First McBlunder gets a pass from the media, now he's being helped by rigged polls

[ProSense]

New Poll: McCain Leading Among Likely Voters!

By Eric Kleefeld - July 28, 2008, 3:35PM

Not so fast with all the gloating over Barack Obama's lead in most of the national polls. A new USA Today/Gallup poll gives John McCain the lead among likely voters -- quite possibly the first poll in a long time to show this -- and Obama only a narrow edge among the wider pool of registered voters.

The numbers: Among likely voters, it's McCain 49%, Obama 45%, with a ±4% margin of error. A month ago, Obama led 50%-44% in this group. Among registered voters it's Obama 47%, McCain 44%, compared to Obama ahead 48%-42% last month.

This runs contrary to all the other polls out there, which have Obama ahead in both the registered and likely voter categories. Gallup pollster Frank Newport speculates that for his poll, Republicans are listing themselves as likely voters in greater numbers because of all the press coverage surrounding Obama's overseas trip and the crowds welcoming him in Europe: "At least in the short term it may have had the side effect of energizing Republicans."

How long before they start telling us this is McBlunder's election to lose?


On edit: Don't these organizations have any controls? Can't they analyze the data and determine if it has been corrupted before they publish it? What sense does Newport's comment make in the context of an already published report?

[depakid]

1. Around DU, the polls are only rigged when they show results people don't like

otherwise, the American fetish with bogus polls is alive and well.

[ProSense]

2. They're deemed rigged (manipulated) when the pollster makes this claim

Gallup pollster Frank Newport speculates that for his poll, Republicans are listing themselves as likely voters in greater numbers...

[depakid]

4. One of many dozens of ways so called "scientific" polls are manipulated

General rule: Polls in the states are meant to influence public opinion, not reflect it.

Corollary: Accurate and honest assessments (AKA valid and reliable public opinion research) is time consuming and expensive (which is why campaigns spend so much money on internals and keep the results held tightly to their chests).

Always a good idea to keep these rules in mind....

[ProSense]

5. "Accurate and honest assessments (AKA valid and reliable public opinion research) is time consuming"

Isn't that the business Gallup is in? If not, they should always include a disclaimer like the bogus online polls. Then, maybe, the news organizations will stop printing the result with such frequency.

[depakid]

6. Gallup will tell you that 70% of Americans believe a guardian angel is watching over their shoulder!

And people will believe it!

In fact and its ilk are often held up in grad school courses as examples of how not to do proper research. Generally speaking though- that's not their business. Their business is selling polls to clients- and tailoring results for what that client wants to see.

Similar to how corporate auditing firms have operated over the past 15 years.

[ProSense]

7. "Their business is selling polls to clients- and tailoring results for what that client wants..."

All the more reason they should carry a disclaimer.

Still, I think that would be news to Gallup, especially in the context of public opinion polling for the general election. Who are they "tailoring" the results for?

[dansolo]

11. Especially since evidence suggests the exact opposite

Voter registration and self-identification amongst Democrats has increased dramatically, and enthusiasm in the GOP is very low. If this poll weights Republicans higher than Democrats as "likely voters", then it is complete crap.

[A-Schwarzenegger]

3. "Republicans are listing themselves as likely voters" ???

I thought the pollsters determined who a likely voter is.

[book_worm]

8. It goes against the grain that republicans are more enthused than dems.

[ProSense]

9. USA Today/Gallup's highly questionable likely voter model

USA Today/Gallup's highly questionable likely voter model

by FleetAdmiralJ

Mon Jul 28, 2008 at 01:03:30 PM PDT

Or at least that's one way to look at this new poll. In it, Obama leads McCain 47% to 44% among registered voters, but trails 49% to 45% among likely voters. There were 900 respondents in the registered voter poll while only 791 people were included in the likely voter poll.

Setting aside the fact that having nearly 88% of registered voters being likely voters is silly to begin with, let's dig into these numbers.

In the registered voters poll, Obama wins 47%, or approximately 423, while McCain wins 396 people. Among likely voters, McCain wins 388, while Obama wins 356 people. All of these are +/- 4 people, depending on rounding.

So, looking at the average, Obama had 67 people dropped by the likely voter filter (with a range of 75 to 59). That represents nearly 16% of Obama's registered voter support. In the meantime, McCain lost 8 people (with a range of 0 to 16). That represents 2% of McCain's registered voter support.

So what USA Today/Gallup is basically saying is that eight times as many Obama voters will stay home than McCain voters, at least based on percentage. If basing it on actual hard numbers, at best it is 3.6 times, and at worst it could be as high as, well, undefined since McCain potentially lost no one in the likely voter filter.

This shows how utterly stupid (and open to manipulation) likely voter screens are (again, igorning the fact that filtering out only 12% of voters is ridiculous anyway). Yet, does anyone doubt that the media will flash up the McCain up by 4 graphic up on the screen?

more

[Fozzledick]

10. The 4% margin of error is a red flag indicating premeditated fraud

The standard for national political polls is to use a sample size large enough to have only around a 3% margin of error, generally about 1100. A 4% margin of error indicates a much smaller sample, around 600. There are three possible reasons for using such a small sample: it's cheaper (which is why it's usually used by local TV stations or newspapers without the resources of a national polling firm),
it's faster (which makes sense if you want something for the 11 o'clock news on a speech given tonight),
and it's less accurate.

For a national polling organization to deliberately use such a small sample when they have no budget or time constraints indicates that their intention is to produce an inaccurate result that will be easier to spin.