Election Forecasts for the German Federal Election 2025

谁将是下一届联邦议院中最强大的政党，谁将成为新总理？选举制度改革后，议席将如何分配？红绿灯联盟将重新获得议会多数席位并继续执政，还是将被中右翼或左翼联盟取代？绿党会再次赢得城市地区几个名义上的选区吗？这样的问题将成为新闻媒体、政治专家和选举研究人员感兴趣的中心。在这个研究项目中，我们在2025年德国联邦选举的背景下设计了一个整体预测努力，包括三个工作包，不仅涉及预测本身，而且还涉及选举预测实践的潜在后果。在第一个工作包(WP1)中，我们在关键维度上创新了德国联邦选举预测模型。我们建议扩展我们在过去的文献中建立的选举前民调数据的动态预测模型，将其与公民和精英的个人调查数据相结合，这些数据也与地区竞选有关。此外，我们计划整合保角推理的最新发展，以更准确地校准基于过去性能的不确定性。在第二个工作包(WP2)中，我们创建了用于科学选举预测的调查数据数据库，与市民实施滚动横断面设计，并与地方精英进行调查。这项针对公民的调查将与一家现有的调查机构合作，从2025年德国联邦选举前几周开始每周进行一次。通过在实地期间对每个选区约100名受访者进行访谈，我们计划建立一个丰富的受访者样本，以稳定当地的预测，并了解当地的竞选动态及其对预测的下游影响。此外，我们计划在选举前几个月招募地方精英，即地区候选人和记者，以激发他们对地区竞选的期望。在第三个工作方案(WP3)中，我们系统地评估了接触选举预测对政治行为和态度的影响。选民、政治精英(地区候选人)和地方专家(记者)如何理解和利用科学的选举预测？选举预测对政治行为和沟通有什么影响？为了解决这些问题，我们在WP2中进行的数据收集的基础上进行了一系列实验，以研究我们的选举预测对收件人的信念、公民的投票意向、候选人的定位和记者对选举竞选的报道的影响。此外，我们希望调查如何有效地传达选举预测，以告知而不是扭曲消费者的信念。

Who will be the strongest party in the next Bundestag, and who will become the new chancellor? How will seats be distributed after the reform of the electoral system? Will the traffic light coalition regain a parliamentary majority and continue to govern, or will it be replaced by a center-right or left-wing coalition? Will the Greens again win several nominal districts in urban areas? Such questions will be at the center of interest of news media, political pundits, and election researchers alike. In this research project, we devise a holistic forecasting endeavor in the context of the German Federal Election of 2025 that entails three work packages and not only addresses the forecast itself but also potential consequences of election forecasting practice. In the first work package (WP1), we innovate German Federal Election forecasting models on key dimensions. We suggest extending a dynamic forecasting model of pre-election polling data, which we established in the literature in the past, by integrating it with individual-level survey data from citizens and elites that also speak to district races. Furthermore, we plan to integrate newer developments in conformal inference to calibrate the uncertainty based on past performance more accurately. In the second work package (WP2), we create a database of survey data for scientific election forecasting, implement a rolling cross-section design with citizens, and conduct a survey with local elites. The survey among citizens will be fielded in cooperation with an established survey institute weekly starting several weeks prior to the German Federal Election 2025. By interviewing around 100 respondents per electoral district over the course of the field period, we plan to create a rich sample of respondents that can be used to stabilize local forecasts and inform on local campaign dynamics and their downstream effects on forecasts. In addition, we plan to elicit expectations about the district races from local elites, namely district candidates and journalists, that we recruit months before the election. In the third work package (WP3), we systematically assess the consequences of exposure to election forecasts for political behavior and attitudes. How do voters, political elites (district candidates) and local experts (journalists) understand and make use of scientific election forecasts? And what are the consequences of election forecasts on political behavior and communication? To address these questions, we build on the data collection carried out as part of WP2 and aim to field a set of experiments to study the effect of our election forecasts on recipients´ beliefs, citizens' vote intentions, candidates' positioning and journalists' reporting about the election race. Moreover, we want to investigate how to effectively communicate election forecasts to inform and not to distort consumers' beliefs.