Polling

As the Obama/Romney head-to-head match up develops, a new poll (or several new polls) comes out nearly every day. With each new poll articles from political trend watchers and journalists alike appear predicting alternatively good things or trouble for one of the candidates. These articles tend to beget more articles on whatever trend the poll may have identified as well as coverage on the daily television news programs. This generally goes on for the length of the news cycle; that is, until the next poll comes out a day or two later.

The astute political observer, however, must approach these polls with caution and know when and where to recognize an outlier from a poll that may have real relevance. Nate Silver at 538 has a great, in-depth piece on how to watch polls in the months leading up to November. It is recommended reading for anyone with an interest in the 2012 races, particularly the Presidency. What follows is my concise reader's guide to polling analysis.

Polling organizations inherently carry some kind of bias; they poll certain groups of people, sometimes more from one side of the political spectrum than the other. The manner in which the polling questions are posed can also affect the result. A poll that begins with a negative premise built into the questions asked can expect a more negative response than questions parsed more moderately. These inherent biases mean that some polls give results that fluctuate considerably from the established norms of polling on a certain topic.

Therefore, it is critical to look at  averages of several similar polls to get a more accurate reading on what direction the political wind is blowing. Aggregate polling information is easily available from many sources in the internet age, RCP, for example, posts daily numbers that combine information from several polls in several categories (i.e. approval rating; Obama v. Romney; likeability, etc.). Looking carefully at polling information can help identify outliers and tone down excitement over surprising numbers that may, in fact, mean very little.


The very recent WMUR Granite State Poll in New Hampshire is a good example. It shows Obama leading Romney 52 to 41 among likely voters and is a good example of  potentially misleading polling information. The numbers in the WMUR poll cannot be compared against an average because no other similar polling information has been gathered recently. Furthermore, the national press and television media tend to draw national conclusions from small state based polling. These kinds of conclusions are, on their face, erroneous. The value of the WMUR poll in N.H. is specific to N.H. and should not be extrapolated to the national electorate. I suspect that the Obama/Romney race in N.H is, in fact, much closer to national averages in polling that show President Obama leading by 4 points than to the 9 point lead he carries in the WMUR poll. As an aside, in the same poll voters asked if they approved of Obama's job performance came in at 50% approval while 47% did not approve.

In addition to looking at aggregate polling information, there lies the issue of polling trends in specific demographic groups. Examining polling in a particular group without context can be misleading.  Jonathan Bernstein recently demonstrated the importance of comparing the specific group trend with that of other groups: A candidate who loses popularity within specific group may well see that reflected in polling because he is losing popularity in several groups or the voting public as a whole. For example, if Obama is losing ground with Latino voters he is probably losing ground with the electorate as a whole. This is not to say that there are not isolated group polling trends, but that they are less common than may be represented in media outlets.

In summary, by focusing on these two areas, aggregate polling and trends in groups v. overall trends, the reader can perceive a more accurate and balanced picture of how the coming election is shaping up. Additionally, it can help the observer know when to view a poll, and the popular reaction to it, with skepticism.