Polling the Titanic


A quartet, to begin with.

Paris is so damned beautiful. Unabashedly beautiful. Pdv has lived in/ known many cities, including Berlin, Tokyo, New York, Boston, Munich, London, Rio, Delhi, Mumbai.  And yet after years that could have led to dull familiarity, I am heartstruck every time I go outside.  Please consider the quartet of photos scattered on this page, snapped in wind and cold and rain, in the so-called dead of winter. There is no possibility of visual kitsch in Paris, with the possible exception of the Eiffel Tower. But even the Tower was a strong and meaningful landmark when the Women’s March set out.


Deja-vu all over again!

Three months before the French Presidential elections, the political landscape is imploding. Up is down, left is right, and so on. Does this remind Brits and Yanks of any recent electoral experiences?

The French version of seismic upheaval began with the landslide primary victory of former Prime Minister (under Sarkozy) Francois Fillon. His image as a hard-working, unspectacular public servant, coupled with experience and by far the most level-headed and detailed program of all the aspirants, catapulted him from the lower rungs up to national favorite by a broad margin (55% of the electorate in December’s polls) for the final elections. Meanwhile, the Socialists, in full suicidal stampede, chose another dark horse, Benoit Hamon, as their standard-bearer. Hamon’s sudden, flakey appeal is his call for a universal salary of around 700 euros a month. A country already desperately mired in debt (98% of GDP, 2000 billion Euros) can dream, can’t it?

Now the fresh bets, which ranked Front National’s LePen, the independent left-liberal candidate Macron, and Fillon in ascending order of popularity, have been swept off the table. On January 24th the paper ‘Le Canard Enchainé’ published the first of a series of ‘revelations’ of FF’s having employed his wife at public expense over decades to the tune of a rough total of 900,000 euros, and his two grown children as legal aides, for lesser amounts.   Were this to occur in the US, it would be a) patently illegal and b) scandalous. In France, it has aroused a storm of outrage—which seems a little exaggerated, since political nepotism, while morally dubious, is legal and has been commonly practiced here for centuries (as it apparently also is in the UK.) The legal question in France is: did the wife and kids actually do any work? Because if Penelope did, she left few traces.

Fillon is now the first choice of less than 19% of the population. Free fall! Pundits and cartoonists liken him to the music conductor on the sinking Titanic. Buffeted by calls to step down and name a successor, he has refused. Who else is there with sufficient profile and experience to carry forth the sober, liberal, pro-EU program that recently made sense to so many voters? Yesterday Fillon gave a calm press interview aimed at expressing a partial apology, and putting matters ‘in perspective.’ He refers to a witch hunt, and there is no doubt that his anonymous denunciators, were they truly outraged, could have made an issue of the family employment deal years ago, rather than await this moment. My sense, watching the interview, was that he only partly succeeded, but that his numbers are likely to rise somewhat in the three months remaining, especially if he can refocus the debate on France, her economy, European cohesion, and the world. But rise by how much?


The whole cock-up is giving pdv, who was in the States during the gut-wrenching reversals of that election, a nasty case of deja-vu. Late hour devastating revelations from unknown sources? A less than perfect candidate? Note that like Clinton, Fillon has served in politics all his life, and one doesn’t come out of that without some spots on the white vest. Whereas, like Donald Trump, neither Macron, nor Hamon nor LePen have ever held executive responsibility.  In a political sense, they have no ‘past.’


A geeky note on polling, or, Tips on Reading Tea Leaves

There are some in France who, rolling an eye back toward the US campaign, consider the polls here suspect, biased, if not meaningless. Whatever their political preference, they hope for a last-minute Big Surprise.

There are those in the US who feel similarly about current polls that show various degrees of (lack of) support for Trump’s recent policy decrees. For pdv, digging deeper into these polls (CNN, CBS, IPSOS/Reuters, and Gallup) was instructive. Three points stand out.

The major prediction error in November 2016 did not come from polling via landline. (CBS and CNN use 50% landlines, and one is surprised that they find anyone at home. Gallup has cut to 30%, with the rest cell phones.) The real problem lay in a poor model for ‘likelihood to vote.’ In other words—a formerly apathetic demographic turned out in unexpected strength. At present, we are seeing opinion polls, not predictions.

