Statistical Politics in the Eurovision Song Contest

In what may be a career-limiting move, I’m going to admit it: I am obsessed with the Eurovision Song Contest. Forget about the glittery-techno-windmachined-keychanging-pyrotechnic-flagwaving-Europop scene – if you’re into granular predictive data, the Eurovision is heaven. Every year, the voting elicits all sorts of theories – ranging from political bloc collusion, to voting structure criticism, to the UK protesting that no one likes them. By virtue of 5,435 raw, independent modelling points from the 2014 contest supplied by the European Broadcasting Union, I hope to address some of these claims.

Here’s how it works: each of the 26 qualifying countries submits a musical act to represent them, in the hope that the 37 other countries vote for them. The voting system consists of two components: a televote (whereby the public in each country vote for another country, and these are ranked), and a jury vote (whereby five members of the voting country’s music industry rank the acts independently; these are combined into a final jury rank). Finally, each country’s televote and jury vote ranks are combined and converted into a final rank, which is then converted to points. This year, Austria’s very talented and very bearded drag queen, Conchita Wurst, received the most points from Europe, and was crowned the winner.


In short, absolutely. I’ve decided to measure this using the Gini Coefficient (G), which takes the ranks allocated to all other countries by a given cohort, and measures the tendency for that cohort to rank their fellow bloc countries higher than other countries. Using this statistic to measure the extent of bloc favouritism, the following blocs were significant:

  • Non-Baltic-ex-Soviets (NBES)
  • The Warsaw Pact
  • Scandinavia
  • Countries with Germanic Official Languages
  • Baltic States
  • Balkan States

The table below indicates these blocs with their overall Ginis, and the individual country Gini contribution to this.


Bloc Ginis are between -100% and 100%: a higher Gini suggests a group’s tendency to favour bloc states, a negative Gini implies a group’s ability to penalise their fellow bloc states, and results close to 0% (roughly between -15% to 15%) suggest an insignificant bloc effect. Countries have individual Ginis, and this is can be aggregated to an overall bloc Gini.

While proponents of bloc voting often cite Eastern European competitors as beneficiaries of bloc voting, this isn’t necessarily the case: with a very high bloc Gini (G=71%), Scandinavians can’t seem to get enough of each other. With the exception of Finland ranking Iceland 12th (out of 26), Scandinavians always made each other’s top ten.

Even without the bias of the aforementioned Scandinavia effect, Germanic-speaking countries exhibited a high discriminatory voting pattern (G=63%). The Danish televoters’ top eight consisted of all other Germanic-speaking finalists except for Germany (G=94%). Also demonstrating Germanic favouritism were combined votes from Switzerland (G=75%), Germany (G=74%), and the UK (G=65%).

Moving to Eastern Europe, the most notable trend was that Armenia and Azerbaijan, who are technically at war, both ranked each other last in their respective televotes and jury votes. For the purpose of bloc analysis, this individual interaction has been excluded. All NBES countries exhibited a high level of internal bloc voting (G=79%), driven by Russia (G=100%), who ranked all NBES states above everyone else. If the Baltic States were included, this would drop to 66%, because Baltic juries appeared statistically neutral to ex-Soviet states.

Despite voting this year, the last time all three Baltics qualified for the final was in 2002, where they appeared to help each other out (G=81%).

There was a visible split in the Warsaw Pact countries between ex- Soviets (who we’ve found have demonstrated internal favouritism) and the non-Soviets (this year represented by Albania, Poland, Hungary and Romania). Although Albania and Poland statistically penalised the Warsaw Pact (G=-39%, -22% respectively), the other two non-Soviet states, Hungary and Romania, appeared to vote neutrally.

Balkan countries didn’t make a large appearance this year, but results from 2012 indicate a significant Balkan voting bloc (G=42%). There were some unsuccessful groupings. The Eurozone was not homogeneous, and interestingly, Greece punished fellow Euro countries (G=-33%). The Schengen Region exhibited varied results by country; counterintuitively, Switzerland, whose new immigration laws may threaten its Schengen status, favoured the region most (G=81%).


Notably, the Georgian jury result was disqualified this year, as allegedly each jury member’s top eight was identical.ConchitaWurst

Besides this, Montenegro and Azerbaijan had significantly lower intra-jury spreads than all other countries. The five Montenegrin jurors had the same top ten countries, but ordered marginally differently – no country’s placing diverged beyond two places from average. Or, perhaps I should have said top 11 – each of the five jury members just so happened to rank France as 11th, who eventually finished in 26th (last) place. Azerbaijan’s jurors also ranked uniformly: the rank awarded by each Azerbaijani juror to any country didn’t deviate from another juror by more than four places – comparatively, for 90% of the other juries, this number was between 15 and 24.


At the 90% level of confidence, a correlation coefficient of 0.27 between the rankings allocated by an individual cohort and the aggregated ranks from across Europe is significant. All countries’ televotes were significantly positively correlated with the final outcome. Despite this, only 22 of the 36 juries (excluding the disqualified Georgia) had a significant positive relationship with the overall result. The word positive here is critical: the earlier point around potential collusion may have driven a significant negative correlation for the Azerbaijani jury (-0.32), while the Montenegrin jury was marginally insignificant (-0.25). And yes – this means that inverting their result would provide predictive insight.



As juries consist of five music industry professionals, juries could be more likely to reward musical integrity rather than pyrotechnically (or otherwise) captivating scenes. Poland’s act this year divided juries and televoters; while the televoters generally rewarded the sight of scantily dressed Slavic milkmaids (based on all televotes, they were ranked fifth overall), the juries did not (based on all juries, 24th). Azerbaijan’s act was conversely divisive: its slow, subtle melody failed to captivate televoters (collectively ranked 26th, last place), but the juries rewarded it (seventh).

Furthermore, while every countries’ televotes placed the eventual Austrian winner, Conchita Wurst, in their top five (aside from Estonia, who ranked her eighth), and most juries following suit, some juries were less generous: the juries from Armenia, Azerbaijan, and Belarus (arguably less drag queen-friendly) placed Wurst in their bottom three.

To address the claim that juries’ ranks can be skewed by political or economic factors more so than the actual performance, I ranked countries by their Purchasing Power Parity (PPP) per capita as a proxy. Sounds extreme? The table below shows countries whose jury votes had a stronger relationship with PPP rankings than the final result – even though their televote had no correlation with PPP rankings1.


So then, who is the perfect country to win this political glitterstorm? They’d need an eastern longitude to appeal to the ex-Soviets, but not be an ex-Soviet to appeal to the rest of Eastern Europe; they shouldn’t use the Euro in order to escape Eurozone backstabbing, but speak a Germanic language to secure a ten-country-strong Bloc vote.

I’m talking to you, Australia2.

1 A positive jury correlation with ppp ranking suggests that the jury favoured countries with a higher ppp, and a negative correlation suggests favouritism of countries with lower ppps.

2 This year, the European Broadcasting Union “reward[ed] Australian Eurovision loyalty” by allowing Jessica Mauboy to participate in an interval act, but not to compete (perhaps this restriction has something to do with Australia not actually being part of Europe).

CPD: Actuaries Institute Members can claim two CPD points for every hour of reading articles on Actuaries Digital.