Replication Data for: A Länder-based Forecast of the 2021 German Bundestag Election

Polls are poor forecasts when elections are distant because they are sensitive to events and conditions that can change before election day. Structural approaches, which fit models on previous elections and fundamentals such as economic performance, are less influenced by idiosyncratic events and can establish baseline expectations about election outcomes. However, they suffer from high uncertainty and underspecification imposed by small samples. We present an early forecast for the 2021 election to Germany's parliament, the Bundestag, that remedies the small-sample problem by predicting general election vote shares in the 16 German states (Länder) using economic and political data as well as state parliament election results. As Länder elections are dispersed over the calendar, they allow us to capture both actual voter preferences and new political issues. We then combine our aggregated national forecast with polling data using a weighted average that increasingly favors polls as the election nears.