Quote:
Originally posted by Mmmm, Burger (C.J.)
If Kerry leads by 8% and the margin of error is +/-3%, and he wins by 2%, then the poll is "wrong" statisitically, but practically is irrelevant, as he still won.
|
Not in the sense she means. If people say one thing to the pollster, and then vote differently, it doesn't mean that poll is wrong, it just means that the poll and the vote are measuring different things. (I think this answers Hank's point too.) She means that the poll is not accurately measuring how the population will answer the poll.
As for Club's point about Congress, congressional districts are so gerrymandered that I'm not sure there's a whole lot of connection between that question on a national level and the breakdown of seats. For more on this, here's a post from Kevin Drum from a
a few days ago:
Quote:
HOUSE HORSE RACE....I just talked to polling guru Ruy Teixeira to find out what he thought about the the LA Times poll showing Democrats with a huge nationwide lead over Republicans in congressional races (see previous post for details). He made a few points:
- His rough guess is that the real Democratic lead right now is 10-11 points.
- However, historically Republicans always outperform the congressional polls. So if the election were held today, the actual Dem lead would probably be several points lower.
- The conventional wisdom is that any seat that was won with a majority of 55% or more in the previous election is not in play. However, if there's a real surge for the Democrats, some of those seats could turn out to be contestable after all.
- There are models that predict House results based on nationwide vote percentage, but it's too early for any of them to be useful. Still, he says, "there's a real chance the Democrats could pick up a bunch of seats."
Well, it may be too early for pollsters with genuine models to make any guesses, but since I have no reputation at stake it's not too early for me. Take a look at the results of the past three elections:
- In the 1998 congressional race, Republicans won 50.3% of the vote nationwide and 51% of the House seats.
- In the 2000 congressional race, Republicans won 50.1% of the vote nationwide and 51% of the House seats.
- In the 2002 congressional race, Republicans won 52.5% of the vote nationwide and 52.7% of the House seats.
I may not have a fancy computer model, but those results seem strikingly consistent to me. If the Democrats can produce a 2-3% nationwide lead in the congressional election this year, it seems to me they might be able to produce an actual win in House seats as well — although a win of uncertain size due to gerrymandering and other uncertainties. You heard it here first.
|
