There have been lots of new state polls this week, and I've rerun my simulations of the general election, one assuming a 4% margin of error for the state polls, and another using a 10.9% margin of error. The first is a snapshot of what might happen if the election were today; the latter uses a margin of error calculated from a linear regression of errors 2004 state polls had compared to the final results, assuming the error would fall as we came closer to the election.

This week's results:
4% Margin of Error
Obama wins 51.6%, averages 270.1 EV (low 212, median 272, high 338)
McCain wins 42.4%, averages 267.9 EV (low 200, median 266, high 326)
Electoral tie 6.1%

10.9% Margin of Error
Obama wins 53.8%, averages 272.5 EV (low 160, median 273, high 362)
McCain wins 44.5%, averages 265.5 EV (low 176, median 265, high 378)
Electoral tie 1.7%

The most common tie scenario involves Obama winning all the Kerry states except New Hampshire, and Obama also winning New Mexico and Colorado.

The polling is now recent for almost the entire country. Only Arkansas and Massachusetts have not been polled since the end of the GOP convention.

With so many polls, it's possible to see some rather large disagreements. For example, two polls last week gave Obama a double-digit lead in Iowa, but a third had the race tied. Three polls of North Carolina showed McCain ahead by 11, 1, and 6. There were 9 new polls in Ohio, ranging from Obama up by 2 to McCain up by 6. Two polls showed a tie in Pennsylvania, but a third showed Obama ahead by 5. Three Florida polls showed a tied race, but three other polls gave McCain leads of 1, 5, and 6. Colorado had 4 new polls, with McCain leading by 2 in 2 polls, Obama ahead by 1 in another, and Obama ahead by 10 in the last.

Much of these fluctuations are probably simply sampling error, but pollsters also can differ in how they weight their raw data.

Things are tight right now, and it will be interesting to see how the week's disruptions on Wall Street affect the next round of polling.