(Mis)Guided by Data? Political Epistemologies of Data-Driven Campaigning in Germany

Parties increasingly use technology and data-based communication strategies for campaigning. These so-called data-driven strategies draw on digital technology, data, and analytics to identify voters and target them with tailor-made messages online (e.g. Facebook, Instagram) or offline (e.g. direct mailing, door-to-door canvassing). In the center of these strategies are political epistemologies which are practices by which political actors construct politically relevant knowledge about specific audiences or individual voters and the criteria by which they evaluate them. In data-driven campaigns (DDC), this knowledge is informed by algorithm-based models, which mine voter groups with similar characteristics, attitudes and behaviors (clustering) or predict unobserved behaviors or preferences (predictive modeling) from datasets, collected through polling or voters´ registration, communication, interaction and participation processes. Mostly, mobilizable or persuadable voters should be identified and targeted. There have been few empirical investigations into DDC and its epistemologies in European electoral settings. This paper provides a first empirical examination for Germany using the example of the Christian Democratic (CDU) and Social Democratic Party´s (SPD) mobilization campaigns in the Federal state election 2017. As a multi-party parliamentary democracy with strict data protection laws, less use of SNS for news consumption (24% Facebook; 5% Twitter) and less indications of a polarized electorate, Germany lends itself as important object of study outside the U.S. context. Three research questions structure the arguments: RQ1) What is DDC´s political rationale and main strategic objective? RQ2) What are the parties´ perceptions of the target “audiences”? How are they being created for which channels? RQ3) What kind of electoral, legal, cultural or ethical challenges do campaign operatives in Germany face? These RQs will be answered by presenting findings from 15 in-depth expert interviews with CDU´s and SPD´s campaign staff on federal and state level. Results indicate that CDU and SPD assign very high importance to DDC and the “predictability of human voting behavior” due to a fragmented media environment, changing society structures and eroding voter milieus. Both parties and even individual party strategists have a different understanding of DDC and use different data sources and models to compute voter targets for their Facebook and door-to-door campaigns. Important for their decisions are the perceptions of “audiences” on different platforms and channels, including their demographics and other characteristics; and candidates’ public personae. On Facebook, only the platform´s basic targeting techniques were used without uploading party´s voter data. The door-to-door campaigns use results from past elections, publicly available sociodemographic data and external data analysis. Apps are used to help track canvassers´ actions and gather voter response data in line with Germany´s data protection laws in voter management databases. However, German DDC is limited by budgetary, ethical or legal restraints, and party culture.