The state polls that have come in, post-debate, are not showing real post-debate movement for Kerry. There appears to be 10 state polls are post-debate. I’ve looked at the ten post-debate state polls and none show much of a change, and certainly no significant bounce for Kerry. Of course, the individual state’s post-debate polls themselves are few in number in addition to being very limited state wise.
Other good sites not yet mentioned in this thread are here -
And here -
Well, I don’t know if you could say the 10 state polls would really show a "post-debate"bounce, if you’re referring to the ten at Electoral Vote.
Of the ten, Alabama and Oklahoma are solid red and wouldn’t reflect any Kerry bounce.
Swing states include Pennsylvania, Florida, Ohio, Michigan, and Minnesota. Of those five, only Florida shows any poll that was done exclusively after the debate. The other four states only reflect a Rasmussen “rolling poll” that began on September 26, the Sunday before the first debate. And even Florida includes that poll in their total, averaged with a Survey USA poll done October 1-3.
The other three states at Electoral Vote are Nevada, New Jersey, and New Hampshire; they all show polls done exclusively post-debate and don’t show any real surprises–Nevada and New Hampshire have nine EVs between them and are leaning Bush (like they did in 2000), and New Jersey is moving strongly back into the Kerry camp (it always was except for one Survey USA poll done earlier this month; it voted for Gore by 16 points in 2000).
Another anecdote regarding the debate is found at Race 2004: "Yesterday I added a Granite Street poll in New Hampshire that was taken partway before the debate and part after. The overall poll has Bush leading 50-45. However, the pre-debate numbers have Bush up 51-43 and the post-debate numbers close the gap at 48-46. Admitted, these have a rather large margin of error attached to them. But it does look like the debate did bring about a change in his support.
Bottom line: There hasn’t been a lot of polling done yet that would reflect the post-debate landscape. And I’d also not put too much trust in the Survey USA poll done last Friday through Sunday; weekends tend to find younger (i.e., Demo-skewing) voters out and about and not answering their phones.
No- according to those numbers, and remembering the margin of error, the race is a deat heat and “too close to call”. When you’re “ahead” by anything within the margin of error, you’re not ahead at all.
It doesn’t follow that just because the state as a whole was solid pre-debate for Bush - that the post-debate numbers wouldn’t be affected to some extent – if those numbers were inclined to be moved by the debate. Still, those numbers show no movement post-debate.
Florida shows two polls besides Rasmussen, both show Bush maintaining the pre-debate numbers.
OK, but isn’t that the point? These states, like the ones mentioned above, just aren’t showing a post-debate bump for Kerry – no matter the spin put on it –
Or Republican moms and dads taking the SUV to their son’s baseball games or their daughter’s soccer practice. Please – we can both try too hard at this -
Perhaps this has already been discussed, but I’m not sure “won the debate” translates into “gained more support”. There may be many voters like my conservative Republican Aunt in Indiana, who easily granted Kerry the win but still doesn’t want to vote for him. One of her objections to Kerry is all the Hollywood Liberals who support him :eek: .
(I’d try to reason with her, but Indiana’s a lost cause anyway & she’s otherwise a wonderful person)
That’s not true. Say we’re talking about a 95% confidence level, since that’s standard. If you’re ‘ahead’ by anything within the MOE, you may or may not be ahead; it’s just less than 95% certain that you’re ahead.
Besides, the MOEs in races that are essentially 2-way races are the answer to the wrong question, because the MOE is for one candidate’s poll number. Here’s the formula for the standard error of a difference:
s(x-y) = sqrt[(s(x))^2 + (s(y))^2 - 2rs(x)s(y)]
where s(x), s(y), and s(x-y) are the standard errors of x, y, and x-y, respectively, and r is the correlation coefficient of x and y.
In an essentially 2-way race like we have now, r is very close to -1. So that simplifies to
s(x-y) = sqrt[(s(x))^2 + (s(y))^2 + 2s(x)s(y)]
and since the MOEs of both candidates are the same, s(x) = s(y), and so it collapses to:
s(x-y) = 2s(x).
