Over the last week, four new national polls have been released on the 2012 presidential race. As seen below, the polls vary widely. At the two extremes, there is an 18-point swing between the Rasmussen Reports poll showing Mitt Romney ahead, and a Bloomberg poll showing President Obama with a large lead. The following article attempts to go beyond a mere listing of polls by providing some analysis as to why each poll might be credible or not credible. Below, a breakdown of each poll is provided based on their sampling, performance in the 2008 presidential election, and any pollster ratings that are available.
According to a simple average of all the polls President Obama has retaken the lead over the last week. A Real Clear Politics average of 9 different polls over the last three weeks has President Obama leading Romney by a margin of 2.2 points (46.6% to 44.4%), compared to a 0.7 point lead at the beginning of this week. An average of the five polls released in the last week has Obama ahead of Romney by 2.6 points (47.6% to 45.0%). Earlier this week the most recent polls had President Obama trailing by one point.
As new polls are released, I will be providing continual updates, posting each poll along with some analysis on each pollster’s past reputation and performance. To receive more updates, follow me on Twitter or Facebook. I will also provide analysis on the latest trends.
Poll of 1500 “likely voters” taken from 6/20/2012 to 6/22/2012
- Romney 48%
- Obama 43%
Analysis: Rasmussen has Mitt Romney performing the best of any pollster against President Obama, with Romney’s a five-point lead over the sitting President. At the beginning of this week Rasmussen had Romney with a three-point lead, so Romney’s lead is expanding according to their data.
Rasmussen’s tracking uses a “likely voter” sampling method, which excludes some registered voters. In the past, Rasmussen has been accused of producing biased results that tend to favor Republican candidates. Rasmussen does not include cell phone users, but attempts to compensate by using an “online survey tool to interview randomly selected participants from a demographically diverse panel.”
In 2008 Rasmussen’s final poll matched up well with the final result in the race, and its polls in the final month were fairly consistent. Rasmussen uses a sample that has “baseline targets” of 35.8% Republicans, 33.0% Democrats, and 31.2% “unaffiliated” voters, despite other surveys that show Democrats’ actually outnumbering Republicans in the general population.
Poll of 3050 registered voters taken from 6/14/2012 to 6/20/2012
- Romney 47%
- Obama 45%
Analysis: At the beginning of this week Gallup had Romney with a one-point lead, so this pollster has Romney’s lead staying relatively the same.
When first released, the Gallup tracking poll swung widely, giving Romney the lead and then giving Obama a large lead within a matter of just a few days. Since that time, Gallup has expanded the date range of its tracking poll, which may make the results more consistent.
Gallup has, by far, the largest sample size, with 3050 registered voters, which theoretically should make their poll more accurate. Gallup’s polls varied greatly in the 2008 race, at one point having McCain ahead by ten points before producing a final prediction with Obama up by eight points. Gallup has not released the internal breakdown of their sampling. Gallup was rated above average by Nate Silver in 2010.
Poll of 734 likely voters taken from 6/15/2012 to 6/17/2012
- Obama 53%
- Romney 40%
Analysis: Bloomberg had the “shock poll” of the week, showing President Obama with a 13-point lead even though other pollsters showed a close race. Many called Bloomberg’s poll an “outlier” and expressed skepticism about the accuracy of Bloomberg’s results. Bloomberg issued a release to defend their data.
As noted by Noah Rothman at Mediate, Gallup and Rasmussen polls have come under criticism for not including enough minorities, something that is not a problem in the Bloomberg poll. In addition, Bloomberg’s poll was done by a reputable organization that used scientific methods. The poll should be viewed skeptically, as any poll that strays this much from the mean should, but it cannot be entirely dismissed.
Bloomberg used Selzer and Co. to obtain its data. In 2010 Nate Silver rated Selzer and Co. well above average compared to other pollsters. There is no public data available to see how Selzer and Co. performed in the 2008 presidential race.
Poll of 878 registered voters taken from 6/14/2012 to 6/18/2012
- Obama 47%
- Romney 44%
Analysis: The Associated Press poll falls within the two extremes of the Rasmussen and Bloomberg polls. The poll includes all registered voters, which may improve the numbers for President Obama.
In his June 2010 ratings Nate Silver placed AP/GFK well above average, and ahead of both Rasmussen and Gallup. In 2008 the AP largely relied on polls from Gallup, so there is no data to analyze how the AP/GFK performed as a separate organization during that time.
Poll of 1563 registered voters taken from 6/7/2012 to 6/17/2012
- Obama 50%
- Romney 46%
Analysis: The Pew poll includes a large number of registered voters and was taken over a long time period. The inclusion of all registered voters likely aids Obama. The length of time over which the survey was taken may even out data more than other polls like Bloomberg.
In June of 2010 Pew was rated well above average by Nate Silver, just above Gallup but below Selzer and Co. and AP/GFK. In 2008 Pew performed very badly in terms of consistency, with their results varying widely in the month before the election. However, Pew scored well in terms of accuracy, with its final poll coming very close the actual Election Day results.