Okay, you're not going to like this but...

[brooklynite]

...for your own health, time to lay off the Presidential Election polls.

The 10 most recent posts include polls that tell us that Romney will beat Obama and that Obama will beat Romney.

And there will be 2-3 new polls a week through Election Day 2012.

Looking at trends over time makes sense. Fixating on each one is a formula for frustration.

[Lasher]

1. I like Real Clear Politics

Since they offer a composite of polls, I can avoid focusing too much on each individual one that comes out.

http://www.realclearpolitics.com/epolls/other/president_obama_job_approval-1044.html

But there is nothing wrong with avoiding them altogether, as you advise.

[karynnj]

2. I do too - even though the articles and the site itself leas right

They were completely fair in the surveys they included in 2010 - and the average is just that - a mathematical average.

This other page, puts the ABC poll in context - it is an outlier and the average of polls, even with it included, is still very good for Obama. (second chart on this page - http://www.realclearpolitics.com/epolls/2012/president/president_obama_vs_republican_candidates.html )

[JoePhilly]

5. During 2008, they were very accurate.

[Lasher]

6. I agree, the RCP articles are Wingnut City so I almost never look at any of them.

The only problem I have with the RCP composite is the strong influence of Rasmussen polls, which are biased in favor of conservatives. That is because Rasmussen publishes polls much more frequently than others do. Therefore, Rasmussen is uniquely represented in every RCP sampling of recent polls.

This is not RCP's fault, though. I don't see how they should play it any other way. Just as you say, RCP is completely fair in their polls.

[karynnj]

7. At least they seem to include in their rules to include only the most recent of any pollster's polls

I liked, during the 2010 elections that Daily Kos with a similar average, did ones with RAS and without RAS.

I didn't mind because they also gave you the list. This was good because you could then look at them in a more sophisticated way if you wanted. I did wonder why they did not use a more robust estimate rather than a straight mathematical average - for the Presidency, there will be enough polls, that something as simple as throwing out the highest and the lowest would likely be better. But the full list, coupled with the dates polled and knowledge of any events that could have had an effect, provides interesting information.

[Enrique]

3. and polls for every state

i think your estimate of 2-3 polls a week is way low.

[charlie and algernon]

4. national polls are worthless for presidential elections

State polls are the only ones that matter.


But even with them, it's WAAAY to early. Especially the Obama vs. Repub polls. We could still be a full year away from learning who the Republican nominee is. And recent history has shown that the person we think is going to be the nominee NOW will most likely NOT be the person it actually is this time in 2012.

[vroomvroom]

8. THANK YOU -- I thought i was the only one thinking this.