So if you really want to be technical about it, if a poll has a 3.4% MOE, by your standards neither candidate is winning until they have a 7% lead.
But that’s not true. You never know for certain whether one candidate’s winning or not; your sureness increases or decreases with the margin in the polls - but all other things being equal, the candidate with a 2% lead in that poll with a 3.4% MOE is more likely to actually be ahead than the one who’s behind by 2% in that poll. It’s just that you’re not 95% sure he’s really ahead until the poll shows him ahead by 7% or more.
Kevin Drum posted a table on his blog about this; he got it from some reasonably heavy-hitter math/stats profs. I’ll see if I can dig it out later.
rtfirefly - everything you said is true, in a textbook. However, the difficulty of obtaining a stratified pseudo-random sample has been fraught with problems. This isn’t an objective measurement of the amount of alcohol in a Guinness; it’s dealing with people who lie, don’t know themselves, change their minds, and weren’t a good cross-section of the true voting public in the first place.
This poll, which I’ve been staring at incredulously since some SMDB’er posted it last week, claimed to have a 3.5% MOE - if we assume that’s 2 standard deviations, then Bush’s lead over Gore was more than 6 standard deviations.
Gore won the popular vote, by the way.
I think much of the variance between polls lies in the determination of the “likely voter”. Pollsters don’t simply ask if people plan to vote, because they would lie. Instead, after you agree to take the poll, a form of self selection that is suspect in my opinion, the first questions are designed to elicit a “probability” of voting. If it falls below their threshold, they discount your vote. These threshold questions will tend to underestimate first time voters, for one. Here is an article from Slate explaining the process. In a race with a large percentage of new registrants first time voters may be a swing group.
If you’re referring solely to Alabama and Oklahoma: Please. Life is too short to have to look for Democrat trends in states with Bush leads of 30% or more.
Not in my browser. When I go to www.electoral-vote.com and hover my mouse over Florida, it cites only the Survey USA and Rasmussen polls that I pointed out in my post, and it averages the numbers from those two polls. The separate page that focuses on Florida is too much of a mish-mash for me to make sense of it.
But at Race 2004, there’s another poll dated October 3, Insider’s Advantage, that shows Bush at 47%, Kerry at 45%, MOE 5%. Even though it wasn’t part of my original cite, and I don’t know the details of the latest poll, that page shows a distinct Bush decline and Kerry rise in the past week or so.
Spin this, then. New Hampshire, if you click on it on the main page, is trending upwards blue (although, Christ almighty, that page is a hard-to-read mess). Also, please re-read my earlier post, especially the section in italics.
New Jersey, as I mentioned, is solidly blue, unless you want to accept the September 14 Survey USA poll (Bush leading 49-45 in that one) as gospel, in which case you have to explain the more recent poll by the same company showing a 50-45 Kerry lead.
Actually, you’re the one straining yourself. As Apos mentioned earlier in this thread, this election is revealing some glaring shortcomings of traditional phone-based polls; weekend polling is just one such example. (And why do you assume Nevada parents of school-aged children are automatically GOP supporters?)
You can poke holes in my suppositions, because polling is an inexact science. But that’s not the same as proving the opposite. If you’re arguing that there was no post-debate Kerry bounce–and you are–then surely you could find some numbers to support your views.
They’ve already been provided Airblairxxx - but it’s clear you’d rather not see.
Update from http://www.electoral-vote.com/:
???
I already explained the 10 polls found yesterday at www.electoral-vote.com. What was there at that point that was post-debate was admittedly meager. But I didn’t find anything to support your supposition that there was no post-debate Kerry bounce, and you didn’t bother pointing out what was there.
And now that post-debate numbers are starting to come in, we can analyze this some more. As BrainGlutton has pointed out, Kerry is trending upwards since the debate in Iowa, Ohio, and New Mexico.
I’m not going to say he’s taken the lead in those states, because the lead is still within the MOE. And there’s also Pennsylvania, where Bush now leads by one in a seven-day poll ending October 3, which obviously includes pre-debate numbers. (Iowa’s and Ohio’s also use pre- and post-debate numbers as well, but the difference is that before the debate they were on Bush’s side by 4 and 3%, respectively, and now they’re on Kerry’s side. Just click on ‘Previous Day’ to see how things looked yesterday.)
