Posted on 10/16/2006 5:13:34 PM PDT by AntiGuv
The pollsters have been struggling with this problem for a while now. SurveyUSA did pretty well last election using automated polls. I think it requires a very good statistical model of the respondents.
But that model would have to change this year to reflect changes in cell phone usage. Every body uses more cell phones now. Many poor Dems use the cheaper cell phone these days (but how many poor Dems vote?).
I take all polls, and especially automated polls, with a grain of salt.
But these Survey USA polls are not exit polls
I understand that, but it is still a poll which have gone horibbly wrong.
BTW, how come the polling companies won't disclose their actual response rate?
And your proof is media polls? I'd rather trust a Ouija board.
In the final analysis, that is my view as well.
Yet I just get the sense that some event will occur that will reverse the current trend, and we'll hold the line.
A tens of millions funding advantage for the GOP in the final weeks certainly won't hurt, and that's what I think may very well turn the tide before Nov 7.
We'll certainly have an answer soon enough!
Even if this poll is accurate at this moment, I still think that Heather Wilson can turn it around. She needs to go after Patricia Madrid on the state corruption issue.
I agree with you and actually that surprises me most of all about Madrid being ahead, because Wilson has already been hitting that angle. It's a somewhat ominous sign that at least Dems are pretty much ready to ignore anything negative about their candidates (Menendez is another good example) due to their being so implacably hostile to the GOP this year. As I noted previously, the most consistent message coming from the polls right now is that Dem crossover is at a minimum.
No, I thought Wilson would lose (51-49, mind you) before I heard about this poll. Call it a gut feeling. But I'm certainly not giving up on the seat.
Deja vu.
I'm guessing an outlier.
Races don't flip 13% in one poll when nothing significant has happened.
I don't know much about the state corruption issue. You have any more info?
IN counts its votes faster than other states. We'll know how bad its going to be Nov 7th by looking at the 3 contested IN seats. If we go 0 for 3, then we lose the House. If we go 3 for 3, then we keep the House with a single digit loss.
If we go 3 for 3 in Indiana, then we're probably taking seats in Georgia and Illinois to compensate. Hostettler's polling is abysmal (yes I know about his past races) and Chocola's not doing too much better. Sodrel still looks ok if the night isn't terrible.
Exactly what I was thinking. My rule of thumb is if it is a poll shows a 10 pt lead for the Dems it is dead even.
I believe that you believe that what you do is a "science".
And I believe that you are serious about your work.
But let me tell you what happened to me the other day.
Honest to God's truth.
A pollster calls me on the phone and says,
"Are you pleased or displeased with the job President Bush is doing?"
I said,
"Well, that is a pretty complex question to answer with a simple 'yes' or 'no'."
He said, "Thank you", and hung up the phone.
Now you tell me.
Were his results valid at the end of the day?
SUSA is overweighting Democrats and are therfor worthless. Rasmussen and the others are alright.
The funny thing is that this has been the battle cry for almost a year now. Nothing much has changed.
The state Treasurer was forced to resign on extortion charges, and Madrid may have ignored evidence against her fellow Democrat. It was the Feds who prosecuted him, not Madrid.
I'm going to reply to you, who posted the thread, rather than to everyone individually.
First of all, Dane's obsession with weekend vs weekday polls is misplaced. We have literally thousands of polls' worth of data -- we have always polled weekdays and weekends. We do too many polls to be able to squeeze them all in on the same day of the week. Our sampling methods minimize the differences between weekday and weekend polling. The effects we saw today are much too large to be explained away like this.
Second, criticism of our polls for being automated misses the point -- we conducted the polls in each of the 4 districts the exact same way in both September and October, any "house effect" would affect both polls and not explain the large movements.
Third, we weight our samples to both census data and voter registration data. Different districts require different sampling and weighting methods, depending on how gerrymandered they are and how racially homogeneous they are. ("Random Digit Dial" is ideal, and is the choice in almost all statewide polls, but is not always possible in gerrymandered districts, and even RDD polls still get weighted to Census data to compensate for women being more likely to answer the phone, old people being easier to reach, etc.)
Fourth, our TV station clients ask us to poll particular races, but we are their experts and we schedule when the polls are in the field and write all the questions ourselves. Their only contribution to the process is to say which races they want to poll (and we may decline to poll exceptionally difficult contests). We may offer them a choice of days to have polls delivered, or consider their requests, but never at the expense of accuracy. (For example, we won't poll over holiday weekends because so many people are away, or on Yom Kippur in geographies with significant Jewish populations because many people we call would be annoyed or not answer).
Fifth, we are fully aware of all the current research on cell phones, caller ID, and voice mail (and do research of our own on these) -- they present challenges, but the track record of PRE-election polling (not exit polling) in recent years shows that accurate polling is achievable most of the time.
Sixth, although "races don't flip 13% if nothing significant has happened" when 4 out of 4 races show similar moves, we think something significant HAS likely happened, and the evidence that the poll is not a fluke is that the same thing happened elsewhere.
Finally, we certainly do not shift our results in order to please our clients, then give the "real" numbers in our final pre-election polls. There are frequently nationwide trends in the final week of a campaign, and sometimes one party has a big GOTV (get out the vote) advantage which will show up in the last few days, but we call them as we see them at the time our polls are taken.
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.