30 Comments

1. George Purcell:

Another data point:

Travis County, Texas (Austin)

We have 584K registered voters.

37% of this population, 217,000 people, have ALREADY VOTED…for comparison, only 51% voted in the entire 2000 election!

If this is indicative of turnout elsewhere, then no one knows what is going to happen on Tuesday.

Oct 30, 2004 - 7:36 am 2. Caroline:

Given the lengthy wait times at the polls you have to wonder what the impact will be in terms of folks grabbed by the collar, given 10 bucks or whatever – to vote. How many will simply walk away (with their 10 bucks and no one the wiser)? Seems to me that its going to take a bit of committment to vote given these early accounts. I would have to guess that would have a bigger impact on Democratic turnout.

Oct 30, 2004 - 7:51 am 3. David Thomson:

ìIf this is indicative of turnout elsewhere, then no one knows what is going to happen on Tuesday.î

We may not know in a epistemological sense, but we can make a good guess: Republicans are voting in droves! They are enraged at the media and other liberal dominated institutions. I know of at least two Republicans who were earlier angry at the president. One was even near rage over George W. Bushís domestic spending policies. So much so, that around six months ago he even sent an e-mail urging his friends to vote against the president. No longer, he now argues strongly for our current commander-in-chief.

George W. Bush should easily win the election by a six point margin. Why do I think that a number of pollsters contend that the election is much tighter? First of all, they are likely committed Kerry supporters who are willing to delude themselves to satisfy their ideological leanings. Also, far too much importance is placed on the newly registered Democrats. I donít think they will show up on election day. This is especially true if the weather turns bad.

Oct 30, 2004 - 7:55 am 4. George Purcell:

“We may not know in a epistemological sense, but we can make a good guess: Republicans are voting in droves!”

Not in the People’s Republic of Austin!

Oct 30, 2004 - 8:04 am 5. Caroline:

David – the tight polls are actually a godsend. If it looked like Bush was clearly ahead it might encourage complacency among his supporters. The tight polls (as well as the hours long wait lines) should bode well for the truly committed. So the question becomes – who is more committed? The Bush supporters or the Kerry supporters?

Oct 30, 2004 - 8:08 am 6. Lem:

OT again

Reading the GQ Kerry interview, reminded me of Woody Allen’s Zelig.

Kerry lacks identity, ergo the flip flop flap. Other than hint’s at his continental doppelganger, there is nothing unique about him.

GQ: Since we’re in a sports bar, let me ask: Who’s your sports hero?

JK: I’ve got so many. Bobby Orr. Jaromir Jagr. Wayne Gretzky. DiMaggio.

GQ: How about Phil Esposito?

JK: Absolutely. I wore number 7 because of him.

GQ: Who’s the better band: Stones or Beatles?

JK: Um…

GQ: And don’t give me some lame political answer and say “I like them both.” Pick one.

JK: Well, I can pick both. I can tell you the truth, and the truth is I love both. I love “Brown Sugar.” I love “Jumpin’ Jack Flash.” I love, you know, “Satisfaction.” I like “Little Red Rooster.” I love every Beatles song. I mean, I love the Beatles. I love the Abbey Road album, I love the White Album.

GQ: But, c’mon?you have to choose one.

JK: But I don’t have to. And that’s the glory of life. I play them both. I do! I play them both. I’ve got them both in my car.

GQ: What about Dylan? Ever met him?

JK: I have. It was like meeting an icon. I love Dylan. He’s brilliant. One of my favorite songs ever in life is… I mean, I can name any number of his songs that I love [long pause]… but you know, Lay across my big brass bed? “Lay, Lady, Lay.”

GQ: Is there anyone you want to meet that you haven’t?

JK: Lots of people.

Oct 30, 2004 - 8:13 am 7. David Thomson:

ìNot in the People’s Republic of Austin!î

I live in Houston and voted a few days ago. My area is highly mixed, but I have been told that the more Republican precincts have seen a sharp increase in voters. Some people had to wait three hours before casting their ballot. Scholars who study voting trends argue that people vote for a preferred candidate. Voting against someone doesn’t get them as excited. John Kerry is not warming up his base. Only George W. Bush excites his supporters—and they are enraged at the unfair treatment by the liberal elitists toward the president.

Oct 30, 2004 - 8:16 am 8. Terrye:

I heard that Zogby went for Kerry already.

The audaciaty of these people. They act as if the election is a fromality, an after thought.

First bush is up, then Kerry, then Bush, then Kerry etc.

They want it to be tight so people will pay attention to them. In truth I think a lot of people are getting annoyed.

I hung up on a polster the other day. How do I know they did not say somehting stupid, like that I was voting for Kerry?

fat chance.

