A case study of the German election 2021
I know, I know… we don’t usually write about politics around here. And we’re not starting today, either. As we’re based in Germany, where a major federal election is coming up later this month, we thought we could use that event to show how you can use social media data to predict outcomes– or at least delve into people’s feelings about public matters.
Whenever they can be done independently, polls are and will stay the most accurate way to do political forecasting.
We do a lot of data visualization and KPI reporting for our clients on a regular basis, so we decided to use our expertise in this field to do an experiment. Our aim is to try out a use of data collection, analysis and visualization that we don’t usually employ in our field and show how it can, with little effort, provide a view of what internet users– which in 2021 means most of the population– like and dislike.
Can social media likes forecast behavior?
Of course, the idea that social media likes alone can accurately depict public sentiment is simplistic and can lead to a somewhat skewed view. We are aware that people with strong feelings for or against a political idea are much more likely to like or dislike it on social media than, say, a person who is not as politically engaged, or simply has less time to stay informed, or would rather get their news from other sources. It will be interesting to see if, in spite of these caveats, candidate visibility and engagement rates on social media are still good at predicting election outcomes.
We’re not the first ones to use social media engagement to predict election results, but we do think that this might be an easy way to analyze pre-election political sentiment in countries or certain constellations where opinion polls are scarce or not available at all. Not that this is the case in Germany, mind you. We’re drowning in polls in this country. And, whenever they can be done independently, polls are and will stay the most accurate way to do political forecasting.
Youtube likes as a forecasting tool
So here’s our hypothesis: we think that visibility and acceptance on Youtube videos displaying the candidates might positively correlate with their result on election day. We have focused on Youtube videos because it’s easy to use their API to connect with data visualization tools (you can find more information here). This is basically something anyone can do. It’s like a DIY poll!
We selected the main Youtube channels covering the 2021 campaign. We analyzed interviews, video presentations and some minor online appearances of the three main candidates. We then connected the Youtube API to Tableau, one of our favorite data visualization tools, to extract views, likes and dislikes. With this data available, we created a public dashboard that will be updated with data from newly uploaded video material every other day until election day (September 26th).
German election candidates on Youtube
For this analysis we focused on the three candidates that were doing best at the time the dashboard was created. We are aware that other candidates / parties might be able to gain enough momentum to get themselves ahead in the time between the publication of this blog post and election day.
The candidates and parties we’re covering are:
- Armin Laschet, from the conservative CDU/CSU (Christian-Democratic Union / Christian-Social Union). Laschet belongs to the same party (CDU) as Angela Merkel, who has been the German chancellor for 16 years and is not standing for reelection. He has been in charge of the federal state of North Rhine-Westphalia since 2017. CDU and its Bavarian sister CSU agree on a common candidate ahead of each election.
- Olaf Scholz, from the social-democratic SPD (Social-Democratic Party of Germany). Scholz has served the government as a finance minister and vice-chancellor in Angela Merkel’s coalition government since 2018. He has to perform the political balancing act of simultaneously being the opposition leader and a member of the current government.
- Annalena Baerbock, from the green party Bündnis 90/Die Grünen (Alliance 90 / the Greens) is the youngest and least experienced of the three. She hasn’t been involved in government but addresses topics (like climate change) that are very important to German voters.
Political newcomers get more Youtube views
The fact that Baerbock is a newcomer is probably getting her more views than the other two candidates, since people might feel they don’t know her as well and want to gather information before making a decision. Scholz, on the other hand, has been around for decades and has a prominent role in the outgoing government. Less internet users will be compelled to go online and find out what he stands for since most people already know him. As for Laschet, he’s better known in his home federal state of North-Rhine Westphalia than in the rest of Germany.
From our sample so far, most videos had a higher ratio of dislikes than likes. But here, too, there are differences between candidates. Interviews are the type of content with the most views, but also the most dislikes.
We are eager to watch our experiment develop as the election nears and will keep an eye on our dashboard. Make sure to check it out and draw your own conclusions as election day gets closer. You can also explore the data on your own by filtering out by candidate, channel, party and content type.