CBS News' guide to understanding political polls
(CBS News) As the 2012 presidential election season heads into its final weeks, expect to see plenty of polls on the state of the race. Here are some explanations, guidelines and information to help navigate the plethora of polls that will be released in the run-up to November 6.
Many of the polls conducted so far this year have been among registered voters. Those are simply people who are registered to vote in their state, regardless of party affiliation or their likelihood to vote in November.
Beginning generally in September, after the party conventions, polls will start to report the results of the race among likely voters. In some cases, "likely voters" refers to voters who tell the pollster they are "very" or "somewhat" likely to vote in an upcoming election, but many polls use a more complex process. CBS News, like many other pollsters, uses a statistical formula that incorporates responses from a variety of questions to identify who is likely to vote. When comparing results from different polls, it's important to note whether the results are reported among registered or likely voters.
Tracking polls are daily measures of public opinion and use a unique methodology that adds some new poll respondents and removes some respondents every day. By their nature, tracking poll results may fluctuate daily. Most tracking polls are conducted in the week or so prior to Election Day; the Gallup Daily Tracking Poll is a notable exception.
The order in which poll questions are asked can affect the responses to questions. Some polls ask the "horserace" question (which candidate the voter supports) at the beginning of the questionnaire, so the vote question isn't affected by other questions, while others ask it further into the questionnaire. This can sometimes be a reason why polls show different results.
The margin of error describes the sampling error in a poll. On a sample of 1,000 respondents, the margin of error is +/- 3 percentage points, which means that 95 percent of the time, the poll results fall within three points in either direction of the results that would be found if everyone in the population were interviewed for the poll. In election polling for two candidates, the margin of error applies to the percentage supporting each candidate in the race. In other words, if the margin of error is three points, then the difference between the candidates needs to be six points to be outside the margin of error.
If two candidates are nearly even in their support, or within the margin of error, then the race can be characterized as close. The phrase "a statistical tie" only applies if both candidates have the exact same percentage - and even then, because of sampling error, it's probably better to call the race close.
When looking at any pre-election poll, it is critical to know who conducted the poll. Campaigns do a lot of polling and have a stake in how their candidate is presented in the polls. Polling done by organizations on behalf of political parties, advocacy groups or other organizations with a stake in the outcome of the poll should be considered carefully.
A good rule of thumb is that well-done polls are transparent about who was contacted as well as how and when their polls were conducted. A good poll should disclose the dates of interviewing, the population being represented, details about how the poll was conducted, the full question wording of all the questions asked in the poll and poll results.
Two polling industry organizations have taken steps toward greater transparency in polls. The American Association for Public Opinion Research, a professional organization for pollsters, launched its Transparency Initiative in 2009. (CBS News is an official supporter of the initiative.) The goal of this program is to encourage "routine disclosure of methodological information from polls and surveys whose findings are released to the public." The National Council of Public Polls (of which CBS News is a member) has also publicly committed to disclosure of its members' polling methods, questions and results.
The National Council on Public Polls has a helpful brochure, "20 Questions A Journalist Should Ask About Poll Results," that provides information useful to anyone seeking to better understand polls.