Congratulations to Sweden’s Loreen for her victory in this year’s Eurovision Song Contest! And to Liverpool for being the best of host cities. And thanks too to the Eurovision statisticians on posting the (almost) complete voting data for Liverpool 2023 on their website within 24 hours of the result!

This post looks at the detailed voting results. It refers back to previous analysis, particularly in this article.

Unusually, Sweden won the contest by winning the jury vote but not the public televote. In every other contest under the current scoring system, the televote has overruled the juries (except when they happened to agree in 2017). This chart shows this year’s jury scores (x-axis) versus the televote scores (y-axis), with the diagonal lines representing the total score – so the winner is whoever gets furthest towards the top right of the chart.¹

As in previous years, the public gave their winner more points than the juries gave to theirs. The public was also more likely than the juries to give songs very low scores (although nobody got nul points this year).

Sweden’s jury total was almost twice that of the jury’s second favourite, Israel – an unprecedented margin in recent contests. This can be seen on the following chart showing the probabilities of individual jurors preferring Song 1 (x-axis) over Song 2 (y-axis). The deep red/blue shading associated with Sweden reflects the fact that Sweden was preferred to every other song by at least 70% of jurors.²

Along the diagonal of the chart are the estimated Plackett-Luce weights, where the ratio of the weights indicate the chance of one song being preferred over another.³ This year, Sweden’s weight was almost three times that of the second place song (Italy). Last year, the UK was top with a weight of 100 and Sweden, in second, was on 92, with a gradual decline along the rest of the pack. Nine songs last year had a weight greater than 36 – the second highest weight this year!⁴

The juries’ preference for their winner this year was therefore extreme, in comparison to previous competitions. This is what gave Sweden enough points to (just) avoid being beaten by the public’s favourite, Finland. Of course, the public also gave Sweden a good total as their second place song, with Norway in third. The voting public seemed unusually keen on Scandinavian songs this year!

This year saw the introduction of the “Rest of the World” (ROW) televote, which further increased the power of the public votes over the juries. The full rankings of the ROW televote have not (yet) appeared on the Eurovision website, but their points were awarded as follows…

• 12 to Israel,
• 10 to Finland,
• 8 to Armenia,
• 7 to Sweden,
• 6 to Albania,
• 5 to Ukraine,
• 4 to Norway,
• 3 to Croatia,
• 2 to Spain,
• 1 to France.

Interestingly, this is exactly the same as the set of points awarded by Iceland’s televoters, and it is also quite strongly correlated with the scores from France, Belgium and Greece.

It is hard to read too much into just one year’s set of ROW data, but the ROW’s preferences appear to be most closely correlated to the western-European (orange) cluster in my previous analysis…

As I previously speculated, the ROW vote is likely to be affected by foreign populations living abroad. It is possible that the large Jewish diaspora may have helped Israel towards its 12 points, although as Israel was among the better scoring songs in any case, we can’t be definite about this without further analysis and/or more years of data.

1. A similar chart for previous years appears in the previous article.
2. The chart for last year appears in the previous article.
3. See the discussion of these in the previous article, and that in Significance.
4. Note that Plackett-Luce weights can be arbitrarily scaled up or down, as they are only meaningful as ratios. These have been scaled so that the highest value is 100.