It appears that the tie I reported before was indeed a fluctuation. Seven national polls conducted September 8-10 give McCain, once again, a median margin of 2.0+/-0.8% over Obama. Applying this as an adjustment to the EV estimator above gives a current estimate of Obama 238 EV, McCain 300 EV. The 95% confidence intervals are Obama [200-270] EV, McCain [268-338] EV. McCain is ahead but it’s close.

This is the measure of where things really stand – not the top-line estimate above, which still includes many pre-convention state polls.

The polls (Pollster, RCP) come from

• Gallup Daily Tracker (McCain +4%)
• Fox/Opinion Dynamics (R) (McCain +3%)
• Democracy Corps (D) (McCain +2%)
• Diageo/Hotline Tracking (McCain +2%)
• Rasmussen Tracking (tie)
• InsiderAdvantage (tie)
• DailyKos/Research 2000 (D) (Obama +2%)

Update: adding DailyKos/R2000 increased the error bar without changing the median.

As state polls pour in, the unadjusted, state-polls-only estimator is catching up with the adjustment. This is important because we do not know where the shift occurred. For example, if the change was in strongly Obama or McCain-leaning states, the true estimate will be closer. The converse is true as well.

By the end of the weekend we will have a precise state-level view of where the candidates stand in the Electoral College.