Are GOP voter purges skewing the poll numbers?

[Gman]

Are GOP voter purges skewing the poll numbers? I read the BuzzFlash interview with Greg Palast posted by struggle4progress, and it occurred to me that all the major polls sample registered voters and likely voters. And, there is a distinction between the two.

Then it occurred to me that the pollster really has no way of knowing if someone is or is not a registered voter and therefore a likely voter. A person being polled self-identifies that they are or are not registered. This has massive implications for Obama's the current poll numbers. In a battleground state like Colorado, where this morning's Washington Post has Obama with a 9 point lead. The Colorado poll was of 1,088 likely voters. And according to Palast, 19.4% of all registered voters in Colorado have been purged.

The WAPO article doesn't state the methodology and maybe it's posted somewhere, but I don't know where and I still haven't had enough coffee. So let's make some assumptions. Let's assume the poll was weighted 35% Democrat, 30% Republican and 35% other. Polls seem to be structured approximately pretty much like that.

Data from the Quinnipiac University on this poll shows that candidate support breaks down as:

	Democrats	Republican	Other
Obana	93%		6%		51%
McCain	4%		89%		40%
DK/DC	3%		5%		9%

Obama/McCain. From these numbers we get:

Obama	354	20	194	568
McCain	15	290	152	458
DK/DC	11	16	34	62
	381	326	381	1088

Now, 19.4% of all voters in CO have been purged from the rolls. Let's assume that 80% of the purged voters were Democrats, 5% were Republican and 10% were Independent or other (sounds reasonable to me). That means that the actual number of sampled Democrats in the Colorado poll after the purge would be:

		Democrats	Republican	Other	Total
Aft Purge	212		316		360	887
Sample %	24%		36%		41%	100%
Difference	169		11		21	201
% Diff		44%		3%		6%	18%

Now we have a poll where Democrats were under sampled. However, the same numbers of voters would turn out to vote because they didn't know they were purged. Assuming no one knows they were purged or else they would not have told the pollster that they were registered, 44% of Democrats would not be allowed to vote.

Using the same preference percentages as above, the new breakdown would be:

	Democrats	Republican	Other	Total	Percentage
Obama	197		19		183	400	45.0%
McCain	8		281		144	433	48.8%
Other	6		16		32	55	6.1%
						887	

We would then see that McCain squeaked by in a 48.8% - 45.0% victory in Colorado. Even if you reduce the percent of the 19.4% purged to 60% Democrat, 5% Republican and 35% Indy, the results are a 47.2% - 47% McCain win. They will call him the Comeback Kid, just as he did in the primaries. Something smells real bad here.

Check my numbers, I think they're correct. Maybe my methodology is off. I've had plenty of coffee now so I think they're correct.

[NYC_SKP]

1. I agree that the purges are developing into what seems to be a deliberate

strategy to cheat their way to victory (again).

As I type this, with three weeks to go, I have to feel confident that between the Obama Campaign and folks like us, these efforts won't be successful or at least their impacts will be minimized, compared to how these things have gone in the past.

I'm in California, quite removed from these actions, but remain confident that we're not going to let them bamboozle us.

Thanks for the heads-up.

Recommended.

[pnwmom]

2. This is exactly why we need MASSIVE turnout.

Many people will turn up to vote only to find that there are problems with either their voting or their registration.

[Gman]

6. Massive turnout that even the pollsters cannot predict is the key

to overcoming this.

[Jersey Devil]

3. Pollers use registered voter lists

You can get them with phone numbers and addresses for all registered voters. They don't rely on the voters' word for it that they are registered.

If they really want to know who is a 'likely voter' they can also get lists that show how often a voter actually votes in every general election for the past 5 years

[Gman]

5. It's true that you can buy lists of registered voters

along with their history. So that would seemingly invalidate my assumption about voters self-identifying (I blame that on not having enough coffee when I started the post!). However, I think it is likely that pollsters could well be working off lists that were supplied before the voters were purged. I know those lists can be somewhat expensive as I've purchased similar lists for block walking efforts and phone banks quite a few years ago now. Most purging seems to have occurred in the last month or so. So I think it's plausible that voters responding in the poll think they are registered when they are not and still intend to go to vote not knowing they are not registered. Pollsters may not be aware they're working from stale lists. I doubt the people doing the purging would go out of their way to advise the pollsters that their lists were stale.

If you were to massage the numbers to this effect, I think the results would still show a much closer race than the poll numbers are showing.

[WCGreen]

11. They should be updating their lists as they are updated at the BOE...

Most of the people who are purges are non voters, people who have died or people who have moved and registered some place else.

I know this because I have worked in the polling industry and have also served as a member on the Board of Elections.

I agree you have a point. However, the polling companies would not jeopardize their reputations by using outdated lists. And as far as the expense goes, they can afford it. They just pass it on in their pricing structure.

[Jersey Devil]

13. In NJ they update the lists every week in the 90 days before an election

Our local candidates often use those lists (which you can buy with mailing labels) when they campaign door to door so they can know who they are talking to, their party affiliation (if any), etc.

[stillcool]

4. I read somewhere that likely voters..

are registered voters who have not voted in the recent past. I also recall reading in the primaries that newly registered voters names are not released until after they vote. Don't know if this is all bullshit, as I don't pay much attention to polls and the statistic analysis behind them.

[Gman]

7. I guess it would depend on how the lists were requested

The pollsters are well aware of the enormous numbers of people newly registered. So I would think they would want to take that into consideration. One way to do that would be to include the newly registered in "registered voter" numbers but not "likely voters" as they have no history.

[WCGreen]

9. Registered voters are available to the people as soon as they can get them

on the rolls. It is public information and available to anyone who wants to purchase a poll book....

[stillcool]

12. yes..I read that those names..

are not released until after they have voted. This was in the primaries that I read that.

[WCGreen]

8. Before you get all upset, any reliable pollster gets their universe

of registered voters from either the local BOE or from the SOS.

Then they decided, based on their model, who is a likely voter and who isn't.

[Gman]

10. Read my reply #5