Predicting election results in Europe can be daunting. In many EU countries, we are witnessing a political turmoil with very affirmed and long-standing parties being replaced by new emergent parties (for example in the 2015 Spanish Elections or the 2018 Italian Elections, Lithuanian 2020 Elections). This rapid change means that established voting patterns might not hold any longer, making election results more unpredictable.
To add to this problem, the validity and accuracy of electoral polls has been recently put in doubt. Famously, polls couldn’t predict Trump 2016, Brexit 2016 or Erdogan 2023. Why is that the case? Polls often use non-representative samples, therefore suffer from selection bias, which in turn harm estimates. Therefore, methods for bias correction are fundamental in order to ensure validity. Although many methods exist, they seem to improve the situation only in some limited cases and can also make it worse in others.
This project aims at evaluating current bias correction methods for election polls, with a focus on emergent parties: parties that gained popularity quickly between one election and the other. To do so, either existing electoral poll samples from past national elections in Europe are going to be collected. Subsequently, a comparison of different correction techniques will be carried out, in order to understand which method is most robust in providing the best estimates with respect to the election outcomes. The project will focus on bias correction for emergent parties, and will provide the scientific community with useful tools to make more valid and accurate predictions when analysing polls.