Saturday, October 11, 2008

Obama lead in Colorado hits double digits

Barack Obama 52
John McCain 42

With the Hispanic vote unifying behind him, Barack Obama is out to his largest lead yet in a PPP poll of Colorado.

Obama is now leading 71-21 with that group of voters, a considerable improvement from a 57-36 edge in PPP's survey of the state taken three weeks ago. The white vote is exactly the same as it was previously, a 48-47 advantage for Obama. Obama's margin has increased by three points since that poll, which showed him with a 51-44 advantage in the state.

The economy may be helping to drive Obama's gains. Three weeks ago 51% of voters in the state listed it as their biggest concern. That number has now spiked up to 60%. Among those folks Obama's lead is 60-34.

Colorado also provides a clear example of how Sarah Palin's popularity across the country continues to fall. Right after the GOP convention her favorability was 41/38. Three weeks later it was 38/47. Now it's dropped to 36/49.

The numbers paint a picture that shows it nearly impossible for McCain to win the state at this point. Only 6% of voters are undecided, and among those who do have a preference currently only 6% say there's any chance they could change their mind. Obama's lead at this point is such that McCain would have to win virtually every undecided, keep everyone who could change their mind currently supporting him in his camp, and win over almost 100% of current Obama supporters who say they aren't totally committed to him. The chances of all those things happening? Not very good.

Full results here.


Anonymous said...


Sreenu said...

As per the voter registration statistics (Sept 2008) available from the State of Colorodo the party ID is 32.1%D, 33.7%R, 33.7%I and 0.5%Others.

The party ID chosen by you for this poll is 41%D, 37%R and 22%I.

This discrepancy in party ID chosen on this poll should not affect end result much because Obama and McCain get pretty much the same support from their respective parties and Obama leads by 21% among the independents.

However I would be interested in knowing the reason for choosing a vastly different party ID for this poll than what is reflected from the voter registration statistics.

Tom Jensen said...

We don't 'choose' any party breakdown. We just weight for fixed demographics- race, gender, and age- and let the party breakdown fall wherever it falls. My guess is that in this Democratic year there are a number of Coloradoans registered as Republicans who are not identifying as such.

Sreenu said...

Thanks for clarification Tom :-)

Anonymous said...

Those numbers in Colorado are right on.

Arapahoe County one of the most conservative counties in the Denver metro area. Democrats out registered Republicans for the first time ever.

Colorado will be blue.

Anonymous said...

srinivas you misunderstand PPPs numbers. PPP did not SELECT the party id breakdown. They called people and asked them what party that were. Oftentimes people give a different party id to pollsters than what is listed on their voter registration. There are two reasons for this: (1) people have changed in their own minds what party they believe is most in sync with them but never bother to change their voter registration, and (2) their is the bandwagon effect, voters as an election get closer start identifying themselves to pollsters as belonging to the party they think is going to win in an election. Reliance soley on voter registration numbers can cause people to misinterpret polls or wrongfully discount them because they think the party id is too out of sync with actual registration. Finally, if you notice PPPs party id for its CO poll showed both Dems and Reps outperforming their share of the registered electorate, this happens nearly all the time in elections.

Sreenu said...

@anonymous, actually they almost out-registered them :-)

328,940(Total Voters)

But still it a big achievement from where it stood in 2004.
108,987 (D)
136,870 (R)
125,755 (UA)
372,631 (Total Voters)

Anonymous said...

The numbers probably haven't been updated.

Sreenu said...

@anonymous, thanks for the explanation. I see the point in what you are saying. I completely agree (1) is possible but wonder about (2) scenario. What if people tell the pollsters that they belong to the likely winner but end up voting their party ID, would it not affect the accuracy of the poll? Is there a way to include the registration data along with demographics while weighting the raw polling data?

Sreenu said...

I got the numbers from here -

They are dated September'30 2008.

Anonymous said...

srinivas the numbers you quote from Arapahoe county are from the end of September. The voter registration deadline in CO was not until Oct. 6, meaning there was almost a weeks worth of information not included in the secretary of state's release of registration information you quoted. Given that the difference was only around 1,000, it is conceivable that anon is right and that Dems passed Reps in terms of registration in Arapahoe county for the first time.

Anonymous said...


you guys need to do more polls


Anonymous said...

So why not weight by party id also? Just curious.

Cugel said...

There are all kinds of arguments for and against weighting by party ID. Rasmussen does, and uses a "likely voter screen" as well. SurveyUSA and PPP don't.

If you DO weight by party ID you have to decide what it should be. You can't just go by registration numbers, even in states which HAVE such data (many don't register by party so there's no way to know).

People who are registered as "Republican" may self-identify as Democrats or vice-versa. What counts in elections is not what's written on a reigstrar's list, but what people think inside their heads, ignoring peer or family pressure, reasons for registering one way or another.

Rasmussen polls for their party ID every week, but even that is subject to error.

NOT weighting by party ID leaves you subject to just getting a bad sample that has too many Republicans or Democrats in it. But, weighting for party ID means you're second-guessing what percentage of voters really identify as Democrats say.

And we know that changes even over the course of a campaign! In 2000 Democratic party ID was around 39%, a few percent higher than Republicans. Then after 9-11 it went down, in 2004 Republican ID hit a low of around 34% and then increased to tie Democrats at 37% around election time (all according to Rasmussen).

So, you introduce an error factor, no matter which way you go. Either approach has validity and advantages as well as disadvantages. It's best to look at an average of all the polls like does.

Anonymous said...

青島 温泉

Web Statistics