For trends, compare apples to apples. The trend is what a lot of people are looking at, week to week. Each poll has its own methodology. Don’t set last week’s CNN (based on ½ landline calls and ½ cell calls) next to this week’s Reuters (online survey.)

Reduce noise. ‘Overall job rating’ is noisy. ‘Approve/disapprove travel ban’ is less noisy.

Be aware of sampling strategy. This is a big one—too big to wade into here. Note as an example that at least one poll samples Democrats and Republicans equally. Not quite a mirror of the electorate.

Be aware of method differences. The Ipsos/Reuters methodology stands apart for two reasons—it’s based (although not always exclusively) on online surveys, and it uses Bayesian credibility intervals rather than old-style statistical margin of error to describe uncertainty around the results.   Interestingly, their January result was tighter than other polls, with 47% disapproving of job performance, and 46% percent approving. Gallup at that time had 50% disapprove, 42% approve.

Go figure.  So to speak.

Thanks for staying with me so far.  A bientôt!


  1. A lovely post. Thank you. I especially liked the shot past Holland & Sherry as it reminded me of the Marais and finding Maison Freres. Like you, I’ve lived in many cities but the most beautiful is Paris. I think an equally exquisite city is Mexico City. Living there was magical.

    1. Thanks, Janet! Noted–I must soon find a road to Mexico City.

      1. A road or a plane? 😉
        (Still waiting for the japanese school uniform photos)
        A bientôt

      2. some fine day… needs a new place. Not here.

      3. Take your time. And meanwhile keep sharing your thoughts on this wonderful place we both love. Hopefully it will not sink…

    2. Hi Janet. Fancy meeting you here. 😉
      (Mexico city has lost a bit of its magic but some places, like Tlalpan, still hod it)

  2. Linda Corwin · · Reply

    Karen, your pictures are beautiful, thanks from a neighbor for posting them and for helping to give a perspective on the election here in France. The American one is too depressing to even think about and impossible not to. Do the French have their heads in the sand (still) this time around?

  3. Linda, thanks for your comment, compliments–and commiseration.

  4. Wonderful to read an explanation of current polling practices by someone who knows the subject, and great advice on how to “read the numbers.” Thank you!

    I do have to defend Wallace Henry Hartley, though, who instead of trying to save his own life, continued playing in an attempt to suppress on board the sinking ship… But it’s nice not to see cliches about rearranging deck chairs!

  5. sorry– meant to write “attempt to suppress panic.” We here in the States could probably use such accompaniment as we watch the chaos fomented by an incompetent.

  6. Willard Boepple · · Reply

    so interesting Kai. a question on the polling observations: how did you determine that the prediction error was in the modeling of the ‘likelihood to vote’ folk?

    1. Willard, apologies, I have woefully neglected my word-garden here. Now back in Paris with fresh digging tools. To answer your question: most of the polling orgs descibe their methodology–tales, well, a little digging. Trust you are well!

  7. margot livesey · · Reply

    Dear Kai,

    as always so witty and informative and frankly worrying. x Margot

  8. Thank you Maristed for your post. Titanic indeed… From a distance of 9,207 kms (just googled it) I cannot believe my eyes. And ears. Brexit? Trump? The latter put Berlusconi in perspective. Hamon? Though a fellow Breton he is basically in favour of “Demain on rase gratis” (Don’t know the english equivalent). Given the margins, he actually might come ahead in the first round. And then what? Hamon-Le Pen in the second round. A choice between la peste et le choléra?
    Polls? I’ve been in the market research business most of my professional life. Current problems may be due to methodology: landlines? Internet? Samples are probably not as representative as they should be. Maybe the questions have to be rephrased: in lieu of “Who will YOU vote for?”, switch to projection. “ten people you know how many do you think will vote for…” At any rate, when the final result is a flip of a coin, 50%/50% no way can we predict the result with a margin of error of +/- 1.7 to 3 points…
    Anyway thanks also for the (cold) beauty of Paris, which I miss so much. I hope that when I come back in the summer, France will still be around.
    Take care.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: