Even this writer predicted that the referendum to overturn Maine's law legalizing gay marriage would fail. And most agreed. Pollster.com had "No" to Proposition 1 leading going into Election Day; and Nate Silver of fivethirtyeight.com created a model, based on several variables, likewise predicting that Prop 1 would fail.
After the election, Harry Enten '11 went back and looked at Mr. Silver's model. With a lot of data mining and some top-notch statistical skills, Mr. Enten created a linear regression model to predict the outcome of referenda concerning gay marriage much more accurate than that of Mr. Silver. To quote Mr. Enten:
The average difference between the model's predicted support for an amendment in an election and the actual support for the amendment was 2.69% (compared with Silver's 4.46%). Importantly, this difference was greater than 2.00% in only 4 instances (Michigan 2004, Montana 2004, North Dakota 2004, and South Dakota 2006) and greater than 4.00% in only two (Michigan 2004 and North Dakota 2004) [of 25 total observations].
In other words, this model is quite good.
Because of his work and the impressiveness of the model, Mr. Enten was picked up by Pollster.com as a guest pollster. If you're curious about what went wrong in Maine, and how we can best predict races like this in the future, I suggest you check out his entry, and keep an eye out for more of his work in the future.
(For an example of a predictive model gone bad, check this out.)