mead berger shevtsova garfinkle michta grygiel blankenhorn bayles
the state of europe
All Bets Are off as French Election Looms

With the first round of France’s pivotal presidential election just days away, polls show the major candidates separated by razor-thin margins. At this point, it seems plausible that Saturday’s voting could produce a final round face-off between any two-candidate combination of the top four—Le Pen, the far-right populist, Macron, the technocratic centrist, Fillon, the traditional center-righter, or Mélanchon, the die-hard socialist.

Meanwhile, U.S. polling guru Nate Silver, who came closer than most mainstream pollsters to predicting Donald Trump’s upset victory, says the real state of the race may be even more unpredictable still. He suspects that pollsters may be “copying one another and over-massaging their data instead of letting their data speak for itself,” and that it’s possible that one candidate or another is actually in the lead.

As the political scientist Matt Grossmann has pointed out, a modest swing toward any candidate in the final days of the race could dramatically change the media narrative after results come in—from a ratification of the nationalism of Brexit and Trump to the rise of a new leftwing revolt against the status quo to a vindication of the centrist establishment. But even if it is the last of those three, the fact that all bets seem to be off so close to election day, with two once-marginal ideologies making serious bids for power, shows just how unprecedented a time this is in French politics, and how fragile the EU project really is.

Features Icon
show comments
  • D4x

    Polls. Elections. At least France has a fallback to barricades and burning cars, unlike the Turks. Wake us when the Brits vote in May.

    • leoj

      For May? In June

      • D4x

        Fixed it as well as can be expected, after a year+ of sleep deprivation. A vote for May in May would be confusing.

  • Pete

    “Meanwhile, U.S. polling guru Nate Silver, who came closer than most mainstream pollsters to predicting Donald Trump’s upset victory, ..”

    Sliver was still wrong.

    • Ellen

      Exactly. He gave Trump a 30-35% chance of winning, which was wrong, wrong, wrong. He even made the qualitative comment that it was highly unlikely that Trump would win because of all the demographic arguments that the NY-Washington crowd have been making for years. They were all wrong. Please don’t quote Nate Silver anymore, along with not quoting Tom Friedman anymore. They have both been wrong far too many times to take seriously as prognosticators, at this point.

      • Y.K.

        Leaving aside that 30-35 chances come true quite often (how can we prove Silver was wrong?), nobody deserves being compared to Tom Friedman.

        • Ellen

          First, I agree with you about Tom Friedman. Sorry for that ignominious comparison.

          Regarding the 30-35% comparison. Yes, sometimes those odds do work out. However, Silver (like the other pitiful posters who got all the elections of the last few years wrong: UK, Brexit, Israeli, US) is hiding behind quantitative statistical gibberish to cover the fact that he agreed with all the less mathematically inclined predictors that it was very unlikely that Trump would win. He gave Trump higher odds than most of the others who put Clinton in the 80-90% bucket, but he was doing this just as a dodge, so he could later say:”look, I was closer than the other fools.” In elections, close isn’t good. Nobody hires pollsters or campaign managers to be close. They hire them to be winners.

          • Y.K.

            I don’t follow Silver closely, but the end result as far as election modelling is concerned is the chance percentage and 30-35 isn’t “very unlikely”. It’s possible his actual method is “guess who’s winning and than add a few % to the likely loser”, but I do recall in 2012 he was willing to be more bold about Obama than most modellers were. I’ll give him the benefit of the doubt for now.

            In any event, I don’t really considered modelling by polls the best method. A simple look at the political events could have told us the odds on what happened were much higher than certain columnists allowed:

            * Miliband couldn’t sell himself to the British voters.

            * Leave had momentum and a much higher chance than the market though they had. The betting market simply lived in complete denial, even denying their own polls even the worst of which gave much higher chances to Leave than they did.

            * Bibi started strong, fumbled a bit by the end, at which point the voters took one good look at the alternative (which had an horrible closing week) and ran screaming back.***

            * Clinton ran a ‘New coalition’ style campaign which completely neglected WWC vote. If she had just listened to her husband more she’d have won.

