact, deceived. Second, and more importantly, why phrase the treatment in such an extreme, zero-sum, way? Why not state that targeting was thought to have benefitted republicans more than democrats? This treatment now is very artificial, unrealistic even. Which is a shame because there was a perfectly natural way to do it. As a result I wonder what the experimental effect really means? Of course republicans are not going to oppose something that gives them, and only them, an advantage. Such an advantage is worth setting aside, say, privacy concerns because it gives them back electoral victory and thus important policies. But in reality, the republican advantage, in 2016, was much more modest than described in the treatment. In a more realistic scenario republican partisan self-interest might have played a (much) smaller role.

Response: Thank you very much for this very important comment. The information we gave participants in the treatment was based on a working paper by Liberini et al. (2). In this working paper, the authors study the effect of exposure to online targeted political advertisements on Facebook during the 2016 U.S. presidential campaign on voter turnout and candidate choice. In order to do so, the authors exploit variations in prices of advertisements on Facebook to estimate the intensity of political campaigning for different audiences, specified by location, political affiliation and demographics. The idea is that given stable sizes of the audiences, more advertisements targeted to an audience should increase prices due to the demand shift. Hence, higher prices indicate higher campaign intensity. The authors then match this data with responses of 2,414 American voters from the 2016 American National Election Study (ANES) sampled once prior and once after the election regarding their Facebook use, turnout and candidate choice. The authors find that “targeted Facebook campaigning increased turnout among core Republican voters, but not among Democratic or Independent voters” (1: p. 5). They specifically state that “microtargeting was ineffective for Clinton, failing to boost turnout or sway voters in her favor” (1: p. 5). Thus, when designing our experiment, we based our treatment information on this working paper, aiming at changing the beliefs regarding the effect of online targeted political advertisements on opposing partisans of one group in our sample to be able to establish a causal relationship between partisan beliefs and support for regulation. Importantly, our search for scientific evidence of similar results of Democrats having been more influenced by online targeted political advertisements in some other election has not resulted in any meaningful findings. Therefore, as we did not want to deceive participants, we decided to manipulate only Republicans’ beliefs based on the evidence that we had at the time. After our survey had been in the field, a new version of the Liberini et al. (2) paper has been released in April, 2020. In this updated version, the authors conclude that being exposed to targeted online advertisement during the 2016 U.S. presidential campaign increased turnout for Trump supporters, and decreased turnout for Clinton supporters (3). Unfortunately, this information was not available to us when designing the study. We agree that the effect shown with our treatment manipulation might be smaller in the real world. Since our results come from a survey experiment, they are indeed somewhat artificial. Our aim was to examine whether altering participants’ self-interest considerations leads them to change their attitudes accordingly. With our experiment we are able to show that the mechanism of partisan self-interest motivating regulation preferences exists and that learning about the advantageous effects of the advertisements for their party leads Republicans to express lower support for a regulation that would oppose their partisan self-interest. We now discuss this aspect in the discussion section, stressing that partisan self-interest considerations might be smaller in the real world and that results of the 2018 working paper have been updated (p. 22, l. 583).

4. Would it not be better to include independents as a control group? Or to compare the scores to a ‘balanced condition’, where both democrats as well as republicans benefit about equally from targeting?

Response: Thank you very much for this suggestion. We have not included Independents as a control group for two reasons. First, most Independents lean towards a party and are not completely neutral. Only about 7% of the American public did not favor a particular party in 2018 (4). Among Independents, a slightly higher share of 44% favored the Democratic party, while 34% favored the Republican Party in 2018 (4). Therefore, Independents would not yield a neutral control population.

The second, and related, reason why we did not include Independents as a control group lies in the nature of the third-person effect (5). Specifically, our experiment was based on the fact that Democrats believe that Republicans are influenced more by online targeted political advertisements, and Republicans believe the opposite. This is analogous to findings showing that people believe that members of an out-group are more easily persuaded than members of their in-group, or themselves (5,6). The perception that out-group members are more influenced by persuasive communication than in-group members increases with social distance to the “others” (6–8). Due to increasing polarization between the parties, this distance is larger between Democrats and Republicans as compared to the distance between Independents and Democrats, or Independents and Republicans (9,10). Hence, the strength of the third person effect presumably varies between these groups and is stronger between Democrats and Republicans as opposed to Independents and Democrats, or Republicans.

