A week before the election I wrote a post questioning whether Rasmussen was the most inaccurate swing state pollster in the nation, based on its consistently bad predictions in the 2008 presidential election.
If you’ll recall, Rasmussen’s 2008 predictions in 11 battleground were heavily skewed toward McCain, with the average “skew” at 4.08%, compared to the actual results. Rasmussen also called four states wrong, and five of its predictions were outside the margin of error, which left Rasmussen with a reliability rate of 55 percent.
Believe it or not, Rasmussen’s performance was WORSE in the 2012 election. Their polls were skewed toward Romney by an average of 4.28 percent. Rasmussen called 5 of 11 states wrong, and they blew the margin of error in 6 of 11 states, for a reliability rate of 45 percent.
Here are those numbers below. Highlighted in blue are the links to the Real Clear Politics results for the state in question. I’ve highlighted in red the states in which Rasmussen wrongly predicted the outcome. In the state where Rasmussen blew the margin of error, the skew is highlighted in green.
Colorado: Obama +4.7; Rasmussen, Romney +3 (skewed Romney by 7.7%)
Florida: Obama +.9; Rasmussen, Romney +2 (skewed Romney by 2.1%)
Iowa: Obama, +5.6; Rasmussen, Romney +1 (skewed Romney by 6.6%)
Michigan: Obama +9.5; Rasmussen, Obama +5 (skewed Romney by 4.5%)
Nevada: Obama +6.6; Rasmussen, Obama +2 (skewed Romney by 4.6%)
New Hampshire: Obama +5.8; Rasmussen, Obama +2 (skewed Romney by 3.8%)
North Carolina: Romney +2.2; Rasmussen, Romney + 6 (skewed Romney by 3.8%)
Ohio: Obama, +1.9; Rasmussen, tie (skewed Romney by 2.1%)
Pennsylvania: Obama +5.2; Rasmussen, Obama +5 (skewed Romney by .2%)
Virginia: Obama +3; Rasmussen, Romney +2 (skewed Romney by 5%)
Wisconsin: Obama +6.7; Rasmussen, tie (skewed Romney by 6.7%)
If you look at the RCP polling numbers in “tossup” Senate races, you’ll also note a consistent Rasmussen “skew” for the GOP candidates.
What conclusions can we draws from this?
- First, you’re nuts if you believe a Rasmussen state poll in a presidential “battleground” state. Fox News apparently came to this conclusion after the 2008 election (they dropped Rasmussen for 2012).
- Second, when you read a Rasmussen poll for one of the above, you’re best off if you deduct a “house edge” from the Republican candidate. That varies widely, but based on the above, you’d be fairly safe if you knocked off 2 points, maybe more, from whatever it said the Republican was polling. The only state they didn’t skew by that much in the 2012 contest was Pennsylvania, which Rasmussen more or less nailed.
- Third, Rasmussen needs to adjust its model of who’s going to vote, at least for presidential races. What they’ve got now appears useful only for fooling Republicans into believing they’re doing better than they actually are. And as we have seen, that doesn’t help the GOP at all.