How To Read Polls Without Hyperventilating
by DemFromCT
Fri Jul 25, 2008 at 01:05:24 PM PDT
Every once in a while, it's important to go back to basics. Now, this is a lesson the talking heads need to learn, but it is helpful for us to learn the same things.
They are not predictive. They are not definitive. They are, however, representative. To insist the data shows a lead today is not to insist that the lead translates into a win tomorrow, next week or next month.
The art of the pollster is in getting this part right. It's not easy, and some pollsters have a track record of doing this better than others. Party ID is a variable, for example, so decisions about weighting affect the results of the poll. However, huge fluctuations generally raise red flags, particularly when no event to match the fluctuations occurs. Gary Langer (ABC):
"Trend is your friend," pollsters say. Look at repeat polls from the same organization to gauge movement over time. And, again, look beyond the horse![[image]](http://mowser.com/web/http%3A%2F%2Fimages.dailykos.com%2Fimages%2Fuser%2F426%2F538_03.jpg)
race to other measures: the levels of commitment and enthusiasm from a candidate's supporters, the groups that are more or less fired up, the factors motivating their support. It takes willpower to trudge off to an hours-long Iowa caucus on a dark winter's night. Who's inspired? How? Why?
It's actually more important since we elect via the electoral college, but it's harder to do. But there is nothing more infuriating than having a blowhard like Chris Matthews insist, as he did today that this election is a dead heat. Take a look at the electoral estimate from fivethirtyeight.com to the right. Or take a look at the electoral estimate from pollster.com. Does that look like a dead heat? See the tracking poll data below for more.
![[image]](http://mowser.com/web/http%3A%2F%2Fimages.dailykos.com%2Fimages%2Fuser%2F426%2Fpollster.jpg)
It is difficult to know, in a change election, who the likely voters are. Right now, LV based on past elections favor older voters, who skew McCain. Before the convention, it's best to rely on registered voters, but always make sure you are comparing apples to apples.
Compared to 2004, there are many more polling resources available on the internet. Invaluable sites like pollster.com, www.fivethirtyeight.com, and www.realclearpolitics.com as well as Mark Blumenthal's Thursday series in National Journal should be consulted regularly (and there are others who specialize in electoral college estimaters). Other terrific resources include News University's online course in how to read polls (partnered with American Association for Public Opinion Research) and 20 Questions A Journalist Should Ask About Poll Results from the National Council on Public Polls.
There are also pollsters who write regular columns about their findings, and the art of polling. Not to be exclusionary, but Scott Rasmussen, Frank Newport at Gallup, Kathy Francovic at CBS and Gary Langer at ABC come to mind as frequewnt posters.
Some sites, like pollster.com or fivethirtyeight.com (personal favorites) allow comments. Read the comments. Ask questions. It's the best way to learn.
Look at them all. If the NBC/WSJ poll has Obama up by 6, and the average of polls has him up by three, then a new Gallup tracker that is tied doesn't 'change everything' - nor does the WSJ/NBC poll, for that matter.
None of that predicts who wins in the fall, but data will enhance your commentary. And the use of data would be a useful habit to get into. But don't be assuming a Bradley effect exists, for example, when that may or may not be the case.
Remember, when polls don't fit the news or the feel you have for the campaign, something is off. It may be the polls, and it may be you.
But if you missed what's going on, and the polls pick up something happening, don't blame the polls and don't blame the pollster. Just use them to expand your world view. And, of course, wait for the next poll to tell you if it's a trend. And, for God's sake, don't compare a 14 point lead in Zogby (internet) to a new Q-poll that shows a 4 point lead in MI and claim this is a remarkable change. The previous month's Q-poll had Obama up by 6, so that's just noise. And as per First Read:
Regarding those Quinnipiac polls, don’t miss this: "Clay Richards, the assistant director of the Connecticut university's polling institute, said the Obama slide [in Minnesota] probably isn't as dramatic as the raw numbers reflect. Still, Richards said McCain is clearly stronger in the state than he was in June."
Rasmussen, meanwhile, says Obama got a bounce from the overseas trip, and the speech was well received by voters. Gallup today has Obama up by 6.
Previous election year polling shows that the conventions have a high (but not 100%) probability of shaking up the race. The impact of events prior to the conventions is certainly more difficult to pin down. As noted, the structure of the race this year appears to have snapped into place in early June after Hillary Clinton dropped out of the race for the Democratic Party's nomination, and it simply has not changed much since. Whether or not it will change at this juncture remains to be seen.
Take deep breaths and stop hyperventilating over those, too. And go back and review what happened in 1980. Ronald Reagan was the risky unknown. Reagan's move in the polls was very late, but it was decisive.
In the meantime, ask yourselves if the networks are cherry-picking polls they want to use to sell their narrative of a close race (particularly the usually ignored Q-polls) and ignoring the Rasmussen and Gallup polls today. The real picture requires looking at all the polls, not just the ones you like.