GWU Battlefield Poll -Obama (D) 50% Romney ($) 47%

[DemocratSinceBirth]

http://images.politico.com/global/2012/09/battlegroundpoll.html



Decent result. Their polls seem to lean a bit right. Some years they are more close than not. In their final 08 poll I believe they underestimated Obama's winning percentage by three or four percent.

IMHO, this isn't always a function of intentional bias but polling methodology.

[ProSense]

1. It had Romney ahead in May and Obama by one point previously:

The POLITICO - George Washington University Battleground Poll is a national recognized, bipartisan political opinion survey focused on election politics in the United States. The poll is conducted by Republican pollster Ed Goeas of The Tarrance Group and Democratic pollster Celinda Lake of Lake Research Partners.

http://www.politico.com/blogs/burns-haberman/2012/09/politicogw-poll-obama-romney-136467.html

[DemocratSinceBirth]

2. Celinda Lake Is A Democrat,Of Course

But that poll tilts right. Their claim to fame is they nailed 96 which was a long time ago. You can check their results/predictions since then. It shows a two to three point Republican advantage as it showed in 08.

[ProSense]

4. Mitt's unfavorable rating is up three points:

Poll: Obama Leads By 3 Nationwide, Outpaces Romney On Economy, Favorability
The latest Politico/George Washington University poll shows the presidential race almost exactly as it has been for months: Obama claiming a narrow but consistent lead, while Romney lags with low personal popularity. In the poll, Obama leads Romney among likely voters, 50 percent to 47 percent — slightly larger than the 1-point lead he held over the Republican nominee in the previous Politico/GWU poll in early August and within the latest poll's 3.1 percent margin of error.

Romney's favorability rating has slipped under water since August, when 46 percent of likely voters viewed the Republican nominee favorably and 46 percent viewed him unfavorably. Today, 46 percent of voters still have a favorable view of Romney, but the number of voters who view him unfavorably has ticked up to 49 percent.

Obama's favorability rating has improved from 50 percent a month ago to 53 percent in the latest poll, compared with 45 percent who view the president unfavorably (down 2 points from August). And the president barely leads Romney on the question of who is best-suited to preside over the economy, 49 percent to 48 percent. The former Massachusetts governor led by 5 points on that same question last month.

<...>

http://livewire.talkingpointsmemo.com/entry/poll-obama-leads-by-3-nationwide-outpaces-romney

[DCBob]

3. Not sure why its so difficult to report state by state results...

rather than adding them all together unless they are trying to hide something.

[DemocratSinceBirth]

5. Then They Would Have To Do A Separate Poll Of Every Single State

That would cost a ton of money.

[DCBob]

6. they have to be already polling every state they have identified as battlegrounds.

Surely they know which data goes with each state. Why not just show those numbers in the details? That wouldnt cost that much more to add that as a breakout.

[DemocratSinceBirth]

7. Because The Total Sample Is 1,000 People

That means they probably called only seventy five Floridians or so. The margin of error would be astronomical.

[DCBob]

8. I would still would like to see the raw numbers.

Maybe one state is pulling the numbers the down.

[DemocratSinceBirth]

9. We Are Getting In The Weeds

I suspect a statistician would say that if the poll was performed correctly with the proper controls another pollster could poll the same universe and get similar but not necessarily identical results due to the confidence level of .095 and the margin of error which is 3.1%.

I suspect Obama's lead is probably a couple of points higher but that can be as much a function of my bias as their bias.

[DCBob]

10. No weeds here, I would just prefer they would summarize the data by state...

just like they do for gender, income, political affiliation, etc. Not really a big deal I would think. They have the data.. its listed as item D17 on the questionnaire:

D17. RECORD STATE/COUNTY CODE FROM SAMPLE _____ _____ _____ _____ _____