Note the work of Serge Galam that recently try to explain the tied election results http://www.citeulike.org/user/xsongx/article/2506399
the factr seems to be "contrarians", people who in discussion about politic oppose the majority (whatever it is). It is the opposite of the classical "followers", who in discussion rather follow the majority,and instead of tight results, lead to strong polarisation... Serge Galam propose physics-like models of opinion transition (similar to spin networks). His previews "followers" model make him think that if a campaign goes long, the result only depend on the respective "stubborn" population, of people that are unable to change opinion when facing opposite arguments. at the end of a long campaing, if one camp have more "stubborn", it will will. "contrarians" on the opposite avoid the polarization... so contrarians and subborn determine the result,.... note that this kind of models are not new. I remember a similar model for racial ghetto emergence in cities.. 2012/11/7 Jed Rothwell <[email protected]> > As I expect I have made clear, I am fascinated by public opinion polls. > Even though this is off-topic, here is another interesting historical note > about the subject. > > My late mother was something of an expert in polls. She was director of > public opinion research at the US Census Bureau. > > Everyone familiar with polling history knows that the 1948 presidential > election was miscalled by every pollster. People often point to this even > today when they wish to discredit or downplay the significance of modern > polling. They say, "you never know who's going to win because look at 1948 > and the famous newspaper photo of Truman holding a newspaper saying 'Dewey > defeats Truman.'" > > My mother was beginning her postwar professional career in 1948, after > World War II work with the Army and others. See Public Opinion Quarterly, > Vol. 33, No. 3, Autumn, 1969: > > > http://www.jstor.org/discover/10.2307/2747532?uid=3739896&uid=2129&uid=2&uid=70&uid=4&uid=3739256&sid=21101260869333 > > Anyway, regarding 1948 she said: "Believe me, nothing like that will ever > happen again. The pollsters learned their lesson." > > In business we would say that was a "wake-up call." It was like Ford's > mismarketing of the Edsel. It became the foundation of modern poll taking. > So don't imagine that pollsters are guessing or that they might be wrong by > 5% or 10%. People are unpredictable individually, and their individual fate > is unknowable. But, taken *en mass* you can make reliable predictions > about people. With polls and actuarial tables we can characterize groups of > people. That is why life insurance companies make a profit. You may find > that depressing but it is a fact. > > Comparing modern methods to the techniques used in 1948 comparing an > estimate made with a slide rule versus a supercomputer. That is literally > true. People used slide rules back then. My mother liked them and used them > to the end of her life. She said they were good for your mind because you > had to remember where the decimal point is. That kept you on your toes. > Plus, the fact that they worked was a good reminder that numbers are seldom > significant beyond two or three digits anyway so why worry about precision? > > Accuracy matters. Precision, not so much. If you don't know the difference > please don't write science papers. > > - Jed > >
