Sunday, March 25, 2012

A Note on House Effects in Polling

I don't have the time or inclination to examine polling differences systematically myself, but those of you interested in polling techniques and possible systematic statistical biases of one method or another may find this post interesting:
Ce qui me frappe est que les écarts concernant Sarkozy sont relativement bien distribués entre sondeurs, sans qu'un impact de la méthode employée apparaisse nettement. Par contre, pour Hollande, l'influence de la méthode semble claire : les 3 sondeurs qui procèdent par Internet voient tous le candidat socialiste plus bas que la tendance et sont en outre plus bas que les 4 autres qui sondent par téléphone. Le constat est le même en prenant les sondages depuis la primaire socialiste (64 sondages au lieu de 44) et la tendance pondérée sur les 20 derniers jours.
Cela ne veut pas dire qu'une méthode de sondage soit nécessairement meilleure que l'autre (le téléphone est par contre incontestablement plus cher). On peut juste dire qu'elles semblent donner des résultats différents, pour des raisons qui ne sont pas évidentes : après le "shy Tory factor", y-a-t'il un "shy non-Hollande factor", c'est-à-dire que certains sondés n'oseraient pas reconnaître par téléphone qu'il ne voteront pas pour Hollande (au moins au premier tour) mais le font plus volontiers si on les interroge par Internet?

Bayrou Relaunches

François Bayrou, whose campaign has been struggling, held a major meeting today at the Zénith. Bayrou invariably makes me think of the Mugwumps. For those who don't know American political history, Mugwumps were Republicans who supported the Democratic presidential candidate, Grover Cleveland, in 1884 against the Republican James G. Blaine, whom they considered corrupt, repellent in character, and morally deficient. Bayrou and his MoDem haven't quite crossed over to the left, but it's not out of the question that they would join a left government after the election.

They share with the Mugwumps a certain sanctimoniousness, especially when it comes to fiscal stringency. So, for example, Bayrou proposes to reduce the deficit by €100 billion, with €50 billion to come from spending cut and another €50 billion from increased taxes. There is already something suspect about the neat symmetry of cuts and taxes, and the lack of specificity about which cuts and which taxes is also characteristic.

At the moment, Bayrou is drawing around 12-13% in first-round polling. At this stage in 2007 he was at 18% . In that year he benefited from anti-Ségolène feeling on the left. Hollande seems to be more acceptable to the left Mugwumps, then, but one might have expected the left of the UMP, social liberals of the right, to flee Sarkozy for a candidate deemed to be less mercurial and "abnormal." Apparently there aren't enough of these to make up for the desertion of the left component of Mugwumpery.

Increasingly, European politics, and not just in France, has become a politics of the marais, as one used to say during the French Revolution. The left and right extremes reject the broad central consensus in favor of the EU and globalization and go their separate ways. In the center there is a substantial majority of voters, roughly 60-65% of the electorate, divided between a party which, for historical reasons, is nominally of the left and another party which, for similar historical reasons, is nominally of the right. These parties divide sociologically  along a number of dimensions (private/public sector employment, age, work/non-work status [retirees favor the right], wealth [richer is righter], urban/rural residence, etc.), but these differences lack the deep identity component of the past. A great deal of emotion therefore settles on personalities: Who is tougher? Who is steadier? Who is more predictable? Who is more tolerant/intolerant of differences? Who is more honest or corrupt?

These are Mugwumpish choices. The grand differences regarding projets de société are a thing of the past. The French language needs a translation of Mugwump. Maraîchers?

Android Poll App

If you're a French political junkie with an Android phone, you'll want this app. I've got mine, and the remarkable thing is that all the polls agree at this point that Hollande will win the second round 54-46. This margin has remained fairly steady for some weeks despite considerable ups and downs in the first-round polling of particular candidates. At this stage in the race, an 8-point margin would seem to be a fairly comfortable lead for Hollande,and the rolling IFOP poll, which includes some post-Toulouse massacre data, shows no surge for Sarkozy. Next week's polls should be more revealing on this score. But for now, if I were a betting man, I'd give odds for Hollande.

But get the app! It's the way to settle any arguments you might get into at the Café de Commerce about who's up and who's down.