In the first article we have spent on this blog in the U.S. presidential election and its relationship to the polls, we made reference city, like many other observers, the site of synthesis of the various surveys in the United united by Nate Silver, FiveThirtyEight blog.
This site, which already gave Obama the winner a week ago, confirmed and even strengthened its forecasts, did not hesitate to say that the strong support% of the popular vote and 303 electoral votes (against 234 for his rival) Obama has a 80% chance of winning the election! While these results are confirmed, but to a much lesser extent, the site Realpolitics, which gives 201 electoral votes for Obama, 191 for his opponent and 146 undecided. The nuance is.
But Le Monde today tells us that part of the electoral debate in the United States these days focuses on the following question: but who is this mysterious Nate Silver, whose comparative surveys blog seems to make more most authoritative among observers?
If the issue is political in the United States, the French point of view lead to also ask in legal terms, and we also noted this problem in our previous article as an issue under the law of holes: understanding how these site of synthesis of the various surveys? Are they reliable, are they controlled what to think from the point of view of democracy? This is a very exotic for those interested in France, legal surveys problem …
The question must be asked: while national polls show Obama and Romney neck-and-neck, the assurance of Nate Silver has something dismounting. In any case, the controversy swells, as suggested by the article in Le Monde, which as “Nate Silver, the young statistician promoted political prediction guru.”
Nate Silver is a statistician for 34 years, known for having developed a software for baseball statistics and reporting with games that can amaze and not unrelated to the latter, since sportsbooks also give Obama paris winner of the election … But what exactly do we?
Nate Silver’s model built based on the following methodology: it compiles the survey of 136 institutes U.S. survey (53 active), in fact the average, then the balance in the polls New York Times / CBS News and Wall Street Journal / NBC News, which include more quantitative than daily series elements. Then he adds in these surveys ten or twenty surveys by five to ten institutes in different states, especially in the swing states, which are vitally important for the election in terms of electoral votes.
Rigorous model or kitchen sondagière implausible? Very attacked by Republicans, who have questioned his political impartiality, Nate Silver, who correctly predicted the outcome of the election in 49 states in 2008, was defended by the Washington Post, considering this model as a simple aggregator surveys. Thursday, November 1, Nate Silver came on the show “Morning Joe” on MSNBC to defend his model: the exchange ended with a $ 1,000 bet on the accuracy of the numbers, as it needs to wait until November 6 to know. The game still ….
However, Nate Silver is wrong or not, the problem remains. Indeed, disgruntled Republicans will not fail to implement against attack: thus they have created their own site, the site Unskewedpolls who therefore seek to establish the truth … straightening the polls in a much more favorable to Romney senses.
No bad puns for the Republican candidate, we can ask: which way to turn?
In any event, the controversy and more generally the debates on websites synthesis of surveys in the United States demonstrate two things, including the French point of view:
On the one hand, those who believed that the use of surveys was problematic in France saw nothing. Further study of the U.S. system is more essential than ever. And there is clearly more work.
On the other hand, the French administrative system, ensuring control from a legitimate independent authority under the law, and statistical and legal competence recognized, may be subject to legitimate scrutiny of observers is probably preferable to such a model of self-regulation.