Hunting for Voters: The Impact of Data-Driven Campaigning on Democracy

The rapid growth of data-driven techniques has greatly enlarged the influence of digital media on political campaigning. Political messages are increasingly matched to individuals through social media. Matching takes place using large databases containing personal information that voters – knowingly and mostly unwittingly – leave online, such as individual characteristics, preferences, and behaviors. Many fear that these political microtargeting techniques on social media are a harmful phenomenon. We know, however, astonishingly little about whether or not political targeting practices actually sway voters, ultimately affecting, and possibly undermining elections and democracies. The dominant view that data-driven campaigning has detrimental effects per se can also be questioned. In this paper, we propose and test a novel dual-processing theory of political targeting in different country contexts. Our model is based on both the Elaboration Likelihood Model (ELM) and Motivated Reasoning Theory. We focus on two important factors when confronted with targeting messages: whether people are motivated and able to process the message. If people are more motivated and able to process the message, then people will engage in deep processing. The matched message is inherently more relevant, thus people engage in issue-relevant thinking and read, remember and store the targeted message better. If people are less motivated and able to process the message, then the citizens will engage in superficial processing. The matched message will function as a cue. People see it is relevant for themselves, yet because they lack motivation and ability, they do not engage in issue-relevant thinking. Whether processing has an influence depends on people’s goals (based on motivated reasoning). When people follow accuracy goals, people attend to the information, process it carefully and invest in reasoning. However, when voters follow directional goals, people want to protect and maintain a predetermined attitude or conclusion and reject dis-confirming information. To empirically test our model, we will conduct already planned preregistered experiments in March in the U.S., India, and the Netherlands. In the experiment, participants will be asked to use a mock social media site on a daily basis for a week where targeted messages will be shown. We will manipulate how many times respondents are exposed to a specific targeted message (ability) and how well this message fits the interests of the respondent (motivation). The arguments in the targeted message will either be in line with the position of the respondent on a certain issue or against the position of the respondent on a certain issue. We will thus create different conditions for different respondents. We will measure the normative outcomes (i.e., political knowledge, voting intentions, interest, issue position) at the end of the week. By making the comparison between three countries with a very different political context, we gain insight in how similar the impact of microtargeting on voters exactly is.