Friday, November 2, 2012

State and National Polls are Diverging

In my models the average of state and national polls conducted by dozens of firms do not add up. In the RealClearPolitics average of national polls, Romney and Obama are tied. The Polls vary from favoring Obama by 3 points to favoring Romney by 5 points. Meanwhile, many of the same firms have conducted dozens of state polls which are also averaged. The average of these state polls is posted on RealClearPolitics (RCP) as well as on my “2012 Election Polls and Models” blog that I post weekly. One would theorize that the average of all state polls should track the national poll averages, right? Well, when I do this, Obama gets 51.2% of the vote compared to 48.8% for Romney (Of course I am not including any third party candidates which will account for about 1 to 2% of the popular vote). This gives Obama a 2.4% win nationally, whereas the RCP average of polls has the race tied. This is a 2.4% divergence of state and national polls. So, which is right, the state polls or the national polls? This is the 1 billion dollar (amount each campaign has sunk into this election) question that we may not know the answer to until Election Day. Democrats say the state polls are correct and Republicans say the national polls are correct.

Think about it, Obama won the 2008 election by 7.3%. If the current race is considered a tie on RCP averages, well that is a gain of 7.3% by the Republican Party. But according to the latest RCP averages for state polls in battleground states we see the following:


2008 Result

2012 RCP Projection



Obama 2.5

Romney 1.2

Romney 3.7

North Carolina

Obama 0.4

Romney 3.8

Romney 4.2


Obama 6.3

Romney 0.5

Romney 6.8

New Hampshire

Obama 9.5

Obama 1.3

Romney 8.2


Obama 4

Obama 2.3

Romney 1.7


Obama 8.6

Obama 0.6

Romney 8


Obama 12.4

Obama 2.7

Romney 9.7


Obama 9.3

Obama 2

Romney 7.3


Obama 10.6

Obama 4.6

Romney 6


Obama 16.5

Obama 3

Romney 13.5


Obama 12.8

Obama 3.7

Romney 9.1


Obama 12.2

Obama 5

Romney 7.2




Romney 6.2

If I weight each of the states based on their population, Romney only has a 6.2% gain, not 7.3% insisted by national polls. These 12 battleground states account for 31% of the nation’s population. This means, Romney will have to gain 8.5% in the other 38 states to account for an 7.3% gain overall, and this is not the case in the state polls. In fact, the battleground numbers would be much worse if Romney did not have ties to Michigan and Ryan did not have ties to Wisconsin. Their ties to these states yield an additional 3 to 5 point gain.

The bottom line is that the state and national polls do not add up. We will find out on Election Day which is correct. My theory is that national polls oversample Democrats by at least 1.2%. However, on the battleground state level, Democrats are being oversampled at higher rates. For instance, in Ohio I believe Democrats are being oversampled by 2.7% and in Florida by 1.6%. This may explain part of the divergence, but we will not know until Election Day for sure.


  1. It's going to be very interesting (to say the least) how this shakes out in the end. I find it odd that both sides seem pretty confident about winning.

    1. Yes, I think if you go by the state polls D's think they are doing good, but if you go by national polls the R's think they have a chance. Only one can be right.