French First Round outcome – Dataswarm system predictions

French Election 2

Above – DataSwarm Analytic Engine – Zeitgeist memes for main participants in French election on Friday 21st april


We predicted it correctly, i.e. Macron & Le Pen, and in that order, but if you look above and read the “Underlying Analysis” below you will see we had a strange piece of data in that Hamon looked like a strong competitor, even leading Le Pen according to our system, and we only got it right by applying experience from previous work, ie:

  • We know from previous elections that  social media massively under-represents the Far-Right and over represents the Left, especially in the low penetration stage that France is in (and these swings can be huge, in the UK 201o elections where UK Twitter penetration was analogous to France today, the Liberals Democrats showed c 60% of support on Twitter – they actually got less thatn both Labour & the Tories)
  • Our system dynamic model of European elections to date shows that the Left is decimated and the centre-right gains support, and we couldn’t see anything changing that in France.
  • We also know (and have calibrated for other elections) that incidents like the Paris shooting drop support for liberal candidates and raise it for conservative, especially anti-migration ones.

So we got it right by applying a lot of judgement, and that is not ideal. We also need to go through the algorithms to see how Hamon scored so highly – was it the above small penetration effect or is there some combination of effects that drove a spurious overestimation?.

But to be fair to us, we used the US primaries to train the system (and us) and the French first round serves the same purpose, and the lower Social Media usage in France (and Germany) means you have to make a lot more adjustment to get an estimate of the overall demos.


We have been tracking the French elections on the DataSwarm engine, to see how effectively it can predict events by finding the election Zeitgeist (our term for memetic dominance, maybe can be thought of akin to “momentum” in US politics). Compared to the US elections in their primaries, for the French first round the only accurate thing we can say is the situation is unclear.

In France, one is not allowed to make any poll forecast from the day before an election (which is on Sunday) so we thought we’d keep to that rule to see what it was like predicting the outcome today (Friday)

As you can see in the chart above, which is a map of the memes of today c mid afternoon, we agree with most of the polls that Macron and Le Pen are up there, and our system thinks Macron has more support than Le Pen. Fillon has a lot of conversation (see the size of his bubble) but it is not as influential as the others. Also Fillon’s bubble contains quite a lot of negative sub-memes. Melenchon is being discussed by more influential media than Fillon, but the conversation is significantly less than any others.  But what we are also seeing is that Hamon is right up there with Macron on Twitter, in fact is more supported than Le Pen in both volume and influence. This is not what the polls say, and last week those positions were reversed on our system too, Hamon has made a great leap forward and Le Pen has fallen back a bit – on social media, anyway.

So, looking at our analysis of Twitter, it would predict that Macron and Hamon will go through. However, there are 3 major problems in predicting that Macron and Hamon will therefore be the winners of the first round, these being:

  • French use of social media is much lower than the UK and US, it’s more akin to the UK of 2010 (when if you believed social media, it said the Liberal Democrats would win – as we noted at the time) and we know social media always trends far more liberal than the overall population, especially when its in early adopter/early mass adoption mode which French adoption rates would suggest.
  • We also know that the Far Right is under-represented in social media, again especially in Social media’s early days – not just because liberals are more likely to be on it, but there are also many “shy Tories” even in liberal circles – right leaning people who keep quiet about their intentions to avoid losing liberal friends, colleagues or patrons.
  • Every time there is an event concerning Islamist terror or Migration, there is a defined and substantial shift to the right in the days following. Yesterday’s Paris events will have an impact, but we won’t see it’s full effect on the system until Sunday.

In short, we think that Hamon is probably over-represented on social media, Le Pen under-represented, and the Paris shooting will push Le Pen forward. Social Media also implies Melenchon is coming up fast but too late, and Fillon has too much negative baggage.

Because we haven’t tracked a French election before we don’t know the magnitudes of these factors. In the US elections we used the Primaries to train our system (and our interpretation abilities) so we were more certain of what we were seeing in the actual Election, and that’s what we’re really using the first round here for as well. (and overall that’s what we’re using elections for, to see what data the system pushes out on mapping memetic trends on social media against known outcomes)

So with all those caveats and uncertainties, then based on our metrics of “noisy liberals and shy rightists” effect on social media in the UK and US, and swings to the right when Paris-like incidents have happened elsewhere we’d predict that Macron and Le Pen will come through despite Hamon’s showing.

But, in politics at the moment, anything could happen in the next half-day….