The average belief of Independents about whether Republicans or Democrats are more influenced by online targeted political advertisements should depend on the political leanings of the sample of Independents. In case of a majority of them leaning towards Democrats, their average belief should go in the direction of the belief of Democrats, but be somewhat less strong. If a majority of them would lean towards Republicans, the opposite would be true. Therefore, the effect of the treatment manipulation of informing Independents about Republicans having benefited more would depend on the composition of the sample of Independents.

To answer your second question, in order to include a condition in which participants would be informed that Democrats benefited more from online targeted political advertisement, we would have needed scientific data to support this to avoid participant deception. However, as we mentioned above, at the time of designing the study, we could not find any evidence of Democrats heaving benefited more from the use of online political advertisements during some election. We did not want to deceive participants in our study, since we incentivized answers to our dependent variable by sending them to members of the United States Congress in aggregated and anonymized form. Therefore, we did not want the answers to be based on beliefs that might have been influenced by deception. Thus, we decided not to include such condition. Following your valuable feedback, we added a paragraph in the discussion section to address this topic (p. 22, l. 565). Thank you very much!

5. Authors did never explicitly measure partisan self-interest. This makes it difficult to show the hypothesized mechanism. Why did they not measure this key concept?

Response: Thank you very much for addressing this important topic! We agree that we did not measure partisan self-interest explicitly. Instead, we propose that individuals are motivated by partisan self-interest if they believe the opposing party experiences an advantage from online targeted political advertisement and, as a consequence, try to mitigate this by supporting stricter regulation of these advertisements. The perceived advantage results from people believing that supporters of the opposing party are more easily influenced by mobilizing targeted advertisement from their respective party, which directly increases their chances of winning elections. The belief that supporters of the opposing party are more susceptible to online targeted political ads than supporters of the own party arises from the third-person effect (5). Hence, we argue that, within the context of our study, partisan self-interest is present when participants experience the third person-effect and act to mitigate its outcomes by supporting the regulation of online targeted political advertisement.

In order to show that partisan self-interest plays a causal role in the support for regulation of online targeted political advertisement, we manipulate partisan self-interest in our experiment. To do so, we inform the treatment group about the advantageous effects of online targeted political advertising for Republicans by telling them that Republicans were more influenced by targeted advertisements than Democrats in the 2016 presidential election. This way, we manipulate Republicans’ beliefs about their co-partisans’ susceptibility to online targeted political advertisements in a way that shifts their partisan self-interest towards these ads being in their favor, helping them to win elections. Therefore, we are able to draw inferences about participants’ partisan self-interest by manipulating the treatment group’s beliefs.

By asking participants about their beliefs a second time after measuring support for regulation, we are able to show that Republicans indeed shifted their beliefs towards thinking that Republicans, not Democrats, are more easily influenced by targeted advertisements. Hence, the treatment manipulation altered their perception of whether online targeted political advertisement would be in line with their partisan self-interest.

Many other studies that examined partisan self-interest considerations in the support of election reforms used cross-sectional survey data to correlate participants’ political affiliation with their attitudes (11–16). This body of research concluded that public opinion towards electoral reforms often reflects party divisions, but could not establish causality. With our design, we follow several other scholars who have used similar approaches to study the causal role of partisan self-interest considerations in attitudes towards regulations or reforms concerning the electoral process. In a recent study about the influence of partisan self-interest on attitudes towards election reforms, partisan self-interest was manipulated using different treatment conditions (17). The treatment conditions were phrased such that the reforms were either in line with Democrats’ or Republicans’ partisan self-interest. Similarly, in a study about the role of partisan self-interest in attitudes towards same-day registration, participants were divided in three groups (18). The groups were either given a text that described same-day registration as being neutral, or favorable for Democrats or Republicans, thereby altering participants’ partisan self-interest considerations. An analogous approach was used in a study about support for voter identification, in which participants were assigned to treatments either presenting voter identification laws as increasing turnout for Democrats or Republicans (19).

