My eyes are glazed over having just spent an hour looking at the state-by-state polls. A number of them have moved quite marginally, but few seem to have shifted dramatically. That doesn't surprise me--while the overall viewership of the conventions doesn't look to have dropped that much, my sense is that fewer curious voters watched it. That is, the really engaged partisans (whose votes we wouldn't expect to shift) watched the conventions on the cable networks, but there was only 3 hours of the convention for the more casual, network-oriented viewer (Phil--correct me if my sense of the situation is wrong). Thus there is probably less of a stimulus here than in the past, and thus less of a response.
There has also been less of an impact from the Edwards nomination than some may have thought. But then again, we don't know about the counterfactual--how would Kerry have done in the period running up to the convention if he hadn't appointed Edwards? Perhaps the vacuum would have been filled by negative news on Kerry.
Frankly, I seriously doubt that there will be any significant event (short of a terrorist attack) that will influence the election from here on in. The voters have already absorbed and shaped their opinions on what's going on in Iraq--now that an interim government has been appointed, there's not going to be too much more to change people's minds there. Voters respond to the economy, but most research suggests that they do so based on where it was earlier in the year. The debates may have some effect on the voters who are still holding out near the end, and if one of the candidates was to really screw up, that might have some effect. But my guess is both candidates play defense in the debates, leaving it a wash. So, my best guess is that while there will be noise in the polls for the next three months, it should be recognized as that--noise.
What that means is that the real thing we ought to be tracking in turnout, and thus party and quasi-party GOTV activities. If a state is basically a statistical dead heat now, it'll probably remain so until the end (with perhaps some advantage for the challenger, so maybe we should adjust all of Kerry's numbers up one or two percentage points, which is enough to cancel out the Nader effect). So what we want to know is, in the dead-heat states, how many people does each party have on the ground? How much money is it spending? If I was trying to predict the outcome, THIS is what I'd be paying attention to, and what somebody ought to start tracking (if the numbers are available).
The final thing I'd want to see is, to the degree it is possible, some estimate of the probable Nader vote better than what we have now. There are two possible interpretations of Nader's numbers, and I can't choose between them. Interpretation One is that, with the drumbeat of punditry about how stupid it is to vote Nader, that some percentage of those actually planning to vote for Nader have just given up admitting it to anyone, thus suppressing his numbes. Interpretation Two is that Nader's numbers are inflated--when it comes down to sitting in the voting booth and pulling the lever, voters in swing states just won't be willing to risk another four years of Bush.
I'd like to think that Interpretation Two is right, but I don't. My sense of the Nader voters is that these are people who are just so alienated by the political process that they have ceased processing information contrary to their prejudices. So no matter how often they are told that a vote for Nader is a vote for Bush, and that the parties really are quite different in the policies they are likely to produce when given executive power, they've just stopped listening. So I really doubt that the Nader numbers will disappear, and there's some chance that they may actually be a little low.
Bottom line--no big, durable shifts in public attitudes from here until November. Nader voters handing Bush a couple of points in swing states. Ground game over air game. Bottom line--we need fewer polls and more sources of other kinds of information.