Persistant Problems With Polls

[A Person]

I have been skeptical of many of the polls released in the last 5 years or so. They tend to overestimate the conservative vote when compared with election results.

Marist have done some research and think they know why:

Cell Phones

Pollsters mostly call people on land lines, i.e. older people. Older people tend to vote conservative

"In Florida, Romney is +3% with landline voters, but Obama is +23% with cell phones."

5/30:Don’t Cell Out!
It should be the battle cry of all pollsters and poll watchers interested in accuracy when it comes to presidential polling. The problems of reaching cell phone only respondents are well documented. But, measurements of the Obama-Romney horse race that rely solely on landline households do so at great peril.

Check out the results from the recent NBC News/Marist Polls of swing states. In Florida, Romney is +3% with landline voters, but Obama is +23% with cell phones. In Virginia, it’s pretty much the same… Romney +1% with landlines, Obama is +18% among cells. Owing to his wider lead in Ohio, Obama is +4% with landline voters but is also +9% with the cell phone electorate.

Two final comments on this issue of methods: First, there’s no doubt that Obama’s advantage with cell phones has a lot to do with younger people dominating the cell phone only electorate. But, the under 30 aged voters made the difference in 2008. It would be a pollster gamble to undercount them this time. Second, this discussion has been only about telephone surveys with live interviewers. Robo-polls which also exclude cell phone households are an entirely different polling battlefield.

The bottom line: cell phone only households need to be included in all accurate attempts at public opinion measurement. For campaign 2012, it’s not a choice, it’s a necessity.

The other thing that has not been addressed is the political makeup of people prepared to sit through a poll. Many (most?) people hang up on pollsters. The people who don't are likely to be more opinionated, are they more likely to be conservative? I don't know but I would like to ...

[The Vicar]

I'd love to think that this is significant, but similar claims have been made in every national election in the last decade, and yet the actual vote counts are predicted accurately by land-line polls, which suggests that either cell phone users don't bother to vote or they are lying when polled, so at this point I am skeptical of the utility of the claim even if it is true.

[A Person]

yet the actual vote counts are predicted accurately by land-line polls

Are they? That's not my impression.

My impression is that Intrade.com does a far better job of predicting winners (Obama by 60%) maybe because people have to back up their opinion with cash.

[Liv]

What is this "land line" you refer too? I think I read about this in a museum or such? People still have them?

[The Vicar]

My impression is that Intrade.com does a far better job of predicting winners (Obama by 60%) maybe because people have to back up their opinion with cash.

As I recall, Intrade's record hasn't actually been as accurate as land-line phone polls overall; it falls down on Congressional contests. (I seem to recall that there was an article about it after the 2010 midterms.)

As for "predicting winners": the land-line phone polls in the last week before an election not only tend to correctly predict winners, but the average results tend to be within a percent or two of having the correct popular vote percentages.

All of which suggests that cell phone users as a group are still irrelevant to politics; the only question is whether this is because they lie when polled (and actually end up voting like land-line phone users) or because they just don't vote in sufficient numbers to alter the election results.

[Liv]

Why can't we have biometric cell phone voting?

App shoots picture of eyeball, you answer a secret question, and vote.

Seems simple enough.