We have added more detailed descriptions of what we mean by partisan self-interest and how we manipulate it in the experiment to the introduction (p. 6, l. 120), the research design (p. 9, l. 200), the results (p.15, l. 381), and the discussion (p. 19, l. 471) sections. Thank you very much again for raising this point!

6. On p. 11 the authors state that perceived desirability of online targeted ads is low. But on a ten-point scale, perceived desirability is 4.66, sd = 2.01. I would not call this low.

Response: Thank you very much for pointing this out. We agree with you that this value is not low. We have rephrased this in the text (p. 12, l. 312; p. 18, l. 442).

I believe this study makes an important contribution to the field of online political advertising research, especially when issues 1 and 3 are addressed.

Response: Dear Reviewer, thank you so much for your valuable and insightful suggestions and feedback. We have done our best to address your comments in this revised version of the manuscript.

Comments from Reviewer 2

Manuscript PONE-D-20-33004 evaluates if Americans’ support for hypothetical regulations of microtargeting change when subjects are informed that microtargeting benefits Republicans while having no effect on Democrats. Results are not surprising based on the many published studies that find support for various (proposed) election laws shifts when partisans are informed that the proposed law disproportionately helps/hurt their party. In this experiment, the effects were only significant for Republicans – though, this is not clearly represented in the current wording of the abstract. I have a few concerns about the experimental components of the manuscript, as well as the article’s framings. Order is consistent with the manuscript’s layout and does not indicate priority.

Response: Dear Reviewer, thank you so much for your thoughtful feedback. Following your comment, we have adjusted the wording in our abstract to better reflect our findings. Specifically, we now state that only Republicans’ attitudes towards the regulation of online targeted political advertisement are partially motivated by partisan self-interest.

Further, while indeed multiple studies provide evidence that partisan self-interest considerations are associated with attitudes towards the election law, many of these studies are cross-sectional and, hence, do not establish causality (11–16). Several recent studies use experimental designs to measure causal effects of partisan self-interest on attitudes toward the election law (17–19). We follow this line of research to provide more experimental evidence of the causal relationship between partisans’ self-interest considerations and their opinions towards the election law. We now discuss these related studies in our discussion section (p. 20, l. 497).

Furthermore, the context of online targeted political advertising is relatively novel, with multiple stakeholders questioning current approaches to its regulations as well as attitudes of the constituents who drive this regulatory agenda.

Microtargeting in the United States is very different than microtargeting in Europe due to the accessibility of publicly available electronic lists of registered voters. States are required by the Help America Vote Act of 2002 (HAVA) to compile and maintain these lists that include basic information (e.g. voter history, party registration, and race/ethnicity). This information is far more valuable for campaigns than data they can purchase or acquire elsewhere. The authors do not seem to understand the distinctions between microtargeting in the United States and elsewhere. And they fail to cite any scholars of US microtargeting (e.g. Hersch, Hillygus, Endres, Nickerson, Rogers, Panagopoulos).

Response: Dear Reviewer, thank you very much for pointing out the specifics of political targeting in the United States and for referring to these important scholars. To make it clearer that the focus of our study lies specifically on online targeted political advertisements, we now more clearly differentiate between “traditional” targeted political advertisements and online targeted political advertisement. In the introductory text that all participants read in the beginning of the study, we referred to online targeted political advertisement as well.

Following your suggestions, we have included the findings of the researchers that were provided and have revised the introduction accordingly. In the introduction, we now mention a widespread use of large amounts of voter data for targeting in political campaigns that has become common since the implementation of the Help America Vote Act of 2002 (20–23). We further point out what is novel about online targeted political advertising in terms of the even larger amount of personal data being used, the increasingly advanced personalization techniques, and cost-effective ways to reach voters (24). We have also added more background information on legislation that has been introduced that would obligate stricter funding disclosure (25) or even ban online targeted political advertisements based on personal data (1).

