07 11 / 2012

Rasmussen polls were, once again, significantly biased in favor of Republicans.

An analysis of polling from 13 of the most important swing states in the 2012 presidential election found that Rasmussen Polls overestimated Republican Mitt Romney’s standing by 3.25%.

None of the 13 Rasmussen polls overestimated support for Democrat Barack Obama.

This 2010 piece by fivethirtyeight.com's Nate Silver about the performance of Rassmussen polls during the 2010 midterm elections found they too were biased:

Rasmussen’s polls were quite biased, overestimating the standing of the Republican candidate by almost 4 points on average.

Rasmussen didn’t do much better this year. Considered the lynchpin to the presidency for both candidates, Rasmussen’s final poll (11/4) from Ohio was off by nearly 2% in favor of Romney. In Michigan, the 11/1 poll overestimated Romney’s support by 2.9%. In New Hampshire (11/4), it was 3.4%. And in Iowa, Rasmussen’s 10/30 poll predicted a 1-point Romney victory: Obama carried the state by 5.6%.

Silver’s 2010 analysis suggests a few reasons why Rasmussen polls were statistically biased in favor of Republicans: the exclusive use of landlines when contacting respondents, and the failure to call back phone numbers if a caller doesn’t pick up (instead moving to a new number) were among them.

Silver also offered this:

Rasmussen also weights their surveys based on preordained assumptions about the party identification of voters in each state, a relatively unusual practice that many polling firms consider dubious 

In 2012’s election cycle, the first concern should have been alleviated: Rasmussen uses Pulse Opinion Research LLC for their polls, and their methodology page offers this:

To reach those who have abandoned traditional landline telephones, Rasmussen Reports uses an online survey tool to interview randomly selected participants from a demographically diverse panel.

During the entirety of 2012 campaign (not just near the end), Rasmussen polls often overstated Republican strength in comparison to other polls. Those less Republican-leaning polls often came under fire for a supposed Democratic bias, notably by Romney’s campaign itself.

Rasmussen polls have now significantly overstated Republican strength in two consecutive election cycles.

The 13 swing states analyzed, using Rasmussen’s last polls in those states ranging from Oct. 21 to Nov. 4, were Ohio, Florida, Virginia, Iowa, New Hampshire, Pennsylvania, Michigan, Colorado, Nevada, Michigan, North Carolina, Indiana, and Missouri. 

All polling data can be found at RealClearPolitics.com.