Poll Dance, 10/29 Edition

In last month's polling update, I coined something I like to call the Rule of Fifty. (Scroll down to the previous post if you have extra time coming out of your ears. Ew. On the other hand, yay, extra time!)

According to the Rule, a candidate's 50-47 lead in the polls is worth measurably more than a 44-38 advantage, even though the latter lead is twice as big. People change their minds less easily these days and the undecided tend to break pretty evenly. Even getting to 49 is a good sign. But 50 is a nice round number.

But a single score in the 50's is hardly significant. Polling companies are biased, imperfect, and subject to statistical variation. It takes multiple 50's showing up in a candidate's polling portfolio before you should consider it a good sign that the electorate has settled on its guy. In that state, at least. National surveys are nigh useless in our arch-undemocratic system.

So, it is with plenty of confidence and a healthy amount of angst that I bring you the most pertinent polling figures I could find. All are recent: the oldest survey dates from three weeks ago, and most of the results are from within the last week. (I get the numbers from here.)

Look at some digits from Virginia, with the President's number listed first:

51-47. 51-46. 50-43. 51-46.

But, at the same time, one notices:

50-48. 50-47. 51-44. All for Romney. The state is not settled at all, and may in fact come down to third-party votes or Hurricane Sandy's yet-to-be-determined impact.

Then, Ohio, where Romney hasn't posted a single 50, EVER, so all of these are Obama leads:

51-47. 50-49. 50-45. 51-46. 50-46. 50-46.

Of course, the Rule cuts both ways. Florida is on its way to choosing Romney:

51-46. 50-49. 51-44. 51-47. And nine more 48's and 49's. All those numbers come from right-leaning pollsters, but the left-leaning ones have yet to hand Obama a single score of 50 or higher this fall.

With that in mind, I prepared a little electoral map. For each swing state, I applied the Rule of Fifty, granting the state to the candidate who crossed that threshold the most often in the past three weeks. Then I gave NH and VA to Romney, granting the tie to the challenger in both cases. Why not? His people are very enthused about removing the President from office. Maybe you've noticed. (Sometimes there's frothing involved.)

The resulting map:

Link to 270-to-win.com

Obama 277, Romney 261.

That map isn't my official prediction. (I do think CO and NH will go blue, and VA will be too close to call on election night.) No, it's just what the polls say, today, with a merciful 198 hours left until polls close on the West Coast.

P.S. -- My experience in manic poll-watching has taught me one thing, which forms the backbone of the Rule. Any single poll is insignificant. Aggregating the surveys is the key to reading them well. This is in stark conflict to over-reaction and sensationalism, which happens to be the key to TV ratings. A Gallup survey that shows Romney up by 6 or 7 is much more advertising-dollar-friendly. A single score in the 50's is interesting, but not meaningful. Several 50's: now we're talking.