Who will succeed Angela Merkel?

Dieser Artikel ist auch auf Deutsch verfügbar.

AFTER ALMOST 16 years as Germany’s chancellor, Angela Merkel will step down after federal elections on September 26th. But who will take the reins of Europe’s biggest economy? Our forecasting model blends the latest polls with other predictive indicators to come up with the most reliable guess at how many votes and seats each party might win—and measures the uncertainty of the contest.

The fragmentation of German politics in recent years has made the outcome of this contest highly uncertain from the outset. It has also been a volatile campaign. At various times Mrs Merkel’s centre-right Christian Democrats with their Bavarian sister party the Christian Social Union (CDU/CSU), the Greens and the centre-left Social Democrats (SPD) have all led the polls. In July our model estimated that the SPD had only a 3% chance of being the biggest party in the Bundestag after the election. Its chances later climbed as high as six in seven, but in the closing stages of the campaign the outcome has again become much less certain. This uncertainty is reflected in the wide 95% confidence intervals for parties’ shares of seats and votes. For example, the interval for the share of seats that might be won by the CDU/CSU ranges from —% to —% (with a mean of —%) .

The model also indicates that several putative coalitions might muster a majority of seats. The most plausible are a left-leaning “traffic-light” combination of the SPD, Greens and the Free Democrats (FDP) and a more conservative “Jamaica” coalition of the CDU/CSU, the Greens and the FDP. In any event, our model suggests that Germany is headed for its first three-way coalition at federal level since the 1950s. It finds a 60% chance that no two-party coalition will be possible.

How does our model reach these conclusions? It aggregates the polls, and then takes into account, from the history of past elections, how those intentions might change before election day. We also calculate the polls’ historical accuracy and explore what could happen if they stray from the mark. In taking into account these extra sources of error, our forecasts thus have a wider margin of error than the polls usually report.

You can learn more about the runners and riders, and many of the possible coalitions, or read more about the methodology behind our polling model.

To estimate the degree of uncertainty about how many seats each of the leading parties will win—and thus the election’s eventual outcome—we generate thousands of “simulation” elections that vary each party’s share of the vote according to the historical error of the polls at that point. Then for every simulation we divide each party’s share of votes by the combined share of all the parties that win representation in the Bundestag: ie, we remove those parties that do not clear the “5% hurdle” and assume those parties will win a negligible number of seats at the constituency level. Averages and uncertainty intervals are calculated from these simulations.

These simulations also account for the tendency for some parties to rise and fall together or at others’ expense—for example, our model says that when the Greens beat expectations, the CDU and the CSU ought to suffer more than the FDP. This way, we have a better estimate of how many seats each party could win conditional on potential movement in the polling average and uniform bias in the underlying polls.

Once votes are cast and seats are allocated, coalition talks get under way. Our model’s simulated seat shares can also easily be translated into forecasts for putative coalitions. For each simulated election, we simply add up the seats won by each constituent party. The average predictions and uncertainty intervals are calculated from these individual additions.

You can read more about the methodology behind our polling model here.

Sources: Wahlrecht.de, The Economist