The presidential election is now less than 100 days away (92 to be exact, but who’s counting…), and people are now starting to pay more attention to the polls. Because there are more eyeballs reading these polls, pollsters are performing polls more frequently, and poll outliers are also becoming more frequent. For those that have paying attention to the polls for some time now, most would agree that the national polls have been fairly tight between President Obama and Republican nominee (although not officially the nominee yet), Mitt Romney. However, in the last few weeks, there seems to be an increasing number of polls with either (i) a systemic bias or (ii) its source media (e.g., CBS, New York Times, Wall Street Journal, etc.) failing to disclose to its readers the partisan identification of the poll participants. Why should a reader care?
By failing to disclose the party identification, readers are potentially misled as to the poll’s ability to predict election results if the election were held around the time of the poll (recall that polls are a snapshot in time). The following are examples of three recent polls where the media source failed to disclose the poll participants’ party identification in its article(s) describing its poll results. All three examples are cited because the party identification differences between Democrat and Republican is relatively significant, resulting in potentially misleading poll results.
- Most recent NBC news/Wall Street Journal poll (July 18 – July 22) has President Obama up by 6 points over Mitt Romney. This difference is certainly more than the average of polls over the last few months. However, the sample polled Democrats over Republicans by 11 points.
- Most recent Pew Research poll (July 16 – 26) has President Obama up by 10 points over Mitt Romney. This difference is well over the average of polls over the last few months. However, the sample polled Democrats over Republicans by 13 points.
- Most recent CBS/New York Times/Quinnipiac poll (July 24 – 30) of some battleground states including Florida, Ohio, and Pennsylvania has the President with an 6 point, 6 point, and 11 point lead respectively. The sample polled Democrats over Republicans by 9 points, 8 points, and 6 points respectively. Some poll analysts have compared these party differences to the party differences in the 2008 and 2012 exit polls and reported that the party differences are less than this poll results.
Because every state does not register voters by party, reliable data for the percentage of registered Democrats to registered Republicans is scant. Some pollsters estimate that Democrats may have a 3 or 4 point advantage over Republicans; in 2008 pollsters contend the advantage was higher at approximately 7 to 8 points, and the chance that this relatively significant difference would have widened since then is not likely. Thus, polls with large party differences may not be indicative of election results if the election were held around the time of the poll.
These large party differences may in fact just be a result of the random sample. However, these differences should be disclosed to the readers to help prevent any misleading inferences from the polls’ results.