I live in a small twon I figure it will take me about ten minutes to vote.

Oct 30, 2004 - 8:27 am 9. Michael B:

The only pollster I even bother with, since I found him two or three weeks ago, is The Horserace Blog. The guy knows his stuff, highly recommended! Too, he explains both his and all the other prominent pollsters’ methodologies in a manner that anyone who has taken one or two statistics courses can readily understand.

He’s another great find, and another example where the blogosphere out ranks the MSM and more traditional outlets.

Oct 30, 2004 - 8:40 am 10. alcibiades:

Voting here in NYC. Usually voting takes all of 5 minutes, maybe up to 15-20 minutes for a presidential race. I expect there are going to be depressingly long lines come Tuesday, with lots of liberals making nasty remarks. I really hope it doesn’t take hours.

I’ll have to bring something to read, something to write and a coffee.

Oct 30, 2004 - 8:45 am 11. David Thomson:

If John Kerry wins, his victory will contradict the prevailing wisdom of voting scholars like Raymond E. Wolfinger:

http://www.polisci.berkeley.edu/Faculty/bio/permanent/Wolfinger,R/

The academic community will have to reevaluate what they have been saying for decades. I just donít sense that moderate Democrats are excited by John Kerry. The radicals residing in Austin, Texas are not the norm. They are members of the Michael Moore fan club.

Many of the Democrats who are poorly educated and financially unstable will vote only if they are provided with a party atmosphere. They are not the type of voter who will stand in line for three hours and endure other irritating obstacles. Am I something of a bigot? Nope, this is the consensus view of liberal scholars. They prefer, however, to keep their mouths shut when speaking in public.

Oct 30, 2004 - 8:49 am 12. David C:

One of the best insights about politics I’ve ever heard is that the Conventional Wisdom is *always* wrong. Always.

You can’t always tell *how* it’s wrong, but it always is, because CW is generally premised on the unstated underlying assumption that everything will happen pretty much the way it did in the last election. This is always pretty questionable, but I think it’s extremely questionable for this particular election, since we’ve had little things like 9/11 and the War for Civilization start between November 2000 and now.

But the CW says that doesn’t matter, and people pretty much rank “national security” and “Iraq” as just random issues in the Big National Pile O’ Issues, not all that much more or less important than school uniforms or how many shovels full of free prescription drugs to throw at senior citizen voting blocs.

I wish I had a lot of confidence about how it’ll play out exactly, but I doubt it’ll look much like 2000.

Oct 30, 2004 - 8:56 am 13. Charlie (Colorado):

Folks, if you recall I predicted Bush 55±2 percent and >300 electoral votes.

That’s my story, and I’m sticking to it.

But I’m beginning to get suspicious that I was a little conservative.

I work right near the Boulder County Clerk; there was a line that must have been an hour long at least on Friday at noon. I suppose that could be ABB voters, but I doubt it’s Kerry voters.

There are six BC04 bumper stickers in my office parking lot, and only two KE. In the People’s Republic of Boulder. Berkeley East.

All the stories seem to be saying that early voting is well above 50 percent Bush, and I can’t think of a credible reason why early voting should be massively slanted on way or the other. If we can’t argue for a specific slant in the sample, least-hypothesis demands we assume that the early voters are distributed much like the total vote.

Kerry is not only not gaining in swing states, he’s having to fight for reliable Democrat states, like Hawai’i.

Pejman Yousefzadeh, just today, says that the DNC and MoveOn.org are running ads in Chicago.

If Chicago is worrying the DNC, they’ve got internal polls on which their worry is based.

And statistically … well, here’s the RCP totals in the average today:

Bush +6

Bush -1

Bush +2

Bush +5

Bush +3

Bush +5

(I’ve made all of those relative to Bush to make it clearer; Zogby is actually shown as “Kerry +1″ rather than “Bush -1″.)

Playing bumpo-Bayesian again, what you want to ask is “what is the probability that the ‘real’ population being sampled would come out significantly in Kerry’s favor (say -1), given these values?” And guess what? The probability that one sample would be +6 Bush when the real value is -1 Kerry is around 1 in 10,000. All of them together, well, way more than one in a million.

On the other hand, the chance you’d have a value of -1, if the current ‘real’ margin is +3 and the margin of error is ±4 is only 1 in 20.

(By the way, here’s a little pop quiz. Here’s what Newsweek/MSNBC had to say today about the polling results:

With a margin of error of plus or minus 4 percent (meaning any support number could be up to four points higher or down to four points lower) anything less than a 9-point lead is a statistical dead heat. So to the statisticians and professional pollsters, BushÔøΩs 6-point lead is a tie.

