Now that the election is over and we aren't hearing about polls hour by hour, here's the result of a study on which polling group was most accurate.
Which poll was dead last? Surprise! Newsweek, followed closely by CBS/NYT's.
Which were best? Rasmussen and PEW Research
Two things: I notice the last two ranked are among the oldest. Newsweek was, in fact, easily the oldest poll having been completed a full 12 days prior to the election.
Second, this doesn't include sites like RCP, 538 and electoral-college.com that use all the other data to do an averaging using various methods each has devised. Don't know about RCP, but electoral-college and 538 were pretty much dead on. In the last few days I think both flipped Indiana to red (as did Intrade) but both called it as 364 electoral votes for BHO. They also variously moved Missouri from blue, to white to red and back to white. I think in the end they both had Missouri blue. Intrade also had 364 EVs with Missouri blue and Indiana flipping red in the last 24-48 hours.
538.com saw the average national polls at about 7%, called that a bit high and reasoned a 6.1% Obama national win (final was 6.5%). I think all three of these sites do a better job than any one of the polls.
Posted by: Rand | November 10, 2008 at 05:13 PM
I did the best predicting the Dunstan/Patton race back in June on the "Partisan Discourse" Blog.
I was almost dead on and Ron didn't believe me.
Posted by: Wayne | November 11, 2008 at 11:17 AM