Heterogeneous voter preferences and fame avoidance behavior. Explaining strategic issue (non-)emphasis on migration policy in Germany 2015-2021

Recent research on “strategic issue emphasis” (Baumann, Debus, and Gross 2021) takes the dynamic nature of election campaigns seriously and goes beyond the traditional reliance on party manifestos. In line with the literature on issue competition and ownership, the thrust is to explain how parties position themselves strategically in relation to other parties in a competitive electoral context (Petrocik, Benoit, and Hansen 2003; Walgrave and Swert 2007). A much smaller part of the literature focuses on how and why parties translate perceived distributions of voters’ policy preferences into emphasizing issues. So far, the difficulty of communicating with divided voter groups has been debated mainly as a problem of centre-left parties in the field of migration (Carxvalho and Ruedin 2018): Social-democratic parties face a particularly challenging situation to serve the distinct preferences of both blue- and white-collar voters. However, divisions among voters on migration are a problem not just for center-left parties (Widfeldt 2015; Odmalm and Bale 2015): If parties restrain communication on divisive issues, as we hypothesize, communicative gaps between parties and citizens emerge. Considering the overall salience of migration and immigration policy after 2015, we take Germany as a case: After the extensive immigration faced by Germany in 2015/16, the issue remained salient for a long time. However, considerable legislative debate and activity notwithstanding, major parties restrained themselves from actively emphasizing the issue. Doubts about issue ownership - the widespread assumption that talking about immigration would play into the hands of Germany’s right-wing populist party (AfD) - may be one explanation. Adding to this conventional wisdom, we pursue the hypotheses of “fame avoidance behaviour”: Parties facing splits in their electorate on issues of immigration will avoid actively communicating on the issue. To address this issue systematically, we combine survey data and text-as-data approaches to assess the effects of the homogeneity or diversity of electoral preferences on parties’ communicative choices. Digital data provide an opportunity to move beyond using campaign manifestos to measure the strategic issue emphasis of parties: Facebook posts of parties are a continuous source of information on the strategic communication of parties. We have built a Facebook corpus (Rüdiger and Dayter 2020) that includes all posts of parties on their Facebook pages from 2015 to 2021. For measuring issue emphasis, we apply topic modelling and supervised machine learning. The methodological challenge is making algorithms for (un)supervised learning work robustly for short pieces of text. A corpus of parliamentary debates (GermaParl) can be used to evaluate parliamentary attention to migration affairs, and to detect patterns of deliberate public non-communication of parliamentary activity. Questions on policy preferences are included in major surveys on electoral behaviour in Germany (GLES), providing us with a measure of the distribution of migration-related policy preferences across electoral groups. Combining conventional survey data with digital (digital) data, we seek to simultaneously make a methodological and substantial contribution.