|
?:abstract
|
-
We have been offering ex ante forecasts of German national elections since 1998, with articles initially appearing in Le Figaro (Jérôme, Jérôme-Speziari, and Lewis-Beck 1998, 2002) and then in other outlets (Jérôme-Speziari 2005; Jérôme, Jérôme-Speziari, and Lewis-Beck 2009). This work, novel for its time, enjoyed some success, encouraging us to continue our focus on the demanding German case, including pre-election publications in this journal (Jérôme, Jérôme-Speziari, and Lewis-Beck 2013, 2017). In the 2013 and 2017 elections, our political-economy model accurately foretold that Angela Merkel would rule in coalition with the Social Democratic Party (SPD). Those predictions followed our previous forecasts from 1998 to 2009, which also were correct except in 2002, when our single-equation vote function did not foresee the narrow Green Party victory over the Free Democratic Party (FDP). During this period of almost 25 years, we did some strategic tinkering with our model, it was hoped, to improve precision. For 2013, we tried to better incorporate the institutional features of proportional representation (PR) and the multiparty system (in particular, the role of smaller, less-mainstream parties). To further that effort, we constructed a Seemingly Unrelated Regression (SUR) model that simultaneously could estimate vote shares in a system of equations, thereby adjusting the estimates in any one equation by the necessary connection to the other vote equations (Zellner 1962). This approach, which explicitly considered parties large and small, was especially beneficial in 2017, with the noteworthy rise in more extreme parties: the Alternative für Deutschland (AfD) and the Left Party. These two blocs garnered almost 22% of the vote share, thereby denying main parties a majority. This also resulted in the formation of a “grand coalition” between the Christian Democratic Union/Christian Social Union (CDU/CSU) and the SPD, which—over time—has damaged their mainstream effort. As we face the 2021 contest, this AfD threat remains, along with a Green Party threat aimed especially at the SPD. Moreover, with Merkel no longer a contender, there are serious leadership issues that the parties must confront. The CDU ultimately selected Armin Laschet, even though polling indicates that voters favored Bavarian Minister President Markus Söder. As for the Green Party, there also has been a leadership struggle, with Annalena Baerbock chosen over Robert Habeck—despite the fact that polls showed them to be equal contenders. Thus, to the extent possible, we have adjusted our model to accommodate these uncertainties under this PR system. As we face the 2021 contest, this AfD threat remains, along with a Green Party threat aimed especially at the SPD. A final important measurement change was made regarding the conception of the dependent variable. We want to simulate the seats of the different parties as well as the coalitions that plausibly could form an absolute majority in the Bundestag. With this in mind, we abandoned the strategy of first predicting vote share using a voting function, then predicting seat share, in a swing ratio relying on the prior vote-share predictions. Instead, seat-share percentages are explained directly by economic and political determinants. This direct procedure allows us to significantly minimize the size of cumulative errors.
(xsd:string)
|
![[This page as RDF]](https://data.gesis.org/gesiskg/static/rdf-icon.gif){kind=icon}