Drudge and other outlets are selectively hyping the handful of polls showing a relatively close race. That’s not surprising given their incentives. The problem is that if you draw enough random samples and you throw in house effects and questionable likely voter screens, a poll somewhere every day or two will show McCain within a few points. But the overall picture of the election is quite clear, as UNC’s Jim Stimson notes, and you should therefore disregard the outliers (key sentence in bold):

This is a race of considerable variability in various organization’s estimates of what should be the same quantity. And at the same time I have never seen such stability in my estimates of the daily lead. A typical day sees about ten organizations report an Obama lead varying between 1 and 14 points. Thirteen points difference is a lot, more than double what would be expected from sampling fluctuation alone. This arises chiefly, it appears, from two sources, (1) initial assumptions about the partisan makeup of the electorate, and (2) varying likely voter assumptions. Both are probably more ambiguous than usual this year. Democratic party identification is trending upward this year. In that context it is harder than in a more stable environment to know what the right numbers are. Any old assumptions will very likely be wrong, as would the practice of just forcing the sample to have equal numbers of Democrats and Republicans, which used to be employed, and maybe still is, by the Battleground poll. Turnout projection is especially difficult this year because the Obama campaign intends to turn out huge numbers of young and African-American voters who are traditional nonvoters. A model that assumes traditional patterns will be wrong if the campaign succeeds and right if it fails. Figuring out the quality of the two sides’ ground games is beyond the normal expertise of pollsters.

Stability: When all the polls are combined to form daily estimates, it is the opposite fact that is most striking. Despite all that daily variation, Barak Obama has held a lead over John McCain of about 7 points over more than a month with virtually no daily variation. In my metric of the two-party vote division, the Obama lead of about 53.5 is just locked between 53 and 54 day after day after day. The organizations that do really large samples are reporting the same fact, remarkable continuity of day to day estimates, as if the race has been frozen since late September. Tracking polls with smaller samples are reporting trends, back and forth, which, while entertaining, appear to be quite false.

Stimson’s estimate, which is currently at 54% of the two-party vote, mirrors Charles Franklin’s estimate of 54% at Pollster.com. Both are quite close to the median prediction of 52% from leading political science models.