            *** Well, there’s the Amit Segal reconstruction which suggests tens of mandates moving around all around upto election day. I don’t buy it at all. To me, it looks like an actual case of pollsters and campaign manager doing CYOA.

      • Jeff

        Which was not a statement that it was highly unlikely. Silver is really good at this and was parsing the polling, which was better last year than in 2012. The result on a state by state basis was within the margin of error. Silver looked at those state polls and assessed Trump’s likelihood of winning lower based on history and the need for A LOT of things to go right.

        Pennsylvania, Michigan, Wisconsin all were inside the margin of error. So were polls in Florida and Ohio. Florida polling ended up WAY off. Had that been more accurate, Silver would have probably put it at 50/50. In fact, as soon as he saw the Florida numbers begin rolling in, he adjusted the projection that night to OVER 50% because he began to adjust based on results. He put Trump up to 70% chance by 8PM Eastern time that night, long before others did the same.

        Silver has been right most of the time. He was dead on in 2012 and he’s gotten the midterms right. In this case, most polls were pretty good at the state level. But he also knew the odds of Wisconsin, Michigan, Pennsylvania, Ohio, and Florida all breaking that way WERE NOT good. The swing votes came from districts that went for Obama TWICE.

        Silver never predicts things outright, he just looks at the data and creates probabilities. And Silver changed his odds even faster than VEGAS did. They didn’t have Trump as the favorite until 9:30. Silver shifted him into favorite status around 8PM based on the Florida results coming in.

        It’s also been documented that Trump’s own people didn’t think he had a chance to win until the evening before the election when they looked at data.

        Silver is only as good as the data he gets and the data was pretty solid, but everything had to break the right way for Trump to win and it did. But you can never put good odds on that.

        • Ellen

          The problem with polls and pollsters goes well beyond the issue of statistical accuracy. The reason many polls are NOT accurate is because certain segments of the electorate (those that are looked down upon by the liberal elites, usually) are not telling the truth about whom they plan to vote for and why? This is now true across many societies (US, UK and Israel, for example) and that is why Silver’s predictions in these elections were all wrong. He was correct regarding many of the state polls, in US presidential elections, but his predictions regarding UK and Israel were as bad as the others. He was hired, unbelievably, to do polling in these other countries on the basis of his 2012 accuracy for the US presidential race, which was his high-water mark.

          If you want to read a really well written explanation as to why pollsters who are trained as statisticians (like Silver) do not get things right, read Janet Daly’s essay on this subject from last spring in the Telegraph. She hit the nail on the head, and I am not going to recapitulate her lovely reasoning and writing. In short, pollsters come from the narrow class and viewpoint as the liberal/left elites who are now being tossed out of office in virtually all Western democratic countries. There is a reason for it and a good one. Let us see how well Le Pen does, and whether that bolsters the Daly argument. Touche!

  • FriendlyGoat

    The world is seeing a spate of political catastrophes bumbled through on very thin margins. One can only hope that the French are noticing this and resolving to prevent themselves from being just another one.

    • Anthony

      I’ve been reading and commenting on ViaMedia/TAI now about 10 years and have noticed both 180 degree change in general postings and more ideological hard edge to inclined commentary (generally). Reading your above comment brought that observation to mind in conjunction with: the sociological core of what we’re all living through – a growing socio-cultural chasm/ a growing sense of split.

      • FriendlyGoat

        This effect is a bit like being in cool-water pot with the burner on low. Oops, it’s warm now. Wow, it’s growing hot. Dang, were cooked before we know it.

        • Anthony

          Apt metaphor! “People often ponder how the Germans, or the Italians, or the Japanese in World War II, or the Europeans in 1914, or the citizens of countries involved in many often idiotic conflicts over the centuries, could sit and do nothing while the agitators, dividers, callous, narrow, distempered, etc. foment chaos/disaster.” Yes, it is easy to imagine that “dang, were cooked before we know it” has historical precedent – And as an addendum, rhetorical hostility benefits whom precisely?

    • Y.K.

