My final predictions are that Barack Obama will receive 332 electoral votes and Mitt Romney will receive 206. This is the same as the Princeton Election Consortium arrived at under different assumptions but using only state polls. The current results for all states are in State Polls 11-5-12. Florida (Figure 1) remains a tossup, but it leans .31% towards Obama, so the best guess is that Obama will ultimately win Florida (though this is only little better than a coin flip).
The popular vote is predicted to favor Obama by 2.67%, which is down from 3% in previous posts for several reasons. First, the difference favoring Obama in California dropped a couple of points, which brings down the national average about .2%. Second, several small states were never polled and I originally used 2008 voting estimates for them, but to be more conservative, I used 2000 voting estimates for those states due to the closeness of that election. Third, the percentage difference favoring Romney in Texas increased a couple of percentage points, which decreased the predicted difference between Obama and Romney.
If we assume 1% of the vote goes to other candidates, then the popular vote should be Obama 50.87% and Romney 48.17%.
One way to reduce the influence of outliers (e.g., large biases in particular polls or fluctuations in enthusiasm), is to calculate the median for each state for the polls taken over the approximately 2-month period pooled for each state. Means-Medians sorts the states into those for Obama, Romney and Tossup with corresponding means and medians. As can be seen, the means and medians are in strong agreement and the predicted popular vote based on the medians should favor Obama by 2.61%.
For each state, the pooled poll can be viewed as one large poll conducted over approximately two months by a number of polling groups. The main assumption is that most voters had stable preference for Obama or Romney by the beginning of September. People can switch candidates, but this assumption requires that switching occurs at low rates and is essentially random. This suggests we should see the polls for states converge on percentage differences rather than continuing to gradually increase or decrease over the sampled period. This appears to be true for Ohio (Figure 2), North Carolina (Figure 3), Virginia (Figure 4) and Colorado (Figure 5).