The actual results from last night -- at least on the presidential level -- provide compelling empirical evidence negating all three of those theories. In the top 10 swing states, compare the predicted result from the final RealClearPolitics average to the actual results in those states:
Swing state RCP Final avg. Actual result Actual result v. poll
Ohio Obama: 2.5% Obama: 4.1% Obama: +2.6%
Florida Obama: 1.8% Obama: 2.5% Obama: +0.7%
Pennsylvania Obama: 7.3% Obama: 10.3% Obama: +3.0%
Virginia Obama: 4.4% Obama: 4.5% Obama: +0.1%
Indiana McCain: 1.4% Obama: 0.8% Obama: 2.2%
North Carolina McCain: 0.4% Obama: 0.3% Obama: +0.7%
Missouri McCain: 0.7% McCain: 0.2% Obama: +0.5%
Colorado Obama: 5.5% Obama: 6.7% Obama: +1.2%
Nevada Obama: 6.5% Obama: 12.5% Obama: +7.0%
New Mexico Obama: 7.3% Obama: 14.9% Obama: +7.6%
With two exceptions (Nevada and New Mexico), the polls were extremely accurate in predicting the ultimate results. And in all 10 swing states, Obama outperformed what the final polls predicted, meaning that there ended up being a better result in counted votes for Obama than the polls anticipated.
http://www.salon.com/opinion/greenwald/2008/11/05/polls/index.html