The German federal election has become a key election not only in Europe but also worldwide. Once again, this election will be closely scrutinized by observers and decision makers around the world. Because of its economic weight, Germany is a key player on the international scene, and its political weight within the European Council makes it a leading player in the governance of Europe. In this respect, it becomes important to be able to anticipate the outcome of German elections, especially in the context of the COVID-19 crisis and its long-term economic and political consequences. What will be the color of the future coalition? What will be the Christian Democratic Union/Christian Social Union (CDU/CSU)’s influence in the post-Merkel period? Will the Green Party supplant the Social Democratic Party (SPD) as a junior partner in the governing coalition? Will the right-wing Alternative für Deutschland (AfD) continue its rise? These are only some examples of many questions that this symposium on forecasting the 2021 German federal election addresses.

Although interest in election forecasting has increased during the past four decades, relatively little research has been conducted on the German case. In her contribution to this symposium, Stegmaier (Reference Stegmaier2021) describes the evolution of German election forecasting, which has been relatively slow since the pioneering works of Jérôme, Jérôme-Speziari, and Lewis-Beck (Reference Jérôme, Jérôme-Speziari and Lewis-Beck1998) and Norpoth and Gschwend (Reference Norpoth and Gschwend2003). This also becomes evident by reviewing previous PS: Political Science & Politics symposia on German election forecasting in 2013 and 2017, which hosted only two (Jérôme Reference Jérôme2013) and four contributions (Jérôme Reference Jérôme2017), respectively.

For the 2021 election, the symposium brings together the work of 19 researchers in the form of eight contributions, of which three already were featured in 2017 (i.e., Graefe Reference Graefe2017; Jérôme, Jérôme-Speziari, and Lewis-Beck Reference Jérôme, Jérôme-Speziari and Lewis-Beck2017; Kayser and Leininger Reference Kayser and Leininger2017). Graefe (Reference Graefe2021a) presents a revised version of the PollyVote method for combining forecasts from different methods, adopting changes that previously were made for the US version of the PollyVote. Jérôme, Jérôme-Speziari, and Lewis-Beck (Reference Jérôme, Jérôme-Speziari and Lewis-Beck2021) modify their political-economy model by directly predicting the percentage of seats going to different parties and better accounting for specificities of the German electoral system (i.e., 5% electoral threshold, proportional representation, and possible coalitions). Kayser, Leininger, and Vlasenko (Reference Kayser, Leininger and Vlasenko2021) update their Länder model, which draws heavily on election results in each German state (Bundesland) to generate a forecast of the federal election outcome.

The symposium also includes three forecasts that were not featured in previous symposia. Gschwend et al. (Reference Thomas, Klara, Simon, Marcel and Stoetzer2021) present a multiparty forecast based on the Zweitstimme model. This model uses a Bayesian approach to combining polls and fundamental data on party competition. Quinlan, Schnaudt, and Lewis-Beck (Reference Quinlan, Schnaudt and Lewis-Beck2021) propose a new type of structural model that does not include polling data. Their political-history model predicts the election outcome based solely on the long-term evolution of the German political system (e.g., grand coalitions, reunification, and party dominance in the Länder). Murr and Lewis-Beck (Reference Murr and Lewis-Beck2021) present a so-called citizen forecast, which draws on survey data on whom people expect to win the election.

For an overview of the vote-share predictions of these approaches, see Graefe (Reference Graefe2021a, table 1), who combines these (and other) forecasts in the PollyVote. As of June 21, approximately three months before the election, there was wide agreement on which party will gain the most votes. Regardless of the method and data used to forecast the election, Armin Laschet’s CDU/CSU is predicted to become the strongest party. However, with respect to the interesting question of who will come in second, the forecasts diverge. Whereas the Zweitstimme model (Gschwend et al. Reference Thomas, Klara, Simon, Marcel and Stoetzer2021), which relies heavily on polls, predicts the Green Party to be one percentage point ahead of the SPD, both the Länder model (Kayser, Leininger, and Vlasenko Reference Kayser, Leininger and Vlasenko2021) and the citizen forecast (Murr and Lewis-Beck Reference Murr and Lewis-Beck2021) expect the SPD to beat the Green Party by a comfortable margin of approximately six points. Based on expected seats, the political-economy model (Jérôme, Jérôme-Speziari, and Lewis-Beck Reference Jérôme, Jérôme-Speziari and Lewis-Beck2021) results in the same order of finish, with the CDU/CSU ahead of the SPD and the Green Party by 82 and 101 seats, respectively. Regarding the remaining parties, the forecasts agree that the Left Party, the Free Democratic Party, and the AfD are likely to again pass the 5% electoral threshold necessary for representation in parliament.

One limitation of vote-share forecasts is that the numbers do not provide direct answers to the questions that voters are most interested in, such as: Who is likely to govern? This is problematic because experimental evidence shows that voters are largely unable to derive this information from vote-share forecasts (Graefe Reference Graefe2021b). In their contribution, Bauer et al. (Reference Bauer, Bender, Klima and Küchenhoff2021) address this issue in showcasing common pitfalls when communicating vote-share results and proposing the communication of event probabilities as an alternative. Some of the contributions already follow that approach. For example, the PollyVote estimates an 88% chance that the CDU/CSU again will become the strongest party, which is identical to the Zweitstimme model’s 88% chance that Armin Laschet will lead the next government. Who will be the junior partner in the coalition remains to be seen.