The official campaign has not started yet, but polls have been around for a long time. And thanks to that, our model goes back to May 2018 - do not thank us, the computer does all the work. Less than two months before election day - May 26 - it is time to take stock of the evolution of the balance of power since November 2018. As the "yellow vests'" first demonstrations started mid-November, we can analyze the correlations with polling results. So here are six takeaways.
But first, two precisions:
- Here, we will use the same prediction intervals as on our forecast page - i.e 83% intervals (5 in 6 chance). The reason is twofold: 1/ this shows that the 95% intervals you are used to see are none other than a convention. You are free to use any limit you want - the last two digits of your birth year, a prime number because you are a nerd, etc. 2/ This range is easy to interpret: a 5 in 6 chance is the probability of getting any number except 6 when rolling a fair die for example.
- For a detailed explanation of how our model works, we invite you to look at this article .
1. LREM lost nothing following the "yellow vests"
In early November - just before the first demonstrations - Macron's party (LREM) had, according to our model and available data, a 5 in 6 chance of getting between 19% and 24% of the votes. At the end of March, the presidential party had the same probability of getting between 20% and 25% of the votes. It has therefore slightly gained since the beginning of the "yellow vests".
5-in-6-chance interval of the popular vote
Solid lines represent the median share of the popular vote of each party. Shaded areas show the range in which the true popular vote is, with a 5 in 6 chance (83%). So a hypothetical range from 20% to 25% with a 22.5% median means that the party has a 5-in-6 chance of getting 20% to 25% of the popular vote, with a median share of 22.5%.
Why take a 5 in 6 chance as benchmark? Look at it as the probability of getting any number but the 6 when throwing a fair die. Hover over the chart to see the details. You can hide/display a party by clicking on its name in the legend.
Two nuances though. On the one hand, LREM lost 2 points at the apex of the demonstrations - from mid-November to mid-January - and did not return to its pre-"yellow vests" level until mid-February. Dozens of variables could explain this bounce - the start of the National Debate mid-January, demonstrations becoming more violent, a natural fatigue towards a movement that lasts, demands that had trouble emerging - but it is very hard to know which ones - if any - do (correlation is not causality).
And on the other hand, LREM is clearly not the front-runner it was in July 2018, when it had a 5 in 6 chance of getting between 23% and 28.5%. It lost about 3 points in 9 months and slid from clear favorite to barely favorite - 5 in 9 chance (55%) of finishing first.
And there is still one and a half month and before election day. The hard choices of the National Debate will inevitably disappoint lots of voters. Hence the presidential majority faces a wide range of scenarios - from a continuous momentum thanks to a well managed National Debate, to reinvigorated "yellow vests" because the National Debate disappointed too many.
2. Macron's popularity is a good predictor of its party's support in the polls
A good leading indicator of LREM's performance is the president's popularity . This may seem pretty obvious but is not often highlighted by conventional wisdom. At 0.93, the correlation is almost perfect: the closer a correlation gets to 1, the more the two studied variables go in the same direction. The closer it gets to 0, the less they are connected. The graph on the left panel strikingly shows that the two series move in the same direction.
Evolution of Macron's popularity and popular support for LREM
Popularity data come from our tracker , which uses all popularity polls and weights them according to several factors . Hover over the graph to see the details.
Thus, when Macron's popularity decreases, there is a good chance that LREM's support in the polls will too - and vice versa. Note that this correlation tells us nothing about the cause: perhaps popularity causes popular support? Maybe is it the other way around? Maybe a third, unobserved, variable, causes these two?
But even when ignoring causality there are two problems with basing a prediction on this correlation. First, popularity data are mainly published at the end of each month, so on election day you may not have all the available data. More importantly, support for each competing party is correlated with the others: what some lose, others win. So forecasting only LREM does not tell us much on the election as a whole.
3. Far-right and Green parties look stronger
RN (far-right) and EELV (green) are the only winners of the timeseries we examine. Early November, the Green party was neck-and-neck with the Socialists (PS) and Nationalists (DLF), with a 5 in 6 chance of getting 3.5% to 8.5% of the vote. Five months later, this range lies between 6% and 11%. And EELV is slightly ahead of far-left LFI, with a 4 in 9 chance (45%) of finishing before Mr. Mélenchon's party.
Distribution of the gap between EELV (green) and LFI (far-left)
Far-right RN saw an increase of similar magnitude, going from a 5 in 6 chance of getting 16% to 22% of the vote, to a 5 in 6 chance of getting between 19% and 24%. For now, RN is almost assured of finishing ahead of the traditional right party LR - while they were virtually tied back in September. According to our model and the data, RN even has a 3 in 10 chance of finishing ahead of LREM - hence of finishing first.
