The Geography of Electioneering in Multi-Party Systems: Evidence from Campaign Poster Locations in Central Berlin

Election posters are an omnipresent yet understudied feature of election campaigns. While some researchers have studied poster design and its processing by voters in lab settings, the spatial dimension of poster placement — where parties place their posters and why — has received almost no systematic attention. This study examines the geography of electioneering in multi-party systems, drawing on a novel dataset of geolocated poster locations from Friedrichshain-Kreuzberg, a dense and politically diverse district of Berlin where evenly spread posters could reach most residents, making it a least-likely case for spatial targeting. By walking every street of the district and photographing all visible posters during the 2025 German federal election campaign, we assembled more than 6,800 geolocated photographs capturing over 10,000 individual posters, to our knowledge the most comprehensive dataset of its kind. We extract and classify every poster with a five-stage automated image-analysis pipeline that combines object detection (Florence-2, GroundingDINO), CLIP-based clustering into campaign motifs, and GPT-4o classification. Its output closely matches hand-coding and offers a reproducible and transferable way of measuring physical campaign material. Linking the full corpus of geolocated and classified posters to precinct-level electoral results from 2021, we show that mobilization is the primary logic of poster placement: Parties place their posters close to where they did well in the previous election. This is not to say that parties do not chase new voters at all; however, when they do, they fish in the pond of parties that belong to the same political bloc as they do and utilize their core issues (i.e., issues important to their electorates). These findings highlight the enduring strategic relevance of offline campaigning and call for greater attention to the geography of party competition.