Political Targeting on Social Media: A Study of Tailored Messages and Ad Success in the German 2021 General Election

The academic literature about political targeting argues that parties adapt their messages to specific voter groups to maximise electoral support (e.g. Franz 2003; Strömbäck et al 2013; Stückelberger, 2022). Social media platforms like Facebook or Instagram are ideally suited for political targeting as parties can draw on personal data and targeting options to send tailored messages only to selected users (Kruschinski et al. 2022). However, we know little about how parties tailor the content of their targeted social media ads and how this tailoring is influencing the ads’ success in terms of how many people see an ad (ad impressions). Based on theoretical expectations derived from the literature about group appeals in electoral campaigning (Thau 2019, Thau 2021, Huber 2022), party competition (Downs 1957, De Sio and Weber 2020) and digital political marketing (Kruschinski et al. 2022; Lees-Marshment, 2019), we will provide empirical answers to the following two research questions: RQ1: Do parties tailor their social media ads by using different group appeals and different issue content for specific voter groups (defined by age, gender, family status, region and interests)? RQ2: Are tailored messages more successful in terms of ad impressions? We tackle these questions by analysing all Facebook and Instagram Ads (n~4200) bought by the five biggest German parties in the last four weeks of the 2021 general election: First, we conducted a content analysis using targeting categories to identify different group appeals and issue content in the ads. Second, we merged the content data with the novel Meta Ad Targeting Dataset, which was released in the summer of 2022, to find out what fine-grained target groups parties selected for their ads, as well as how many impressions and how much spending each ad received. We match the strategic selection of target groups with the targeting criteria age, gender, family status, region and interests to the content analysis coding if the ad was tailored to these criteria. This gives us the opportunity to explore which target groups the parties wanted to reach, and how they adapted the message content to these target groups. To measure the success of a tailored ad, we conducted a negative binomial regression model with the ads’ impressions as the dependent variable. The findings will provide novel insights about the strategic use of targeting tailored messages on social media and how effective this strategy is.