National poll: Obama/Romney deadlocked at 46% but Obama is favored in big states

[buckshot24]

I read that article and nobody has explained what the weighting actually represents. They just assume it's by date because that terrible journal says so.

It could be something else. Sample size for example. Margin of error. Someone find out.

I think Silver's model is slow to react to changing data. So if a poll seems to represent the trend of the prior polls of the same company he gives it more weight because there is corroboration to the latest result. He probably gives the ABC national poll just released a lower rating because it shows a change of 4 points from it's previous poll. Just a guess.

 

[buckshot24]

Taking a look at 538 polling weights.

The Quinnipiac poll in Ohio has Obama up 5. Silver gives this poll a 1.73 weight.

• Suffolk has it tied and Silver gives it a weight of 0.83
• Angus Reid has it tied and Silver gives it a weight of .61
• PPP has it Obama up 1 Silver gives it a weight of .87
• SurveyUSA has Obama up 3 with a weight of 1.11

It seems to me that Quinnipiac is given too much weight and it so happens to give the largest Obama lead.

Pulse Opinion Research 10/8 has Obama up 1 with a weight of 1.00
Pulse Opinion Research 10/15 has Romney up 1 with a weight of 0.34

The newer poll is given less weight.

Here is the bullshit of his system.

• There is a Quinnipiac poll from 8/21 with Obama up 6 Silver has it weighted 0.60.
• Marist poll on 9/11 Obama up 7 Weighted at 1.05
• PPP poll 9/9 Obama up 5 weighted 1.13 which is higher than the PPP poll for this week!

He has polls from May that have weight.

 

[buckshot24]

https://www.youtube.com/watch?v=c0p05FRR1zc

You have Silver saying Republicans would get around 45-50 seats in 2010. He was off by 13-18 seats.