Here’s your quiz question: is this correct, and if not, why not?)

Oct 30, 2004 - 9:35 am 14. George Purcell:

“Bumpo-Bayesian” I like that!

Here’s my Bumpo-Bayesian take:

Given this potential turnout, no poll that controls for party ID should be taken seriously. We simply cannot assert that the voting population in 2004 is the same as previous yesr.

Oct 30, 2004 - 9:55 am 15. Charlie (Colorado):

George — I don’t know if I’d quite go that far, but if you follow Roger’s link to the Horserace Blog there’s a lovely discussion of the issues.

It turns out the weightings have got some very odd issues, but overall the weightings are weighted toward the Democrats fairly significantly.

Actual party identification is supposed to be close to even, most of the polls — by depending on previous elections — are weighting the sample between 5 and 10 percent toward the Democrats.

This would suggest any poll results could be skewed toward Kerry by as much as 1 or 2 percent.

Just for fun, what’s that do to RCP?

Bush +6

Bush +1

Bush +5

Bush +7

Bush +5

Bush +7

That would put the (questionable validity) RCP average about +5 percent — or something like Bush 53 Kerry 47 — and my bumpo-Bayesian approach would predict something around +6, ie, 53/47 (again).

Oct 30, 2004 - 10:22 am 16. George Purcell:

Charlie, I’ve been following Jay’s blog since the second day he went live. He is really doing some top notch work there.

My point on weighting is this:

A pollster has two choices…simply report party ids in their sample or weight their smaple based on party ids.

If turnout is heavier than expected, party id weights based on historical data are unlikely to be correct, and using them will introduce a non-random, systemic bias in the analysis.

On the other hand, not weighting by party id bay decrease the precision of the estimate but will NOT introduce systemic bias.

I think we’ll find that the less precise but unbiased estimates are better.

Oct 30, 2004 - 11:18 am 17. Charlie (Colorado):

Absolutely — I’m just saying that we don’t need to completely discount the weighted version if we have some measure of the systematic bias. If my supposition is correct, then we have a systematic bias of about -2 pct in the Kerry direction.

Oct 30, 2004 - 11:21 am 18. George Purcell:

Fair ’nuff! I guess my point is that, given unknown turnout patterns, I’d be uncomfortable trying to estimate that bias.

Should be an interesting post-election project however!

Oct 30, 2004 - 11:23 am 19. David C:

The variable I always wonder about is the increasing number of people who aren’t covered by polls, and what characteristics they have.

“People who only use a cell phone” have been talked about a bit, but to me, the more interesting and probably larger group is “people who use Caller ID to screen their calls.” I’m in the latter category, and pretty much never pick up the phone unless I know who’s calling. I think this sort of thing is becoming more common. Is there a “poll of people who don’t answer polls” out there somewhere? I wonder what similar characteristics, if any, there are in this group?

Oct 30, 2004 - 11:33 am 20. Ed Poinsett:

I live in rural NE Georgia, we’ve had 4-5 hour waits every day this week. I finally gutted it out yesterday. Got to the poll at 1 PM, voted at 6:30 PM.

They locked the Courthouse doors at 5PM, but all inside were permitted to vote. Nobody left, last votes finally cast at 10PM. GWB got my vote and all the rest if you looked at the bumper stickers in the parking lot.!

Oct 30, 2004 - 11:54 am 21. Charlie (Colorado):

The variable I always wonder about is the increasing number of people who aren’t covered by polls, and what characteristics they have.

That’s always the kicker — does your sample have some kind of error you don’t account for? Is it really representative of the population that will really vote?

But I’ll note a couple of things: first, the KerrySpot a few days ago had an analysis of what the effect of different splits of the undecided would be; for Kerry to win, the undecided have to go almost exclusively to Kerry, more than 4 to 1. For Bush to win, the undecided have to go only 1 in 4 to Bush, and if they go about 1 in 3 to Bush I get my 300 electoral votes.

Second, someone last night (on TV) mentioned what they call people who are depending on undecided and unsampled voters to win.

Losers.

Oct 30, 2004 - 12:08 pm 22. Fresh Air:

David C–

Historically, Republicans have underpolled due to a dispropotionate percentage of people who are too-busy-making-money-to-answer-the-damn-phone in the GOP column. (This did not hold for some reason in 2000. I don’t know why.) The cell-phone-only bunch is an 18- to 30-year-olds’ phenomenon. Everything I’ve seen is that Bush leads among this group by a substantial margin.

Both of these factors argue for systemic underpolling as Charlie (C) noted earlier.

