August 24, 2013

"Do it Again, Mommy!"

by Thomas Gschwend and Helmut Norpoth


Woody Allen could have written a better screenplay: The ruling coalition consisting of CDU / CSU and FDP is obtained at the next general election with a 51.2 percent share of the votes, a majority.

We owe this insight we developed a prediction model, the Chancellor model, which proved itself in the past three federal elections. Derived from theoretical approaches to the explanation of choice behavior, we have a statistical model developed, which is already in the summer before the last two general elections could provide 2002 and 2005 accurate predictions and tapped the right winner, while the results of opinion research institutes, based on the polls, the parties , were systematically wrong. The Chancellor's model provided a month before the election even more accurate predictions for the percentage of votes for government coalitions as all the established polling organizations, including some 18-clock forecasts on election night itself, although the surveys in 2009 actually were for the first time closer to the percentage of votes of CDU / CSU and FDP model predicted the Registrar at least one early Black-Yellow majority.

For the development of our predictive model, we wondered what similarities we can use for the prognosis of a coming election from the previous parliamentary elections in the history of the Federal Republic. We were interested particularly the common share of the vote each government coalition. This is a most relevant size, as a reigning coalition gets a majority on election day, they will continue to govern. 
Whether you may be hoping for a victory of a coalition government, we explain the interaction of long, medium and short-term factors. First, there is the long-term voter support of the governing parties - measured as the average electoral success in the last three federal elections. Moreover, secondly, the medium effective process of wear in office - measured by the number of terms (logarithmic). Third, the average popularity of the acting chancellor is a, measured by values  in surveys over the period of one to two months prior to an election.

Using statistical analysis methods, we can determine the interaction of these three factors and their weightings for voting in favor of the government coalition as follows.

Forecast for Black-Yellow 2013 = -7.12 + 0.72 * (PAR) + 0.40 * (KAN) - 2.72 * (AMT)

PAR: Long-term voter support of the governing parties (agents of the vote share in the last three federal elections) 
KAN: Registrar Support (excluding undecided) 
OFFICE: Abnützungseffekt (number of terms) logarithms,

For the forecast, we still need the current values  for 2013. For the longer-term voter support, the black and yellow enjoys among voters, resulting in a value of 46.4%. This corresponds to the average share of the vote, the CDU / CSU and FDP could win in the last three federal elections. There is also a support of 67% for Chancellor Angela Merkel on the basis of survey results in July and August.This value is calculated as the average percentage of respondents in the political barometers in July and the first two in August, preferring instead Steinbrück had as Chancellor Merkel, to the exclusion of undecided. Finally, Merkel has two terms behind him, creating a non-trivial Abnützungseffekt.

If these three values  into the above formula, which gives a forecast of 51.2 per cent of second votes for the combined share of the vote CDU / CSU and FDP on 22 September 2013. This forecast, confirming a preliminary prediction, which we already 2013 issue of the journal PS July on this blog and in the July: Political Science & Politics had published.