Yesterday’s polls pointed toward a continued divide between the state and national surveys and a persistent Romney bounce showing few signs of fading. Just look at the chart: the whole right column is red, showing Romney making gains since pre-debate polls, as is most of the “margin” column while you read through the national polls. But the remainder of the "margin" column is mainly blue once you start entering the state polls.
One point that wasn’t mentioned in yesterday’s article on the battleground states: While Romney has gained more in the non-competitive state polls than the battlegrounds, the non-competitive states still haven’t shown as much movement toward Romney as the national surveys, which might suggest that the divide is something more of a state poll versus national poll issue than a battleground versus non-competitive state question. On the other hand, a handful of recent polls in California and Massachusetts hint at the possibility that Obama suffered outsized losses in the deep blue states, including yesterday’s poll of California by SurveyUSA. There just aren’t enough polls in the truly non-competitive states to be sure.
Yesterday’s Mason Dixon poll in Florida received quite a bit of attention, at least on twitter. Yes, outlying results are flashy and interesting, but they’re often flashy and interesting because they're not especially representative. This winds up adding to the perception that the polls are neurotic and broken as readers try and reconcile “Obama+6” in Ohio with “Romney+7” in Florida. In both instances, the average of polls suggests that the race is closer in both states than those two polls suggest, even if the leaders of those two polls appear to hold an edge in both states.
Support thought-provoking, quality journalism. Join The New Republic for $3.99/month.
There is, however, one type of survey worthy of the scorn increasingly heaped on pollsters: the post-debate flash polls. The CNN flash poll, for instance, interviewed just 381 registered voters who watched the debates. So not only was the margin of error quite large, since 381 voters is only about 40 percent as large as the average national poll, but the survey only considered an unrepresentative pool of “debate watchers.” These polls will be able to catch a blow out victory or defeat, but these are not especially precise surveys.