There is no question the turnout will be massive this year. The question is, which side will have the stamina to wait in mile-long lines for hours? This also points to a problem for vote fraudsters: How many of them will wait in line for four hours to vote twice?

Oct 30, 2004 - 12:24 pm 23. mwalls:

I live in Norman, Oklahoma. I voted during the Oklahoma vs. Oklahome State game (roll up the sidewalks kind of game) and the wait at the County Election board was 45minutes. I shudder to think what it’s going to be like Monday or Tuesday.

Oct 30, 2004 - 1:21 pm 24. Sandy P:

Aren’t there also quite a few dems, not only here, who won’t change their affiliation at least to Independent????

How does that skew the polls?

I once read after 2000 that about 15K people in Louisiana changed from dem to pub affiliation, there were others who went the other way, but not like that.

Oct 30, 2004 - 2:41 pm 25. Charlie (Colorado):

Aren’t there also quite a few dems, not only here, who won’t change their affiliation at least to Independent????

Sandy, you’re exactly right — in fact, that’s one of the things that may be confounding these models, as George and I were talking about. Historically, not too many years ago, there were more Democrats than Independents, and more Independents than Republicans. Now, Republicans are tied with Democrats, or even ahead a little.

Being appropriately conservative, the pollsters’ models often reflect the previous election — but there is very good reason to suspect that registration has moved toward Republicans. And (with no data at all, just a feeling) I suspect that someone who changed their registration to Republican in the last four years is going to vote for Bush this week.

Oct 30, 2004 - 3:28 pm 26. Rick Ballard:

Charlie (C),

It could be substantially worse for the Dems than appears at the moment. Using the actual ‘00 turnout numbers for each state and current polling percentages for each state The Kerry numbers come out at 3.2 million less than Gore received in ‘00. W is up about 320,000 using the same method for a gross change of over 3.5 million. I know the model is simplistic but it does provide another perspective.

Oct 30, 2004 - 3:57 pm 27. Charlie (Colorado):

Rick, I’m with you. It all sort of feels like the ‘94 election to me — you remember, the one where the R’s won the House and Pataki beat Cuomo?

I don’t recall the polls even suggesting that — although I was out of the country and wasn’t seeing the press as much.

Oct 30, 2004 - 5:05 pm 28. Rick Ballard:

Charlie, I think ‘02 was closer to ‘94. There have been 3 landslides since ‘60. ‘64 for Johnson was primarily a reaction to the Kennedy assasination and some great campaigning against Goldwater. 7 million fewer Reps showed up for Goldwater than did for Nixon in ‘60 and 8.6 million more Dems showed up for Johnson than did for Kennedy. In ‘72 15 million more Reps showed up for Nixon than had shown up for him in ‘68 and 2 million fewer Dems showed up for McGovern than had shown up for Humphrey in ‘68. In ‘84 10.5 million more Reps showed up for Reagan than had shown up for him in ‘80 and 2 million more Dems showed up for Mondale than had shown up for Carter in ‘80.

Thematically, this race is closest to Nixon/McGovern. We are at war, there is a clear choice in visons for the future, there is a noisy anti-war, anti-Bush crowd that has succeeded in infuriating most moderate conservatives and Kerry is a MA liberal to the left of Kennedy. The extremism of the BDS crowd has undoubtedly sickened moderate Dems and will significantly damage Dem turnout. The Acorn/ACT/Moveon morons have had all the impact of the Deaniacs in Iowa – which is to say none if not negative. Putting Moore in the box with Carter had to have been as stupid a move as Dukakis in a tank. Letting Sharpton into the primaries offended a fair portion of the Dem’s black constituency as well as an equally fair share of their Jewish constituency. Tack on the 1.4 million more Dems lost to the disparity in mortality demographics and I believe there is a fair chance that we are looking at a 44-46 million cap on votes for Kerry and 58-60 million votes for W. That would give a range of 56/44 – 58/42 which is within the range of probabilities that both you and I have been stating for six months.

And I didn’t even use any of those nasty formulas where the sigma gets tired and has to lay down.

Oct 30, 2004 - 7:02 pm 29. Charlie (Colorado):

Andy Card — the President’s Chief of Staff, so consider the source, but still — just said on Fox and Friends that undecideds are breaking 60/40 to the president. Now I can’t find the post at NRO, but there was a tabulation of what would happen at various different break ratios, and 60/40 for the President would imply a blowout.

Oct 31, 2004 - 5:08 am 30. Charlie (Colorado):

Wow, and you thought I was being optimistic.

Oct 31, 2004 - 6:02 am

Write a Comment

Name:	(required, displayed)
Email:	(required, not publicized)
URL:	(optional, displayed)
Comments: