ABSTRACT
The 2012 presidential election turned out to be strikingly predictable. 1 This was all too clear at the end of the campaign, when various “quants” successfully predicted the result in 49 or 50 states (see, e.g., Bartlett 2012). The most celebrated was Nate Silver of the New York Times blog FiveThirtyEight, but there were numerous others, including Simon Jackman at https://Pollster.com" xmlns:xlink="https://www.w3.org/1999/xlink">Pollster.com, Drew Linzer of Votamatic, and Sam Wang of Princeton Election Consortium. Much of their success at the end of the campaign was due to the pre-election polls themselves, with basic RealClearPolitics averages of poll results from the last few days correctly predicting the winner in 49 of the 50 states, narrowly missing (by 0.3 of a percentage point) Florida. However, there is value in taking seriously how to combine polls from different survey organizations, or “houses” (Erikson and Wlezien 1999; Jackman 2005).
