If an internal Democrat memo can be trusted, Arizona is the clue that this is going to be a significant reversal of power in US politics. The memo is http://www.dscc.org/img/AZearlyvote.pdf (look out, you all, it’s a PDF file).
According to the Democratic Senate Campaign Committee, approximately a third of Arizona’s electorate has already voted. And their tracking poll indicates that 44% have voted for Pederson (D), 40% for Kyl (R), 4% for other. The remaining 12% declined to say for whom they were voting.
Now, let’s analyze and add caveats. First big caveat – tracking polls can be off. However, Pederson was supposedly running about five points behind Kyl, not neck and neck. So the normal 3 to 4 percent error already says things are not good for Kyl.
Secondly, the 12% ‘no declaration’ need examined. Historically that margin breaks one of two ways in almost all cases (there are rare exceptions). First, and most usually, it reflects the known voter pattern within a percent or two. They’re just telling the poll takers that their votes are their own business. This leaves Pederson with an edge. The second most common effect is that the voters are embarrassed by their votes – they voted against type and don’t really want people to know. This is encouraging to Pederson as the state is nominally Republican and Kyl is the incumbent. So the 12% is either mirroring the known ratios or it’s making the upset even larger.
Again, the big thing is that Kyl was supposed to have a small but comfortable lead – 5 or so points. If he is running neck and neck, that raises questions about the polls in every other state with a race.
Ask a simple what if. What if every race’s polling algorithm that identifies likely voters is off, skewed wrongly toward the Republican candidate? What if it’s only two points – the max swing of margins of error for the combination of previous polls and the vote tracking polls? What if it’s only three points, or five points, both compromises between previous lead and current tracking?
What if the error is the full nine points? What does that come up with in the upcoming partitioning of Congress?
I think we’re seeing a tsunami. I don’t think nine points is a nationwide algorithm error. But neither do I think the models are right. I think – maybe – that slipping the numbers by two or three points might not be out of line. We’ll see.