The Internet has been a boon for pollsters — everybody's results are touted worldwide seemingly instantaneously. But for observers it sometimes leads to confusion. How can the AP be calling the race as near dead even, when CNN has Obama with a significant lead at the same time? Well, poll experts say that it has little to do with strict numbers and everything to do with who you ask...and what you ask them. The same pollsters publish very different numbers for samples of registered and "likely" voters (those who have voted regularly in the past.) But it's even more complex. For example, some polling organizations use cell phone numbers in their samples, some don't. Cell phone users tend to be younger and more affluent — factors that also play into voting patterns.

The Internet is a great place to sort these discrepancies out though poll aggregators — sites which gather the disparate polls and re-calculate all the results offering up some sort of believable average. The serious analysts don't just add and divide, but weight different pollsters based on past accuracy and other factors. However, the aggregators often don't agree among themselves. There's been a blog-based battle between two of the most popular sites, RealClearPolitics.com and FiveThirtyEight.com over which polls should be in, or out, of the sample. Then last week there were the numerous articles about American racial attitudes and "The Bradley Effect" — which suggests that predicting whether people are telling the truth to either themselves or pollster is always a losing game. Our best advice is self-education — read the tips below and read the each site's FAQ.

In addition, get advice from Daniel Yankelovich, the founding father of public opinion research in an excerpt from the MOYERS DIGITAL ARCHIVE:

"I think that if I had to make one single suggestion it would be to ask yourself the question, when you look at the poll results, is this an issue where people have made up their minds? You may not know, but if you see inconsistencies, if the wording of the question changes the response. And you can ask yourself, have you made up your mind about Social Security, or Medicare or — drugs for seniors, or more money for schools and things of that sort. If you haven't made up your mind, the poll the people that are polling are like you and they haven't made up their minds, you can't rely on the poll results." -- Daniel Yankelovich

The National Council on Public Polls has a list of questions all journalists should ask before citing a poll. They are useful questions for a media-savvy public as well.

• Who did the poll?
• Who paid for the poll and why was it done?
• How many people were interviewed for the survey?
• How were those people chosen?
• What area (nation, state, or region) or what group (teachers, lawyers, Democratic voters, etc.) were these people chosen from?
• Are the results based on the answers of all the people interviewed?
• Who should have been interviewed and was not?
• When was the poll done?
• How were the interviews conducted?
• What about polls on the Internet? (There's no sample control on the Web.)
• What is the sampling error for the poll results?
• Who's on first? (The ordering of choices, especially candidates, can affect responses.)
• What other kinds of factors can skew poll results? What questions were asked? These questions are annotated at the National Council on Public Polls web site. You can also find an Ethics and Standards guide to polling at the American Association for Public Opinion Research.

Published October 24, 2008.