Today I wish for fresh data from Missouri and North Carolina. In one case, you get a vote in whether a survey is done!
Nate Silver points out that PPP is asking people what state to poll in. Missouri is one of the choices. Vote for your favorite!
Today’s histogram is very spiky. This is a symptom of only a few races being on a knife edge. Right now is an exceptional case because Missouri and Indiana both have 11 electoral votes. The highest peak contains the possibility that Obama and McCain split Indiana and Missouri, in either direction. One permutation is more likely but they’re both in there.
Debate #2: Obama won again. I found it a relief to hear about issues, more or less, for 90 minutes. As expected, Fallows is very good (other pundits; update – the conservatives speak). And now back to the awful endgame.
All that more polling will do is reduce the number of spikes in the histogram, which is a statistically correct input of the polling results.
However, what happens when the weighting factors for inputs, such as party affiliation and age, are misleading, or the turnout bears no relation to expectations? Anticipating turnout is what skewed your results in 2004, and also skewed results in 2000. GOTV vs. Sit This One Out might affect outcomes more than expected.
Your poll model abstracts away the weightings. This, by itself, doesn’t matter if they’re wrong, because they can be dialed back in later. An interactive map that dials in GOTV or dials out sit-homes would be much like the one you have already, albeit rather than a vote going from one voter to another it would be one vote additional vs. one more taken away.
This could expose the potential for GOTV or the looming results for stay at home in places like Texas, which are hard red in your tables.
Criminy, my grammar.
Sam’s analysis is a statistically correct interpretation of the input polling results.
My question is: what happens when the input weighting is wrong?
No one has commented so far on Tom Brokaw, so I will: He was awful!
I don’t mind that he asked his own questions–as moderator he’s entitled–but as moderator he’s also entitled to cut off a candidate that’s running off at the mouth, as Gwen Ifill did in the VP debate, and she only had to do it once.
Brokaw chose not to do that. Fine, but then there’s no sense in constantly whining about candidates running over time. I hope he was embarrassed that McCain had to actually try and do his job for him by suggesting a wave!
There were two poor performances last night, only one of which was from a candidate.