Online attention to a party is often read as a sign of its support. But there are different kinds of attention. People click and react on a party’s Facebook page. They also look the party up on Wikipedia. I compared the two.
The paper
Looking Them Up: Wikipedia Attention, Facebook Engagement and Votes for European Parties, 2015 to 2020
Read the paper (open access) · Abstract and PDF on my site
What I did
I followed 112 European parties in 17 countries from July 2015 to September 2020. For each I took views of its Wikipedia articles, in the local language and in English, from the Wikimedia pageviews data, and interactions with its Facebook page, from CrowdTangle (Meta’s former tool for tracking public pages).
What I found
The two kinds of attention were related but far from identical: within country-months, the median rank correlation between parties’ shares of views and of interactions was 0.50. Both rose around national elections, with different timing. Facebook interactions rose by about 160% in the month before an election and stayed at about that level in the election month. Wikipedia views roughly doubled in the month before and were almost five times their usual level in the election month, when results bring people to look parties up.
In 20 elections in 14 countries, parties’ shares of Wikipedia views in the two months before the election month tracked their vote shares more closely than their shares of Facebook interactions did (a correlation of 0.84 against 0.62). But both were less close than their previous vote shares (0.92). Adding either measure to the previous result cut the average error of out-of-sample predictions only slightly, from 4.0 points to 3.8 or 3.9. Parties without a previous result drew about twice as large a share of attention as of votes.
Wikipedia attention was a better guide to parties’ relative size than Facebook engagement, but neither added much to knowing how they had done last time.
Caveats
The comparison covers 20 elections in 14 countries. Attention is not support, and the paper is a preprint that has not been peer reviewed.
The views in this post and the paper are my own.
