Negative Campaigning in German elections: Measurement, Dynamics, and Determinants

近年来，对负面竞选（即对政治对手的口头攻击）的研究引起了相当大的关注。研究表明，全国代表大会可能对民主产生功能失调的后果（例如两极分化，对政治的信任下降）。这尤其适用于那些以不文明的方式攻击对手的负面竞选策略。然而，NC的原因和影响很少在美国以外进行分析。本项目调查了在德国联邦和州一级的选举活动中使用NC的决定因素。目的是测试一个综合的负面竞选传播模型（IMNCC），该模型包括微观层面的解释因素（使用负面竞选传播的政治家）和宏观因素（竞选期间的结构性条件）。除了检验NC是理性成本效益考虑的结果这一经典假设外，IMNCC还检验了其他解释的可行性（例如价值观的作用、对NC的态度、个性、形象管理）。为此，我们首先使用候选人调查的自我报告来调查候选人对NC的使用和评价。我们借鉴了作为2013年、2017年和2021年联邦选举德国纵向选举研究（GLES）的一部分进行的候选人研究。此外，测试我们的综合理论模型需要新的变量的操作化。因此，我们将对巴登-符腾堡州、梅克伦堡-西波美拉尼亚州、莱茵兰-普法尔茨州和萨克森-安哈尔特州（约2400人）参加2021年州选举的所有相关政党候选人进行邮寄和在线调查。为了衡量NC的实际使用情况，我们收集了上述选举中候选人的整个Twitter通信。我们手工编写了一个高质量的训练数据集，作为大规模检测NC的机器学习模型的输入。与政党宣言等已建立的来源相比，Twitter的数据在更细粒度的层面（单个候选人）和更大的数量（竞选期间成千上万的帖子，以跨党派的标准化形式）上可用。这使得在综合模型中同时检查微观和宏观层面的各种理论解释成为可能。这两个数据源将在候选级别链接，并使用多元回归分析进行分析。我们将根据实证结果对负面竞选传播的整合模型进行修正。因此，该项目为更好地理解NC的决定因素做出了重要的理论和实证贡献(1)随着时间的推移，(2)在不同的联邦层面，(3)取决于不同的政治家个人特征。

Research on negative campaigning (NC) - i.e. verbal attacks on political opponents - has gained considerable attention in recent years. Studies suggest that NC can have dysfunctional consequences for democracy (e.g. polarization, declining trust in politics). This applies in particular to negative campaign strategies that attack an opponent personally or in an uncivilized manner. However, the causes and effects of NC have rarely been analyzed outside of the United States. This project investigates the determinants of the use of NC in German election campaigns at the federal and state level. The aim is to test an integrated model of negative campaign communication (IMNCC) that includes explanatory factors at the micro level (politicians who use NC) and macro factors (structural conditions during election campaigns). In addition to examining the classical assumption that NC is the result of rational cost-benefit considerations, the IMNCC also examines the viability of other explanations (e.g. the role of values, attitudes towards NC, personality, image management). For this purpose, we first use self-reports from candidate surveys to investigate the use and evaluation of NC by candidates. We draw on the candidate studies conducted as part of the German Longitudinal Election Study (GLES) for the 2013, 2017 and 2021 federal elections. In addition, testing our comprehensive theoretical model requires the operationalization of new variables. Therefore, we will interview all candidates of the relevant political parties who will take part in the 2021 Landtag elections in Baden-Württemberg, Mecklenburg-Western Pomerania, Rhineland-Palatinate and Saxony-Anhalt (N~2,400) in a postal and an online survey. In order to measure the actual use of NC, we collect the entire Twitter communication of the candidates in the above-mentioned elections. We hand code a high-quality training dataset that serves as the input for machine learning models for the detection of NC at a large scale. In contrast to established sources such as party manifestos, Twitter data is available at a more fine-grained level (of the individual candidate) and in larger numbers (hundreds of thousands of posts during an election campaign, in standardized form across parties). This makes it possible to examine various theoretical explanations at the micro and macro levels simultaneously in integrated models. Both data sources will be linked at the candidate level and analyzed using multiple regression analysis. We will revise the integrated model of negative campaign communication based on the empirical findings. The project thus makes an essential theoretical and empirical contribution towards a better understanding of the determinants of NC (1) over time, (2) at different federal levels and (3) depending on different individual characteristics of politicians.