Conclusions of our case study on Germany’s 2021 election
As Germany woke up to a tight election outcome, we at advalyze rushed to see how Youtube engagement and likes/dislikes ratio of the three major candidates correlated with their parties’ official election results.
A few weeks earlier, we had set up this experiment where we extracted likes, dislikes and views from Youtube videos featuring the three candidates most likely to gather the highest amount of votes: Olaf Scholz (SPD), Armin Laschet (CDU/CSU) and Annalena Baerbock (Bündnis 90 Die Grünen). You can read about our methodology in our latest blog post “Can social media behavior predict election outcomes?”.
The best predictor for the election outcome was the likes/dislikes ratio, more specifically the dislikes (thumbs down).
Our goal was to use our data visualization tools to see if people’s behavior on social media could be used as a predictor of election results. Our hypothesis was that there would be a correlation between election results and Youtube behavior, but we couldn’t say which of the metrics (views, likes, dislikes) would be best at predicting voting behavior.
How does social media engagement correlate with voting behavior?
It turned out that the best predictor for the election outcome was the likes/dislikes ratio, more specifically the dislikes (thumbs down). The number of dislikes negatively correlated with the candidates’ results in the polls. This means that the social-democrat Olaf Scholz, whose party garnered the most votes, got the least number of dislikes. Armin Laschet was second and the green candidate, Annalena Baerbock, received the most dislikes and also came third in the polls.
As for views and likes, other factors probably affected the results. The average age of supporters of the different political parties and the comparatively young age of Youtube users did presumably play a role. Views seem to be more linked to a candidate’s prior public visibility rather than to voting behavior. As we explained here, the awareness level of the three candidates ahead of the election was very different.
On the other hand, people are more eager to click on the thumbs (either up or down) when they feel strongly about something. As voters appear to feel antipathy towards opposing politicians more than passion for the ones they actually end up voting into office, the thumbs down turned out to be a more reliable forecaster of the election result than any other indicator. Put simply: people feel “dislike” more strongly than they feel “like”.
See the dashboard, including election results, here.