The Eurosearch Song Contest 2019
What if Google searches were used to award points in the Eurovision Song Contest? And how close is this prediction based on search activity to the real Song Contest’s outcome?
The Eurovision Song Contest is one of the biggest popularity contests in the world. Anyone in one of the 41 competing countries can vote for their favourite song through a continent wide televoting system. But today, popularity can also be measured in another way: by counting online searches. So, how close was the Eurosearch Song Contest to the Eurovision Song Contest?
The ranking
If countries awarded points based on search activity for other countries’ contestants, using Eurovision’s famous 12-point voting system, the ranking would look like the one on the left of the below chart. The overall ranking of the 2019 Contest is displayed on the right, the televoting ranking is in the middle.
Final winner Duncan Laurence from the Netherlands was fourth in the ranking based on search activity and Russia is third on both the search and the overall ranking. But the most searched participants, from France and Iceland, did not end up high in the televoting and overall ranking. People probably searched for those candidates for other reasons than liking their songs and considering awarding them a televote. Norway won the televoting, which was also not picked up by search activity before the final.
On the map below, you can see which countries searched the most for the contestant of your country, and for which contestants your country searched the most. Switch to "Televoting" to compare search activity with the televoting results, and switch from the geographic map to the grid map to better represent smaller countries.
Search and televoting mapped
A look back
Search activity proved itself to be a good predictor for the Contest outcome in previous years: in 2017 search activity predicted the Contest winner correctly, last year the predicted runner up won it. This year the signal from the search activity was a bit weaker. So, can we confidently say that searches can reliably predict the Eurovision Song Contest? Or were the good predictions based on coincidence?
To check this, we gathered the data from the past 15 editions of the Contest. Because voting rules changed quite considerably during this period, we focused on the televoting data. The televote winner was the Contest winner by definition from 2004-08. Since 2009, scores from a professional jury have also contributed to the outcome of the Contest. But usually the televoting results are in line with the overall results: only twice, in 2016 and this year, the televoting winner was not the overall winner.
The chart below shows a strong correlation between the average points for search activity and televoting results between 2004-19. The Russian candidates were the most searched for, while Bulgaria received the most televoting points on average. Czechia, Australia and the Netherlands are below the line of equal search and televoting points: they received less televoting points then could be expected from the search activity for their candidates.
Televoting versus search activity
Average points, 2004-2019
But not all countries have participated in all finals of the Contest: after all, only 26 countries take part in the final every year. Countries that have participated in many finals will have scored more points and will have had more exposure to drive searches. We need to look at the data for each year separately to see if predictions based on search result can be reliable.
Searching vs voting by year
For some years the correlation between search and televotes was pretty strong (see 2007, 2008 and 2014-2017), in others it was a bit weaker. In 9 of the 16 years, the top-searched candidate also received the most points from televoting. In 8 of these 9 years, the televoting winner was also the overall winner, so a prediction based on search activity would have been correct half of the time.
Critics of the televoting system have suggested that countries fail to vote objectively for the best songs. It is suggested that Eurovision voters are heavily influenced by historical, cultural and linguistic ties with other countries when choosing where to cast their votes. This is backed up by the data. The following chart displays the cumulative televoting points awarded from one country to another during the 2004-2018 period. Notice how the country pairs with the most search activity (right of the chart) and most awarded televoting points (top of the chart) have shared borders, language or history.
Who searches/votes for whom?
Search and Eurovision
Online search activity usually doesn’t have too much predictive power. You can’t predict elections and the outcome of things like sports events based on what people search for. But the Eurovision Song Contest is a special case: search activity is strongly correlated with the outcome of the Contest. There are two main reasons for this.
The first is that people go online to find information about the artists and songs they like. They don’t do that for songs they don’t like. This is unlike search behaviour for politicians, for instance: people also also search for information about politicians they don’t like (like in the case of political scandals, for example). So the signal in the search activity data is strong.
The second reason is that people have a direct way of influencing the outcome of the Contest through the televoting system. In sports, search activity for a team can’t predict the outcome of next weekend’s game. But in the Eurovision Song Contest, people searching for a candidate they like can have a direct say in the outcome.
So do we still need the televoting system? Or should we just use search data to determine who gets our "douze points"?