Distribution of the gap between RN (far-right) and LREM (center)
4. LFI (far-left) and DLF (nationalist) lost the most
LFI and DLF are the only two parties that lost ground between November and March. LFI's share persistently declined from 10%-15% to 5%-10% - a substantial drop that the far-left party tries to explain by supposedly biased polls . Why not.
But the thing is, on dozens of elections, we could not find anything else than a pro far-right bias . And there are many other, much more parsimonious, potential explanations. Let us just recall that Mr. Mélenchon's popularity sharply dropped after a video of his lashing out at police officers searching his party's offices went viral last October. He still has to revover from that fall. So if the correlation between the leader's popularity and the party's popular support is as strong as for Macron, the slow but steady decline in LFI's support is not that surprising.
DLF's trajectory differs from LFI's, as Mr. Dupont-Aignan's party was trending upward up until February. It particularly suffered in March, as evidenced by the evolution in the number of seats:
5-in-6-chance interval of the number of seats
Solid lines represent the median number of seats of each party. Shaded areas show the range in which the true number of seats is, with a 5 in 6 chance (83%). So a hypothetical range from 20 to 25 with a median at 22 means that the party has a 5-in-6 chance of getting 20 to 25 seats, with the median outcome at 22.
Why take a 5 in 6 chance as benchmark? Look at it as the probability of getting any number but the 6 when throwing a fair die. Hover over the chart to see the details. You can hide/display a party by clicking on its name in the legend.
Projected at 5 median seats until February, it dropped to 0 median seat in just one month. This means that the distribution of scenarios is centered at 0. In other words: the party has at least a 1 in 2 chance of getting no seat at all - in this case a 55% chance. Some of Mr. Dupont-Aignant's potential electorate apparently did not enjoy his early March controversial appearance in a TV show.
5. PS (left) and LR (right) are convalescing, but one of them has a restless life
Finally, the two former big parties' progression was very similar - they did not budge one bit. The Socialist Party (PS) has been stuck at the same level for months now - 5 in 6 chance of getting between 3.5% and 8.5%. More worrying, with the current data, the model estimates PS has a 4 in 10 chance of not getting any seat - it got 13 seats in 2014 .
Distribution of the number of seats obtained by PS (left) and associated probability
Each bar represents the probability that the Socialist Party gets the indicated number of seats. The higher the bar, the higher the probability. When a party gets less than 5% of the votes it gets 0 seat, which explains the threshold. Hover over the chart to see the details.
The Republicans' (LR) campaign is more hectic. Its median support fell to 13%, then rose to 16% at the end of March - basically the level it has in November. Although LR won 2 points in March, we don't know if this support reflects a conscious choice or a default solution depending on its competitors' media setbacks. The latter case would involve an enduring volatility - rise in March, fall in April. LR's emotional rollercoaster may not end anytime soon.
6. Unlike 2017, there are only two main parties
This article is getting long and I can feel your impatience. So let us conclude by noting that RN's and LREM's position in the polls is very similar to their score in the first round of the 2017 presidential election. It looks like Emmanuel Macron and Marine Le Pen kept their voter base over the last two years.
Conversely, the other two presidential contenders find it hard to position themselves in the new political landscape - LR is still about 4 points below its 2017 score and LFI is 11 points below.
So here is a strange paradox: LREM and RN have not increased, LR and LFI have gone down, but nobody benefits - for now.
This article was quite long, but we think this kind of coverage - less frequent but more in-depth - suits our purpose better - introducing the scientific method into campaign coverage. It takes more time to write them, but it is the price to pay for less descriptive and more analytical work. Feel free to ! In particular, how often would you want to read these articles?
Until the next one, here is a summary of the progression of each party's popular support:
5-in-6 chance interval of the popular vote (%)
November 2018 | March 2019 | Median support progression | |
---|---|---|---|
Party | |||
RN | 16-22 | 19-24 | 2 |
EELV | 3.5-8.5 | 6-11 | 1.6 |
LREM | 19-24 | 20-25 | 1 |
PS | 3.5-8.5 | 3.5-8.5 | 0 |
LR | 14-19.5 | 13.5-19 | -0.6 |
DLF | 4-9 | 2-7 | -2.3 |
LFI | 10-15 | 5-10 | -4 |
Share it when you see a fallacious use of polling. Because remember: a probability distribution is better than an average, which is better than an isolated poll.