Polls are imprecise, especially this year, and you can’t take one poll and assert Candidate A is the leader if his lead is within the MOE. And, of course, the only poll that ultimately matters is on November 2.
But it is safe to say that polls can indicate trends. And if you continue to claim there’s no post-debate bounce, contrary to the numbers that have come in (however imprecise and meager they are), then perhaps it is you who would rather not see.
(Of course, it could also be that you can’t take 25 minutes out of your workday to actually dig into this stuff, like I can. I’m nothing if not fair. But it’s not my fault if your office doesn’t have a door.)
Well, yeah. But that’s a whole 'nother issue. Whether we’re dealing with the amount of oxygen and nitrogen in the atmosphere, or dealing with the number of Bush and Kerry supporters - if the two things we’re estimating have a correlation near -1, then the MOE of the difference is twice the MOE of either one.
You’re still using the standard deviation of one of the two measurements as your yardstick, for no apparent reason.
But more important is the word “Gallup”. His poll has lost all integrity. Take a look at his last two polls, for instance. Gallup’s 9/26 poll gave Bush a 13-point edge among registered voters; his 10/3 poll knocked that lead down to 1 point. Did Kerry have a good week? Maybe, but not that good. The party-ID distribution in the poll went from (D-R-I) 31-43-26 to 39-35-26. Not surprisingly, when the poll went from 12% more Republicans to 4% more Democrats, the level of Bush (and Kerry) support moved radically as well.
Who knows what Gallup’s sample looked like in that pre-election 2000 poll? Garbage in, garbage out.
Gallup’s still a ‘name’ in polling, the way David Broder is a ‘name’ in the commentariat. Neither one’s worth a damn, most days.
Don’t worry, I’m not likely to forget before they check me into the nursing home. Thirty years after Watergate, I can still quote chapter and verse. And this buncha looters and thugs make Nixon and his henchmen look small-time.
Then Kevin posts a table showing the probabilities that Candidate A is really ahead, if s/he’s ahead by X% with a published MOE of Y%. The formula that generated the table was provided by a math professor and a stats professor at Cal State-Chico.
Missouri: Bush 49, Kerry 47. 683 likely voters, polling done 10/2-4.
Washington: Kerry 54, Bush 43. Same dates, 640 likely voters.
Ohio: Kerry 49, Bush 48, same dates, 761 likelies.
California: Kerry 51, Bush 43, same dates, 748 likelies.
North Carolina: Bush 52, Kerry 45, same dates, 628 likelies.
ARG:
Florida: Kerry 47, Bush 45, 600 likelies, polling done 10/2-5.
I still wish that average sample size was larger. I always like to see at least 1000.
Update: According to http://www.race2004.net/ today (10/7/04), the electoral vote tally is now 215 for Bush, 239 for Kerry, 84 undecided.
On the other hand, http://www.electoral-vote.com gives the tally as 253 for Bush, 264 for Kerry. I guess that’s because of different counting systems: race2004.com divides the states into five categories: Strong Bush, Weak Bush, Undecided, Weak Kerry, and Strong Kerry. While electoral-vote.com uses seven: Strong Bush, Weak Bush, Barely Bush, Exactly Tied, Barely Kerry, Weak Kerry, and Strong Kerry. And, as of today, electoral-vote.com counts only two states as Exactly Tied: Michigan and New Hampshire. But race2004.com counts eight states as Undecided: Nevada, Iowa, Missouri, Arkansas, Florida, Ohio, New Hampshire, and Maine. When we consider the margins of error, race2004.com’s method is probably more realistic; electoral-vote.com gives the impression we know the picture more closely than it really can be known.
The new AP poll, fresh from the oven, has a 50-46 lead among likely voters, >MOE=3%, for
[drumroll]
Kerry.
BG, isn’t the current tally on your site Kerry 253 Bush 264?? Am I not reading this thing correctly or has it changed since you last posted???
-XT
THis was interesting from BG’s site as well:
Go Badnarik!! 
-XT