The study’s motivation feels like a straw man argument. The authors note on page 3, “a heated public debate calling for stricter regulations has accompanied the emergence of such ads.” However, none of these citations reference a “heated public debate” about regulating microtargeting in the United States, and I am not aware of a debate or proposed legislation in any US state or nationally. The merits of microtargeting are more frequently debated outside the US than inside. After all, the United States Congress paved the way for political microtargeting through their HAVA legislation (see Hersh 2015). The authors appear to be conflating debates about political misinformation, dark money in politics, and false statements on social media with political microtargeting. Further, the authors are somewhat cavalier with their references throughout the manuscript.

Response: Thank you very much for this very helpful comment. Instead of speaking about a heated public debate, we now base our argument on different actors who have raised concerns regarding the use of personal data for online targeted political advertising. We now refer to scholars who critique the use of personal data and the lack of disclosure and transparency (24,26–29). We also mention a private initiative that seeks to raise awareness about personal data use for political targeting (30) and we have added more recent evidence of the public’s negative attitude towards the use of personal data for online targeted political advertising (31). In addition to this, we cite two federal level legislations that were recently introduced and that aim for increasing disclosure obligations (25) or even banning targeted political advertisements on platforms (1). Further, based on your comment, we have revised all references of our manuscript. In the introduction, we have deleted those that referred to the harms of online targeted political advertising also in terms of misinformation and foreign interference.

I do not consider the treatment, as described in the text, to be truthful. Though, the exact treatment language is not quoted in the text. The authors note, “we truthfully informed a randomly selected sample of participants that the Republican party benefited more than the Democratic party from the use of targeted political advertising in the 2016 presidential election” (page 6). I am not aware of any legitimate, peer reviewed publication that support this claim. It is possible, that Republicans benefitted more than the Democrats in terms of vote choice / persuasion, fundraising, (de)mobilization, but one CANNOT simply assume that the party of the winning candidate benefitted more than the party of the losing candidate. Later in the paper the treatment is described slightly differently as, “Participants in the treatment group were informed that controlling for the number of ads people saw, targeted political advertising on Facebook significantly increased voter turnout for the Republicans in the 2016 presidential election, while having no effect on Democrats ” (Page 8). The working paper cited to support this claim does not support it. In fact, it argues that targeted advertising had a negative effect on the turnout of Democrats.

Response: Thank you very much for this comment. We agree that it cannot simply be assumed that the winning candidate benefited more from the use of online targeted political advertisements. In our treatment, we informed participants that these advertisements “significantly increased the number of votes for the Republican party, but not for the Democratic party” and that they “influenced Republican voters, but did not influence Democratic voters”. We have added the instructions to the appendix. They are also available, together with the data and analysis files, at our data repository: https://osf.io/tynp7/?view_only=ccb6dcbfbb5a44ba985b572f959fb011. As noted in our response to Reviewer 1, the information we gave participants in the treatment was based on a working paper by Liberini et al. (2). In this working paper, the authors study the effect of exposure to online targeted political advertisements on Facebook during the 2016 U.S. presidential campaign on voter turnout and candidate choice. In order to do so, the authors exploit variations in prices of advertisements on Facebook to estimate the intensity of political campaigning for different audiences, specified by location, political affiliation and demographics. The idea is that given stable sizes of the audiences, more advertisements targeted to an audience should increase prices due to the demand shift. Hence, higher prices indicate higher campaign intensity. The authors then match this data with responses of 2,414 American voters from the 2016 American National Election Study (ANES) sampled once prior and once after the election regarding their Facebook use, turnout and candidate choice. The authors find that “targeted Facebook campaigning increased turnout among core Republican voters, but not among Democratic or Independent voters” (1: p. 5). They specifically state that “microtargeting was ineffective for Clinton, failing to boost turnout or sway voters in her favor” (1: p. 5). Thus, when designing our experiment, we based our treatment information on this working paper, aiming at changing the beliefs regarding the effect of online targeted political advertisements on opposing partisans of one group in our sample to be able to establish a causal relationship between partisan beliefs and support for regulation. Importantly, our search for scientific evidence of similar results of Democrats having been more influenced by online targeted political advertisements in some other election has not resulted in any meaningful findings. Therefore, as we did not want to deceive participants, we decided to manipulate only Republicans’ beliefs based on the evidence that we had at the time. After our survey had been in the field, a new version of the Liberini et al. (2) paper has been released in April, 2020. In this updated version, the authors conclude that being exposed to targeted online advertisement during the 2016 U.S. presidential campaign increased turnout for Trump supporters, and decreased turnout for Clinton supporters (3). Unfortunately, this information was not available to us when designing the study. In the discussion, we now report that results of the 2018 working paper have been updated (p. 23, l. 589) and we make it clear that the treatment text was written based on the results available at the time (p. 22, l. 581.).

