A Bloomberg poll released June 20 has President Barack Obama polling well ahead of his Republican challenger, Mitt Romney.
Among likely voters polled, 53 percent stated they would vote for Obama in November, and only 40 percent stated they would vote for Romney.
While most of the polls conducted thus far in June have had Obama edging Romney, this 13 point margin is unprecedented, since five other polls have placed Obama’s lead between one and four percentage points.
Is Obama on his way to a major polling surge, or was the Bloomberg survey flawed?
Bloomberg Poll Methodology
The poll was conducted by Selzer and Company, on behalf of Bloomberg National News. A pool of 1002 respondents were contacted between June 15 and June 18, 734 of whom identified as likely voters.
While some poll questions were presented to the entire sample, the “for whom would you vote” question was only presented to the smaller sample of likely voters. Respondents were contacted by live interviewers via landline and cell phone. Results were weighted for age and race to be consistent with recent census data. The margin of error for the smaller sample of likely voters was +/-3.6 points.
Was the Poll Flawed
Although some conservatives in the blogosphere are quick to assume the poll must have been biased, pollster J. Ann Selzer points out that the methodology and sample were similar to a Bloomberg poll conducted by Selzer and Company in March, which had the candidates tied. Both polls had demographics that closely mirrored U.S. Census data. The samples were, however, broken down to include only likely voters, so it is possible that the demographics shifted, as information about the sub-sample has not been released. Polling expert Mark Blumenthal concludes that the high margin for Obama should be questioned, unless other polls start showing a similar spread. He does not, however, suggest that the methodology was flawed. The most likely scenario, according to Blumenthal, is that the poll was a statistical outlier.
Polling Science: What is an Outlier?
Most mainstream polls, including the Bloomberg poll, have a 95 percent level of confidence. Blumenthal points out that this leads to 1 in 20 polls statistically falling outside the margin of error, even with methodology that follows good polling practices. The poll may have been conducted during a time when favorability for Obama had temporarily shifted due to events in the news. The cause is more likely statistical, however, since the poll was conducted over several days and there were no major news stories that might have caused a spike in Obama’s polling.
Interpreting the Poll Results
Polling expert Nate Silver with the New York Times suggests that this poll should not be completely disregarded, even though it is inconsistent with other polls conducted over the same period of time. Silver says the pollster has a good track record, and the data is valuable when averaged in with other polls to gather aggregate data. Statistically, there will be more outlier polls conducted between now and November that may sway the balance in the other direction. “You should not throw out data without a good reason. If cherry-picking the two or three data points that you like the most is a sin of the first order, disregarding the two or three data points that you like the least will lead to many of the same problems,” said Silver, in his coverage of the poll, before pointing out that unknown pollsters, or those with poor track records, are exceptions.
Aggregate polling data currently has Obama with a slight lead over Romney. His lead varies between 1 and 3 points depending on whether the Bloomberg poll is factored in.
RealClearPolitics. General Election: Romney Vs. Obama. (2012). Accessed June 21, 2012.
Blumenthal, M. Bloomberg Poll Shows Big But Questionable Obama Lead. (2012). Huffington Post. Accessed June 21, 2012.
Silver, N. Outlier Polls are No Substitute for News. (2012). New York Times. Accessed June 21, 2012.
Decoding Science. One article at a time.