Interesting WSJ article on why national voter polls have such varying results

Not a debate as such, but likely to be of more interest to this forum than elsewhere. Mods please move if you think it belongs elsewhere.

Some Surveys Indicate Tighter Presidential Race - Differences in Predicting Outcome Result From How Pollsters Gauge Voter Turnout and Weight Party Affiliation

Their description of the margin of error is wrong in that image. I guess it is close enough, but these kinds of things bother me.

Thanks for the article.

Worth remembering though that not all polls and psephologists are created equal.

They’re all still showing a consistent lead, and the chief examples of there being a tighter race are not from the biggest and most respected outfits.

Zogby’s claim of “That’s just not America, period” isn’t particularly compelling. Such from-the-gut conventional wisdom may not be appropriate given the shifting dynamics of new registrations.

Also, we know that telephone based polling can actually under-represent Democratic support. The problem in the past has been getting these people to care, register and show up. I don’t think the Obama GOTV effort will disappoint.

Per the notion in the article that cell centric younger voters are they only ones without land lines (thus being inaccessible to polling) is also kind of (IMO) incorrect. Lots of single middle aged people, including myself, have dumped their house land lines. It’s was $ 400 a year to maintain a land line even with minimal use. It was a waste of money for the number of times I used it.

I, too, thought that was a very strange comment, as presumably it is sampling efforts that inform his picture of America in the first place. It will always be a close race if we only consider proportions that align with our preconceived notions about political distribution when that notion is a 50-50 split… in fact, it will be worthless polling, because we will fail to obtain random sampling at all, instead basing polls on self-selecting samples! Weird. I’m sure he must mean something else by his comments, as he’s way smarter than me, but it is hard for me to dig down and understand what he’s actually talking about.

A guy I go to school with informed me, “I don’t listen to those polls.” All of statistics, wiped away by the slightest bit of ignorance. Ah, well. It will definitely take longer than we think.


Some pollsters, well, very few pollsters apply weights to compensate for the cell phone effect.

This early line makes me suspicious. Real Clear Politics is run by conservatives (it was purchased by Forbes last year), and Nate Silver of has accused them of cherry-picking polls that favor John McCain.

I would tend to disagree with you,a person should vote the way that they personally believe in even if they are perceived as being wrong by the majority(perhaps wrongly)

I think that it is decidedly unhealthy for people base their vote on what everybody else thinks or to be on the winning side.

Apart from the fact that polls are not necessarily accurate by reason of their poll sample being wrong,question bias or even the people polling lying the polls themselves influence the voting.

Many years ago in a British General Election the polls showed that the Labour Party would have a landlide and that the Conservative party mightjust aswell fall on their swords en masse.

Result the L party voters stayed away in droves as their votes wouldn’t be needed whereas virtually every C.voter who wasn’t actually on their death bed turned out to vote.

Much to the astonishment of everyone including the themselves they were elected into government.

I’m not sure but I think the law was changed to prevent polls being conducted within a certain time close to elections after that.

Dont get me wrong I hope that in this case the polls are spot on,if Obama doesn’t get in then the U.S. will be in serious trouble but your school mate is in the right of it in his own personal decision making.

Oh, he should vote however he wants. He was talking about not trusting statistics.

Well the issues to debate are not the obvious things mentioned in the article, that sampling variations explain a lot of variation in the polls and that polling accuracy as a predictive measure depends on how well the sample reflects those who actually come out to vote on election day, a more difficult task perhaps in this very atypical election cycle, but how much each factor is distorting polling results in which direction.

In that regard another inaccuracy in that article is noteworthy:

No. Gallup polls one group and then reports it three ways; all registered voters; applying a “traditional” likely voter screen (as described); and applying a different “expanded” likely voter screen that ignores whether or not someone has voted before but goes by intention to vote as its main screen. The last is felt to potentially be more accurate as it does not filter out the newly empowered younger and Hispanic voters - if they actually show on election day. It is of note that the expanded screen has run anywhere from 2 to 4 points higher Obamaward than the traditional screen, sometimes even higher than the all RV results.

So how much will past voting behavior predict turnout this time? Or will it actually be very misleading? How big will the cell phone effect be? How much will the different ground games effect turnout on election day? Will there be any Bradley effect or a reverse Bradley effect for that matter? Will some states be more effected by some of these issues than others? If so will they lead to any particular surprises in which polling was way off?

I am confident only that there will be some states with surprises but have little expectation that we will figure anything else out with so many factors going on at once.

Whoops,sorry mate,my mistake.