More information is needed about the survey. Was there any quota sampling? If so, which variables were used? What was the time interval between each survey wave? What was the attrition rate? Some aspects/concepts are incorrectly described (e.g. a response scale from "not at all" to "to a very great extent" is NOT a Likert scale). Do you have a citations for the description of L2 as “the largest voting tracking service in the United States.” How did L2 and/or Dynata identify Democrats and Republican? Is this based on self-reports from an earlier survey? Party registration (not all states have party registration)? Votes in primary elections (most registered voters do not vote in primaries)?

Response: Dear Reviewer, thank you very much for raising these important questions. Quotas were applied on age, gender, region, and party affiliation. Following your comment, we have added this information in the sample description (p. 11, l. 262). Participants completed the survey within one session, no time passed between the different parts of the survey. We have made this clearer in the experimental design section (p.8, l. 173). We have further re-checked the information on L2. The information on being the largest such provider was based on self-description as provided on the website of L2 and we have not been able to independently verify this information. However, we found a description as “one of the largest” in Cappelen et al. (32). Therefore, we dropped this description. We also added additional information on L2’s methodology in the sample characteristics section, stating that respondents’ party affiliation was partially verified by their actual voting behavior and partially derived from other known attributes about the participants (p. 10, l. 257). Thank you very much for pointing out the wrong labelling of the scale, we have adapted the wording in the experimental design section (p. 8, l. 191).

The "manipulation check", “all subjects were then asked to make an estimation of the number of interactions (likes, shares, comments) that social media campaigns on Facebook of both Republicans and Democrats received relative to each other prior to the midterm elections in 2018” does not appear to directly evaluate whether the treatment group read and/or processed the information in the treatment.

Response: We kindly thank you for this remark. We agree that we did not ask participants directly about their beliefs regarding the influence of targeted political advertisements on partisans in the 2016 election. With this question, we aimed at presenting participants with a similar, but not identical context as in the treatment. We chose the 2018 elections because this way participants had to infer from information they learned about the 2016 elections to an election in the future. Thereby, we could test if participants generalized from the treatment text to Republicans being more susceptible to online targeted political advertisements in general. This is important because of our dependent variable, which asks for attitudes for regulation of targeted advertisement. If participants thought Republicans were only more easily influenced in the 2016 election, they would not use this information for forming their support for regulation of online targeted political advertisements. However, we agree with you that this measure is not a strict manipulation check. Based on your feedback, we have rephrased and reflected this in the text (p.7, 144; p. 16, 358; as well as captions of S12 Fig.; S13 Fig.; S14 Fig.).

Finally, all control variables should be collected pre-treatment (See Gerber and Green 2012; Montgomery, Nyhan, Torres 2016; and many others), but it appears many were collected in wave 3 after the treatment was delivered (wave 2)

Response: Thank you very much for pointing that out. In our results section, we report the result of a Welch t-test which compares support for regulation between Republicans in the treatment and Republicans in the control group to report a result that is not influenced by the inclusion of control variables: “With Republicans, we found significantly lower support for stricter regulation of online targeted political advertising in the treatment than in the control group (two-sided Welch t-test, t(776) = 2.08, Cohen's d = 0.15, p = 0.04).” (p. 16, l. 389.). However, we have to admit that the discussion of the regression results in table 3 might raise concerns over the control variables having influenced the treatment effect. In response, we have deleted table 3 to make it clear that our main treatment effect is robust and not potentially influenced by the inclusion of control variables that were measured after the treatment was administered. For the same reason, we excluded the S10 table in the appendix.

