Objective
Forecasting electoral results is, in general, a very difficult task. The D’Hondt rule makes forecasting Spanish elections even more challenging. The 2015 Spanish General Election is going to be especially difficult since there are new parties that have no electoral history in general elections.The polls taken in recent weeks show how difficult this task is, because they show very different proportions for the parties participating in the elections. The explanation used by pollsters is the fact that many electors (between 20% and 30% depending on the polls) are still undecided.
The team of researchers participating in this project needed to find a challenging endeavor and decided, probably unwisely, to take on this difficult task.
We are convinced that polls, as happened recently in Argentina or the UK, are not enough to get an accurate prediction of an election. Polls have to be weighted by the quality of the pollsters and the time since they were run, and they must be combined with a baseline model. A very successful experience on this approach is Nate Silver’s success in the prediction of electoral districts in the last two US presidential elections.