I thank Burt for bringing this up:
It’s worth bearing mind when we talk about things like “momentum” and “changing the dynamic” that since well before the conventions, Romney has never been projected to actually win. Very recently there have been a few isolated polls showing him with very narrow leads in the popular vote. But just as many recent polls give Obama narrow leads, too. And as we all know, where those voters are matters a lot. Not once have weighted polling averages put Obama behind in popular vote projections and not once has Romney been forecast to get a majority of votes in the electoral college.
I actually think that hinging predictions based on the state-by-state polls is a mistake. I’ve been meaning to write on this since Kazzy asked why we were even paying attention to the national polls. The answer to that question is that because the national polls get us the best look into the mood of the nation, and that mood most definitely filters down to the swing states.
After Romney had his bounce from the first debate, there was quite a bit of talk about how Obama was still winning the swing states and so it didn’t really matter. The swing states, however came around and it became a closer race there, too. According to Nate Silver, as of 10/12, the widest diversion between the Electoral College and National Popular vote totals was less than four percent. Memories of 2000 aside, crossed votes are an anomaly.
The state-by-state vote is worth keeping an eye on in the event of a really close election, but on the whole I would give them less reliability when gauging the dynamics and direction of the race. The national polls are ongoing by several organizations. Mood shifts are picked up more quickly there and we have more data with which to look from to see the changing dynamics. The state polls are more likely to be conducted by local news outfits and not on an ongoing basis. That’s one of the reasons that there is a lag.
In the history of voting, it’s just rare that you see national movement without seeing movement in the individual states. If you pick up voters nationally, you’re going to be picking them up in swing states. If the national popular vote is outside of 2% difference*, it becomes almost statistically impossible for the NPV loser to win. If Obama or Romney is ahead by less than 2%, as is often the case, it really becomes anybody’s guess in my view. Again, it’s worthwhile to look at the state polls, but I would definitely not be making any predictions based on them. The polls that have Obama up or down have margin of errors that usually extend beyond the 2% threshold.
Right this second, with most polls showing Romney or Obama ahead by a point or two, it’s anybody’s guess. But mostly for that reason, in my view, and not because of the state polls. If Obama pulls back some of that vote, it’ll still be his election to lose. But either way, I’d bet pretty good money that if Obama wins the EV, he will also have won the NPV (and vice-versa).
* – This is according to Silver and a college professor who more specifically said 2.2%. Had John Kerry pulled off Ohio, it would have been a far, far more remarkable event than 2000.
We’ve a respectful disagreement going on this point. I’m sure there are some numbers from past elections floating around out there comparing polling in particular critical states with national polling and the eventual results, which we could each use to arm our respective thoughts here. But for the time being I’ll leave that to Nate Silver, who gets paid by the New York Times to do that sort of thing.
Fair enough. I’ll just note as it stands that there are polls suggesting that Romney is competitive in Wisconsin and Pennsylvania. I don’t think Romney is competitive is Wisconsin and Pennsylvania. Or, if he is, he’s going to win the national vote by a pretty hefty margin and so the real swing states are not as competitive as they appear.
(Note: This does not apply to states with wicked demographic shifts or the potential thereof. We could see wild swings in states like Nevada or Arizona. I don’t expect that of states with more population stability. And, of course, I could be wrong about all of the above. I’m squishy that way.)
Will is right on this, and I don’t think Nate Silver has said anything to the contrary. Like Will said, the Electoral College only comes into play if the differerence in the popular vote is less than 2%.
With one caveat: it’s more likely now than before the first debate that the popular vote will in fact be within 2%. But not necessarily. It’s important to realize that the campaign is dominated by message until the very end, at which point it becomes dominated by logistics. Ie, the focus of the race changes from the nation to the states. But that hasn’t happened yet. We still have about a week, maybe a little less, where message is king.
And whatever message-based movement we expect to see in the next couple of weeks is much more likely to break in Romney’s favor that the President’s. That’s why IMO it’s something of a misconception to think that the race is actually tied now.
For that matter, it’s much less clear than people think which states actually matter. In particular, for Romney the state that really matters is Florida, not Ohio. There’s no way he can make up the EV’s for losing Florida, but there’s several ways around Ohio.
The democrats see things differently, of course. Their ground game starts in September, and that’s when they need logistics STAT.
The democrats have an advantage over the Republicans in terms of logistics — their voters are crammed closer together, and also, crucially, the people who don’t vote tend to want to vote Democrat.
Yeah, but most of the ways around Ohio lead through Pennsylvania, which Romney isn’t even contesting.
One thing I don’t understand–and perhaps the answer is as simple as “that’s not how you get a statistically generalizable sample”–but why can’t the national pollsters just make a note of which state the respondent resides in and then we can get the national- and state-level polls at the same time?
Different sample sizes needed.
Also, in general, different questions asked.
My governor is now less popular than Jerry Sandusky in this part of the state. You’d never get that from a national poll.
As far as sample sizes, I think I see where you’re coming from.
As far as questions, it’s a bit murkier for me. I do realize, however, that local-level politicians’ performance might affect whether or how a state resident votes in a national campaign.
I actually kinda doubt that, barring the whole “pay raise scandal” or something else that EVERYBODY hates.
I think the state-by-state poll is much more interesting than the national which is much more prone to outlying fluctuations. The right have jumped on the results of the recent Gallup poll as evidence that Romney is pulling ahead; I’m yet to be convinced.
After his performance in the second debate, with major policy details still lacking, I’m not really able to see any strategy for Governor Romney to get to the White House. Indeed I think the gap on election night will not be as small as some pundits predict.
Check out my predicted electoral map here: http://tommygilchrist.wordpress.com/2012/10/18/obamas-electoral-path-to-victory/