Polls! Polls! Polls!
(This started as a quick update, but slowly evolved into the monstrous analysis you see before you. Read at your own peril.)
This is one of those moments where I’m going to say what I think, which has some factual basis, but could also be wildly off the mark.
I take in all of my daily political news at RealClearPolitics.com, and you should too. You can literally get everything there. But, back to the point at hand.
RealClearPolitics, to get a sense of the relative position of each candidate, simply takes all poll results and averages them out. Using this method, they currently have Obama ahead about 6%. My sense is that this number is not representative of his current lead.
I believe two things about this race right now:
1. Almost all of the polls are overstating Obama’s lead, but...
2. Obama is, in fact, in the lead by a few points.
Just as a bit of background information for anyone not in the know, election polls in general never simply sample 900-2000 people and report back their answers in simple percentages. And for good reason- the results wouldn’t be very informative if that’s all they did. Say they called 900 people, and 500 said they’d vote for Obama, 300 for McCain, and 100 hadn’t decided. Just taking that at face value would mean Obama would be leading with a whooping 56% of the vote to McCain’s 34%, with 12% undecided. However, if 200 of the people surveyed aren’t registered voters, and 150 of those wanted to vote for Obama, that would skew the result- taking them out, the result would be Obama 50%- McCain 36%, with 14% undecided. The resulting survey would therefore tally the registered voter preference; this is the minimum criteria for any poll measuring the electorate.
However, because it’s a fact that a great many voters don’t actually bother to vote, most polls try to measure likely voters, and this is where things get tricky. Pollsters can have entirely different ideas of what constitutes a likely voter. When you look at the Gallup poll, for instance, you see that using their traditional likely voter model, Obama’s up 2 to 3 points, but with their expanded model, he’s up 6 or 7.
There’s no doubt there will be additional voters this year, so I don’t think either of those results are totally out of left field. However, you’ll notice that, above, I said the RCP average showed Obama up by about 6 points. That’s because there are a great many polls out there that show Obama ahead by something like 12 or 15 points. It’s those polls that I have the real problem with.
Again, I could be totally wrong, and this election might end up seeing a huge shift like that. But I doubt it- the last several elections have been decided almost 50-50 down the line. It’s likely, given the country’s mood, that Republicans have suffered defections and Democrats have gained, but I still don’t believe that represents a 15% swing to their side.
So, back to those polls that have the huge Obama numbers. Aside from kicking out non-voters and trying to determine likely voters, many polls actually weigh their sample to try to make it as representative of the electorate as possible. This means that, if a certain pollster finds out that in his poll he sampled 300 Republicans, 300 Democrats, and 300 Independents, he might actually give more statistical value to the Democratic replies, since there are more of them in the electorate. While this is done with accuracy in mind, I don’t need to tell you it can also lead to all kinds of problems.
For instance, poll respondents can and do lie, for any number of reasons. If I wanted to attempt to skew a poll in a small, personal way, I could tell the pollster that I was a Democrat who had been leaning Obama but had just decided to switch my vote to McCain. This would still reflect my absolute preference (a vote for McCain), but because the pollster weighs Democrats more heavily, my answer has a disproportionate effect on the survey. This example would have the effect of inflating McCain’s numbers, but the converse is equally easy to imagine. Many Republicans switched their affiliation to Democrat during the primaries to prolong the Democratic race. Many of those people, despite being Democrats officially, would probably identify themselves as Republicans to a pollster, and therefore their candidate preference would be reduced in weight in the poll.
Those two things, likely voter model and weighting, are two examples of things that, I believe, are inflating Obama’s numbers in some of the outlier polls. I also think there are several other methodological problems with some of them that are as yet unknown, as the pollsters in question won’t release their polls’ methodologies. But the point is, if you have two polls, one of which shows Obama 12 points ahead and the other of which shows 2, somebody’s wrong.
Why does all this matter? Well, if Obama is only up a few points, which is the case in many battle ground states, a late surge to McCain could very possibly overtake him. And lest you dismiss that possibility, it’s been known to happen to the frontrunners. Bush ended the 2000 campaign up about 6 points in Gallup, and a late surge to Gore nearly cost him the election (it did cost him the popular vote, and he barely squeaked out the Electoral College). Carter ran ahead of Reagan until the very last week of the campaign, and Reagan crushed him nationwide. In both of those cases, an outside event tipped the leaners into one column or the other (Bush’s 20-year old DWI made the headlines the last weekend before the election, and Reagan out-debated Carter). Given that the race is tightening in the last week, I’d say a major embarrassment from Obama’s past does have this possibility. (UPDATE: This Obama video that the LA Times is hiding in an oh-so-nakedly partisan way might be just the thing)
One final thought: If the unthinkable does happen and Obama loses… well, let me just paraphrase Gandalf. “The Left’s wrath will be terrible! Their retribution, swift!” America will be denounced as racist eight ways from Sunday, and there will be endless allegations of electoral fraud on the part of Republicans (literally, endless. You’ll be hearing about it 50 years from now). Hearings will be convened to figure out why polling organizations could possibly be so incredibly wrong, and legislation will be proposed that any public poll has to release its methodology.
And there’ll be riots. Oh, yes, there will be riots.




OK, the restaurant had a name, I just can’t recall what it was.
As I guess most readers have picked up on by now, I recently got a new phone when my old one disappeared under mysterious circumstances. It was time anyway, as the old one had a bad crack in it, flipped back and forth freely with no resistance, and randomly powered off for no apparent reason (but I’m my father’s son, and I’d of kept using it anyway if I hadn’t ended up losing it). This new one has a 2 megapixel camera and 



