segunda-feira, abril 23, 2012

The polls in France

The last French polls did well, by comparative standards at least. In 2007, no pollster managed to estimate the vote for the first four candidates with an average absolute error below 2 points. In some cases, like in the CSA last poll, things were even worse, with Sarkozy's vote underestimated by almost 6 points and Le Pen's vote overestimated by more than 6 points. Things went better this time: Hollande and Sarkozy did slightly better than expected, and the major problems came with Le Pen and Mélenchon doing, respectively, better and worse that what the polls were suggesting, but still with deviations that were not so large as the ones that occurred in 2007.

A pollster-by-pollster look suggests that, in most cases, whatever may have been behind the over- or under-estimation of candidates' results was common to all pollsters: Le Pen was consistently underestimated, Mélenchon was consistently overestimated and, with few exceptions, the underestimation of both Hollande and Sarko were also common (and relatively small).

Things continue to look bleak for Sarkozy. In 2007, for the 2nd round results, Sarkozy got a majority of the votes in almost every single poll published since the beginning of the year:

Now, Sarkozy faces the exact opposite situation: he has been behind Hollande in every single poll published since at least February, and the polls conducted yesterday place him behind Hollande at least 6 points.

2 comentários:

Unknown disse...

Electoral maps and exit polls tell us the same story: Marine Le Pen's electorate was poorer, more working-class/blue collar (CSP-), less ideological, younger, less Catholic and more concentrated on traditional left-wing areas than ever. Gauche-lepénisme at full force. No wonder she won the department of Gard, a Communist stronghold. Why is this important? Traditionally these FN the constituencies tend to sit out or even vote for the socialist candidates in the run-off at a much higher rate. Therefore any scenario of the FN vote coalescing around Sarko is now more unlikely than ever. Socio-economic demographics is destiny in this case.

Looking at a Marine 2012 vs. Jean-Marie 2002 vs Jean-Marie 2007 map is fascinating. Sarkozy retained most of the CSP+, upscale, traditionally right-wing voters he regained 5 years ago; the blue-collar voters happily went back to Le Pen (and were joined by some of their peers who had been sticking with the left).

Let's face it: barring a total screw up by Hollande (it's still the French left), Sarkozy is screwed. He hasn't lead on a poll since 2009. Stick a fork on him.

Pedro Magalhães, alguma opinião sobre o que possa ter acontecido ontem em Alberta?

Foi um rotundo falhanço das empresas de sondagens:

Parte foi voto estratégico por parte dos eleitores de esquerda que votaram PC para impedir uma maioria da Wildrose.

Isso não explica tudo: a Wildrose teve menos 6% do que qualquer uma das sondagens dos últimos dias sugeria. Houve uma sondagem que capturou algum movimento, mas nada nesta ordem de magnitude.

Há muito que não via um falhanço destes por parte dos pollsters.

Pedro Magalhães disse...

I think it's an excellent point about the unlikelihood of Sarkozy collecting the FN vote. More evidence on that:

On Alberta, I know next to nothing. I'll have a look.