      There’s a talking point about ‘narrow margins’ but it simply isn’t true. Lets take a look at some recent votes considered ‘catastrophes’ by some:

      * The Brexit vote support was comparable to that of recent British governments (for example, the 2007 elections were won with 35% of the vote). If the UK’s FPTP standard were used instead it would have been a crushing win of over 400 seats.
      * Trump won by a lower margin than most but still comfortably (Clinton could have won PA or FL without changing result). Trump, of course, lost the popular vote bigly.

      * The Italian referendum was a narrow 59-41 win for No.

      * Turkey had a fair vote, where the only voter decided to give himself 51% of the count… Still I think it’s more accurate to say Erdogan won 100-0.

      None of these cases were really ‘narrow’.

      • FriendlyGoat

        Trump won the electoral college by fewer than 100,000 votes total spread over WI, MI, PA—–less than one tenth of one percent of the people who voted nationally. Brexit was 52,48. Turkey is closer than 52-48, or whatever “they say” it is. I’ll stick with big errors on small margins.

        • Y.K.

          Well, if we selectively switch people around we can shift quite a few elections. 200,000 votes in Ohio in 2004. A few hundreds in Florida in 2000. A few thousands in IL in 1960. To me, this isn’t so much a measure of election closeness but of “interesting ways to gerrymander votes if it were legal”. And given that we’re talking about statewide votes, I don’t think that the comparison to number of people who voted nationally is valid. If 2 million less people had voted in California, would it have made the elections less close in WI or MI?

          52-48 is quite normal for contested elections. Not decisive, but pretty respectable by historic standards if you compare to other elections.

          Turkey looks to me more like a case of ‘one man, one vote’, or at least so remote from a real democratic vote than speaking of ‘elections’ or ‘closeness’ is a case of category error.

          • FriendlyGoat

            I know conservatives do not acknowledge that Donald Trump’s election was a “barely, barely, barely” squeaker——but it was. Some 125,000,000 voted and less than 100,000 put him in.

          • Y.K.

            I’m not a conservative or even an American. I’m pretty sure that had Clinton been less sure of herself she’d have won. But she didn’t and there you are. As for your point, given that the US has a statewide system, the national vote total doesn’t really say much about ‘closeness’.

          • FriendlyGoat

            You sure do seem pleased with an ultra-conservative outcome in the USA for someone who claims not to be a conservative. Whatever.

    • Angel Martin

      Nate Silver’s observation observation about pollster result “herding” means we don’t really know how close this will be.

      One of the reasons the pollsters are having problems is that the old left/right categories are breaking down.

      In my case, I would vote for Le Pen as long as she is on the ballot. But in the case of a 2nd round of Melenchon vs Macron or Melenchon vs Fillion; I would vote for Melenchon.

      Melenchon may be Hugo Chavez in a scarf, but he represents at least the attempt at change on: EU/euro/NATO/globalization/trade/immigration

      Macron and Fillion are bankers and crooks who represent nothing but the EU and the Ile de France elite.

      • FriendlyGoat

        I never understood people who seemed to “kinda/sorta” like both Donald Trump and Bernie Sanders at the same time. And I don’t much understand coin flips between Le Pen and Melenchon.

        • Jim__L

          Easy. Both Trump and Bernie could tell there was something very, very wrong with the way things have been going. I’m not surprised a die-hard, bubble-dwelling Hillaryite can’t figure this out.

  • Beauceron

    I would love to see a run-off between Le Pen and Mélanchon.

    For my own amusement, if nothing else.

    I know it’s wrong. But I can’t help it.

    • Angel Martin

      Given the high levels of voter dissatisfaction and the problems with candidate polls, this is currently my base case.

  • Angel Martin

    Nate Silver is sure right about herding. If all these pollsters actually had true random samples from exactly the same “likely voter” population there would be more variation in the published polls due to sampling variability alone.

    Pollsters in France have to report on their “methodology”.

    There are lots of problems with what they are doing but the biggest i see is doing online polls only, when 15 percent of the French population is not online.

© The American Interest LLC 2005-2017 About Us Masthead Submissions Advertise Customer Service