Further, it is important to emphasize that the correlational results that are reported in tables 1 and 2 are not influenced by the treatment as they are measured on respondents who did not receive the treatment information. These results show a correlation between privacy concerns and support for stricter regulation.

Dear Reviewer, thank you very much for your insightful suggestions and valuable feedback. We have done our best to address your comments in this revised version of the manuscript.

References:

1. Banning Microtargeted Political Ads Act, H.R.7014, 116th Congress. 2020.

2. Liberini F, Redoano M, Russo A, Cuevas A, Cuevas R. Politics in the Facebook Era. Online Working Paper Series No 389, Centre for Competitive Advantage in the Global Economy, The University of Warwick. 2018;72.

3. Liberini F, Russo A, Cuevas Á, Cuevas R, others. Politics in the Facebook Era-Evidence from the 2016 US Presidential Elections. CESifo Working Paper No 8235. 2020;

4. Pew Research. Political Independents: Who They Are, What They Think. Pew Research Center - U.S. Politics & Policy. 2019. Available from: https://www.pewresearch.org/politics/2019/03/14/political-independents-who-they-are-what-they-think/

5. Davison WP. The Third-Person Effect in Communication. Public Opinion Quarterly. 1983;47(1):1.

6. Perloff RM. Third-Person Effect Research 1983-1992: A Review and Synthesis. International Journal of Public Opinion Research. 1993;5(2):167–84.

7. White HA. Considering Interacting Factors in the Third-Person Effect: Argument Strength and Social Distance. Journalism & Mass Communication Quarterly. 1997;74(3):557–64.

8. Jang SM, Kim JK. Third person effects of fake news: Fake news regulation and media literacy interventions. Computers in Human Behavior. 2018;80:295–302.

9. Iyengar S, Lelkes Y, Levendusky M, Malhotra N, Westwood SJ. The Origins and Consequences of Affective Polarization in the United States. Annual Review of Political Science. 2019;22(1):129–46.

10. Bordalo P, Coffman K, Gennaioli N, Shleifer A. Stereotypes. The Quarterly Journal of Economics. 2016 Nov 1;131(4):1753–94.

11. Gronke P, Hicks WD, McKee SC, Stewart III C, Dunham J. Voter ID laws: A view from the public. Social Science Quarterly. 2019;100(1):215–32.

12. Wenzel JP, Bowler S, Lanoue DJ. Citizen opinion and constitutional choices: The case of the UK. Political Behavior. 2000;22(3):241–65.

13. Bowler S, Donovan T, Karp JA. Why politicians like electoral institutions: Self-interest, values, or ideology? The journal of politics. 2006;68(2):434–46.

14. Tolbert CJ, Smith DA, Green JC. Strategic voting and legislative redistricting reform: district and statewide representational winners and losers. Political Research Quarterly. 2009;62(1):92–109.

15. Alvarez RM, Hall TE, Levin I, Stewart III C. Voter opinions about election reform: Do they support making voting more convenient? Election Law Journal. 2011;10(2):73–87.

16. Stewart III C, Ansolabehere S, Persily N. Revisiting public opinion on voter identification and voter fraud in an era of increasing partisan polarization. Stan L Rev. 2016;68:1455.

17. Biggers DR. Does partisan self-interest dictate support for election reform? Experimental evidence on the willingness of citizens to alter the costs of voting for electoral gain. Political Behavior. 2019;41(4):1025–46.

18. McCarthy D. Partisanship vs. Principle: Understanding Public Opinion on Same-Day Registration. Public Opinion Quarterly. 2019;83(3):568–83.

19. Kane JV. Why Can’t We Agree on ID? Partisanship, Perceptions of Fraud, and Public Support for Voter Identification Laws. Public Opinion Quarterly. 2017;81(4):943–55.

20. Hillygus DS, Shields TG. The persuadable voter: Wedge issues in presidential campaigns. Princeton University Press; 2008.

