Can a like predict electoral results?

As experts and polls have failed to accurately predict the outcome of everything from the American presidential election to Brexit and municipal elections in Denmark, the polls as a method of predicting is heavily doubted. In Nextwork we share this distrust. And we find that the problem of the predictions origins from using obsolete methods to understand voter preferences. Simply but, regular polls used to small and imprecise data samples and focuses, almost solely, on the voters own reflections of a party, candidate or the political situation in general.

Instead of asking for reflections, we argue people’s digital behavior is a much better predictor of actual voting behavior. Combining online and offline data, we have searched for patterns between party preferences and public Facebook activity and apply machine learning classifiers to explore the link between their political liking behaviour and actual voting intention.

Along with our partner, Analyse & Tal, Nextwork uses publicly available data to create engaging strategies for political organisations.