It's a secret trick of the news business. Nothing interesting to report during an election campaign? Commission a poll! Congratulations, you just bought yourself a front page story.
Polls can have news value: they help voters understand trends and feel the direction of political winds, but not all polls are created equally, even if the media tends to report on them as if they were.
Perhaps the most common sin by news organizations is to report on marginal shifts in polling data as if they were substantial. But if Candidate A is 6 points ahead of Candidate B in a poll with a 3 point margin of error, that's actually a tie, a statistical dead heat.
The key is in knowing the critical factors that determine if a poll should be believed. A “no” answer to any of the following questions would call into question a poll’s reliability.
- Was the poll conducted by a credible organization?
- Was the poll conducted within the past 10 days?
- Were “likely” voters polled?
- Were landline and cell phone numbers of poll respondents randomly selected?
- Was the sample size at least 1,000?
- Was the sampling error no greater than plus or minus three percentage points?
The Texas Observer's Abby Rapoport just published a great Polling Survival Guide to help voters navigate the murky waters of opinion surveys.
How pollsters choose to write their questionnaire can have major implications on the results. The American Association for Political Opinion Research offers some key warning signs for unreliable questions. First there's the order in which you ask the questions. If the poll begins asking which candidates someone will vote for, it risks favoring the most well-known candidate: the incumbent.
Our favorite polling sites include Real Clear Politics, which aggregates polling numbers, hopefully to give a more accurate reflection of voter sentiment. We also like the Texas Tribune, and not just because we're reporting partners with them. The Trib includes reports on others' polls, and runs a blog on the challenges of conducing opinion surveys.