Today, CNN released a poll with a topline that read Sharron Angle 49%, Harry Reid 45%. Not a great showing on its face. However, I looked a bit closer at the poll and noticed this:
Registered Voters:
Reid: 48
Angle: 44
Likely Voters:
Reid: 45
Angle 49
Now, I understand that there is a enthusiasm gap and that we can all expect some level of undervoting among Democrats when compared to their Republican counterparts. However, it just seems wrong on some level to take a Reid +4 RV spread, and flip it on its head to a -4 spread. Could the composition of the voters really be that far off from the composition of the electorate?
The good news is that with the vote now approaching 50 percent of the 2008 totals in Nevada, we now have a good sample of who will be voting in this election. We also have crosstabs on CNN's results. With that, we can take a closer look.
Both GMU and CNN provide nice crosstabs that break down the results by various groups. For the purposes of this diary, I will look at the voters by party ID, and look at the CNN results for those same groups. The results show a voting population that doesn't quite match CNN's assumptions of a drastic Democratic lack of turnout.
With a count equalling 46.1% of the 2008 count already tallied, we see the following percentages for party ID in the vote:
Dem 42.4%
Ind 15.1%
Rep 42.5%
Now, in the CNN crosstabs, we see what they tallied for these subgroups among "likely voters":
Dem: Reid 96% Angle 3%
Ind: Reid 38% Angle 53%
Rep: Reid 10% Angle 82%
So, doing a little math, you get:
Reid: 50.7% (.424*.96 + .151*.38 + .425*.1)
Angle: 44.1% (.424*.03 + .151*.53 + .425*.82)
Whoah! And this is a poll that had a topline of Angle +4! What does that mean? Well...it means something very, very important: CNN's likely voter screen is not even remotely reflective of the actual voting population that is showing up. The subtext here is that all kinds of polling operations have been doing the same thing with likely voter screens, and if the change is this dramatic in Nevada, this very well could reflect a general error in polling methodology.
So there you have it. Real votes in an actual election just don't mesh with the state of the world as envisioned by this particlar polling operation. And as Nate Silver skillfully demonstrated, if polls in general are systematically off by just a few points, it totally changes the game. And just to give some historical context, the polls overestimated Republican turnout by 5 points in 1998, so it's not like this would be unprecedented. A result like that of 1998, if applied today, would make Republican gains FAR less dramatic than everyone is expecting!
Now, just to be clear, I'm not saying that this is a nationwide trend, or even true in Nevada. But it sure does give a pretty solid argument that maybe, just maybe, certain people have gotten carried away in the "Democrats won't show up!" narrative without doing the proper legwork to see if these claims would come true.