21. Nickerson DW, Rogers T. Political Campaigns and Big Data. Journal of Economic Perspectives. 2014;28(2):51–74.

22. Hersh ED. Hacking the Electorate: How Campaigns Perceive Voters. Cambridge: Cambridge University Press; 2015

23. Endres K, Kelly KJ. Does microtargeting matter? Campaign contact strategies and young voters. Journal of Elections, Public Opinion and Parties. 2018 Jan 2;28(1):1–18.

24. Chester J, Montgomery KC. The digital commercialisation of US politics — 2020 and beyond. Internet Policy Review. 2019;8(4).

25. Honest Ads Act, S.1989, 115th Congress. 2017.

26. Burkell J, Regan PM. Voter preferences, voter manipulation, voter analytics: policy options for less surveillance and more autonomy. Internet Policy Review. 2019;8(4).

27. Rubinstein I. Voter Privacy in the Age of Big Data. SSRN Journal. 2014

28. Boerman SC, Kruikemeier S, Borgesius FJZ. Online Behavioral Advertising: A Literature Review and Research Agenda. Journal of Advertising. 2017 Jul 3;46(3):363–76.

29. Wood AK, Ravel AM. Fool Me Once: Regulating Fake News and Other Online Advertising. S Cal L Rev. 2017;91:1223.

30. ProPublica. What We Learned From Collecting 100,000 Targeted Facebook Ads. 2018. Available from: https://www.propublica.org/article/facebook-political-ad-collector-targeted-ads-what-we-learned

31. Gallup. In U.S., Most Oppose Micro-Targeting in Online Political Ads. 2020. Available from: https://news.gallup.com/opinion/gallup/286490/oppose-micro-targeting-online-political-ads.aspx

32. Cappelen AW, Haaland IK, Tungodden B. Beliefs about behavioral responses to taxation. Working Paper; 2018.

Submitted filename: Response to reviewers.pdf

Click here for additional data file.

8 Apr 2021

Partisan self-interest is an important driver for people's support for the regulation of targeted political advertising

PONE-D-20-33004R1

Dear Dr. Baum,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Natalie J. Shook

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: All comments have been addressed

Reviewer #2: (No Response)

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: (No Response)

Reviewer #2: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: Thank you for your elaborate and convincing rebuttal letter, and thank you for your contribution to the targeting literature.

Reviewer #2: I appreciate the authors’ response to the feedback provided by myself and the other reviewer. I agree the manuscript has improved overall.

I recognize the authors believed they were providing factual information (based on the cited working draft) when they conducted the survey. Now that they know otherwise, they should consider contacting the survey respondents assigned to the treatment group to notify them about the deception used in this study. The survey vendor (Dynata) should know which members of their panel participated in this survey, and, as such, it is possible to notify experimental subjects about the use of deception in this study. Further, the authors continue to use phrases such as “truthfully informing” in the text. At minimum, they should refrain from describing the treatment as truthful/accurate/honest. The content of the treatment can be presented to readers without claiming it is factual.

I agree with first reviewer that it is an overreach for the authors to claim they manipulated partisan self-interests. Instead, they randomized which subjects they exposed to information stating that targeted campaign communications benefitted one party and not the other. This information may or may not have influenced partisan self-interest.

The null effects for Democrats in the experiment is somewhat surprising. Generally, and based on the published experimental studies in this domain, you would expect Democrats and Republicans to move in opposite directions when provided with information that an aspect of elections benefits one party and not the other. After all, a campaign tactic that helps only one party mobilize their base has a net negative effect for the other party.

The exploratory analyses testing for heterogeneous treatment effects based on self-placement on a liberal-conservative scale is interesting. I am surprised the authors did not look at strength of party identification, which seems like a more natural choice.

The streamlined introduction is greatly improved. However, the last paragraph (of the introduction) is confusing (especially, the 2nd sentence).

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: No

16 Apr 2021

PONE-D-20-33004R1

Partisan self-interest is an important driver for people’s support for the regulation of targeted political advertising

Dear Dr. Baum:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Natalie J. Shook

Academic Editor

PLOS ONE