As Barack Obama enters the last two weeks with an average lead once again expanding into the neighborhood of 7.2%, more consideration will be given to whether these polls can actually be trusted. The lead seems nearly insurmountable with so little time left and Obama enjoying the money advantage. But is there a lurking "Bradley factor" or a a "Dewey Defeats Truman" reprise in there somewhere?
The short answer is, in all likelihood, no. The polls are probably quite accurate, and are more likely to be underestimating Obama's strength than overestimating it. If we go back to the primary season and examine the results we find that Obama actually outperformed the polls by more than 3% ten times, underperformed the poll's predictions by more than 3% six times, and ten other times met the polls' expectations when the margin between himself and Hillary Clinton fell within 3% of what the polls predicted.
So that's 73% of the time meeting or exceeding what the polls predicted and only failing to do so 27% of the time. And of the six times the polls overestimated Obama's performance, the only two that incorrectly chose him as the winner happened quite early in the primary season. Clinton won in New Hampshire and California, where polls had shown Obama 8% nad 1% ahead respectively. After Super Tuesday, February 5, that never happened again. The pollsters were able to improve their models and devise more accurate turnout projections for various types of voters after that. In general, (14 out of 16 times) when the polls were off by more than 3%, Obama wound up simply winning by more than already predicted or losing bigger in states where he was already predicted to lose.
As far as thinking of a shock such as 1948, Dewey led in the last Gallup poll by 5%, conducted eight days before the election. Truman would up winning by that same margin, 5%. George Gallup's numbers were probably about right on October 25, and his was the only extensive polling being done. Based on his experience in the previous three elections he felt another survey so close to the election would be unnecessary. He therefore missed the late surge to Truman.
But today we see no signs of a McCain surge, though numerous polls are being taken. A tightening of the race, noted here October 19, has apparently been reversed, as Obama's aggregate lead has gone up from 5.3% to 7.2% over the past three days. McCain needs a seismic event to recover the kind of momentum he would need to make up that kind of deficit, for he now needs to pull off a consistent shift of half a percent a day every day until the election.
The short summary is that Obama looks very, very strong at present. An unanticipated Bradley factor shift away from Obama or massive reordering toward McCain is not currently happening. Meanwhile, the days tick off the calendar one by one. Yes, we should believe the polls.
3 comments:
There's another factor I'm concerned about: some type of election fraud. I strongly believe that this was a decisive factor in the last two presidential elections, in Floria and Ohio. I have also heard several people talk about how the touch screen and even paper screening systems are every bit as vulnerable to tampering as they were in the past. On the other hand, I also believe that the best insurance against this happening is for Obama to have such a strong lead heading into election day, that any fraud resulting in a McCain victory in a key state such as Virginia or Colorado would be more obvious. I wonder what your thoughts are about this.
Cameron, I agree the best defense is a strong lead. Based on recent history there is reason to be wary of chicanery. But I'd contend the amount that can be gotten away with is relatively marginal. A strong lead and vigilance should be enough to prevent it actually changing the winner. Thanks to its remarkable organization and flush financial condition the Obama campaign seems to be as well positioned as possible to combat and overcome shenanigans. Its legal teams are numerous and active. The recent case in Ohio short-circuting GOP efforts to purge voter rolls is a case in point. This campaign will do a much better job on this that the previous two. See this: http://news.yahoo.com/s/afp/20081017/pl_afp/usvotefraud_081017194202
On an education-related note, Mike Rose, the noted teacher and author, had an article in the Pittsburg post Gazette on Sunday, Rosie's Dream that is a well-reasoned and succinct analysis of the candidates' education policies. Obama's policy, it turns out, is 33 pages, McCain's is 4. You can probably see where this is going, but it's a great, even-handed read and I recommend it.
Post a Comment