If you were a bookie... Polls and bets on the 2004 elections

Kerry is bidding higher on Tradesports.com than he ever has - something like 48.

Bush was at 60 on Saturday. Now he's at 51. Your trends don't look too good for the pres. as we pull into the home stretch....

Kerry was at 40 on Saturday. Now he's at 48. You tell me who is doing better.

Cycloptichorn

realjohnboy wrote


I don't think they could so long as Kerry's main thrust was being a military guy who is tough on terrorism. Edwards voted against every single initiative re the military, the war on terrorism, the invasion of Iraq, etc. and he never served in any kind of military capacity. So the southern charm and oratory skills suddenly became a liability in many places.

I just saw John Zogby on the daily show tonight. John Stewart asked him, "Who is going to win on Tuesday?"

Without hesitation, Zogby's answer was...JOHN KERRY. Case closed. Pack up the Bush plantation. It's over.

Well, Zogby isnt the pollster most universally respected for a lack of bias ...

He called this race "Kerry's to lose" a long time ago already, and was widely scorned for it. So he as a 'vested interest', so to say.

On the other hand, gotta agree with Cyclo. If, like Moishe says, the highest Bush has "gone for" on Tradesports was 66.6, the lowest 48.8 and he is now at 51.2 (close to his lowest point), that doesn't look too good for him then, does it? With Kerry up from 40 to 48 since Saturday ... has he got the all-important last-minute momentum?

RealClearPolitics, with this morning's Reuters/Zogby 47-ea tie factored into the 3-way shows a 2.3 Bush lead, 48.6 - 46.3, while Head-to-Head, its a 3 point Bush lead, 49.3 - 46.3. Looking back through the 3-way results, Kerry has not held an overall lead since mid-Summer. More recently, in the 50-some polls reported since the beginning of September, Kerry's best individual poll results have been one 3-point lead, one 2-point lead, and 3 ties, all of which were at odds with the majority of the corresponding polls and none of which were sustained in subsequent reportings of the particular polls. Looking at the histogram in the bottom panel of RCP's chart reveals that Kerry has managed only sporadically to achieve a lead of as much as 2 points, and has been unable to sustain that 2 point lead longer than a single week. On the otherhand, Bush has held an average lead of greater than 2 points since early September.

Interestingly, Bush's weakest days have been those which included samples drawn Saturday and Sunday, typically posting his lowest performances on Monday mornings, with Tuesday results showing a rebound. Pretty much the same phenomonon is reflected both in the daily Rasmussen daily tracking, and by the WaPo/ABC daily tracking, with Bush's best performances clustered in the midweek periods, as weekend results roll out of the tracking averages. Dunno as I can see any reason to conjecture this coming Tuesday's Poll-That-Counts results might diverge much from what seems to be the well-established pattern. Could happen, of course ... but the odds sure don't favor it.

Uh, ohhhhhhhh



That's 7 to 5 (a drop from 7 to 4 two days ago ). Hope ya'll got your bets in! :wink:

just listened to a u.s. pollster being interviewed on CBC-TV. his point was that we are all free to guess and bet on the outcome of the election, but that no one can reliably predict the outcome of the election. if it should turn out that your guess was right, it doesn't mean that you actually knew what the outcome would be, you just bought a lottery ticket that turned out to be the winning ticket. hbg

If you believe that; take the points. :wink:

That pollster would be wrong.

I can reliably predict the outcome of the election. The question is whether you believe me or not, but that is your problem, not mine.

Kerry will win with between 280-301 electoral votes. Of that there is no doubt.

I'm with you, ebrown!

As usual, I won't contest the general point Timber made. Bush does indeed appear to have the edge still, and that edge does indeed hover somewhere around the 2%.

And of course, he was ahead, most of the time comfortably so, throughout September and October, no argument there either. Though one can ask oneself whether seeing such a lead preceed one of merely 2% just before the elections is actually a good or a bad sign for the guy.

But even as I won't contest Timber's main point, I again have to do some factchecking on the details.


Odd, that. I count quite a few more individual Kerry leads and ties, some of which continued into the subsequent poll. Perhaps the overview RealClearPolitics gives is simply not complete (but we'd already established that), or you've overlooked a few.

Eg (and you can check them back on my own graphs):

- AP-Ipsos had Kerry up in both its last two polls, in week 41 and week 43, by 4% and 3% respectively

- Democracy Corps had Kerry up in all its last six polls, spread over the past four weeks, by anything in between 1% and 3%

- The Economist/YouGov poll had Kerry ahead in four of its polls since early September, including the last three ones, with his lead increasing to 4% in the last one

- Gallup had Kerry tied with Bush in week 40 and up 1% the week after

- Newsweek had Kerry up 2% in week 40

- The Harris poll had Kerry up 1% in week 37

- ARG registered a tie in its last two subsequent polls, in week 36 and week 40

- IBD/CSM/TIPP registered a tie twice, in week 37 and 39

- CBS/NYT registered a tie in week 40. NBC/WSJ registered a tie in week 42. Pew registered a tie in week 42. Both a LegerMarketing and a LATimes poll registered a tie last week.

Then there are of course this month's daily tracking polls, three out of four have shown Kerry ahead at some point in time, although all four have shown Bush ahead more often.

So, let's rephrase Timber's sentence to make it correct:

"More recently, in the 50-some polls reported since the beginning of September, Kerry's best individual poll results - apart from those registered in daily tracking polls - have been two 4-point leads, three 3-point leads, five 2-point leads, six 1-point leads and 11 ties, recorded by a spread of 13 polls, all of which were however at odds with the majority of the corresponding polls within the week they appeared and only five of which were sustained in a subsequent reporting of the particular poll."

Hm. Sounds different ;-)


It's been a while - mid-August was the last time - but Kerry had in fact a 2 point lead even in the RCP graph for over a week twice; he had one almost on end between early July and mid-August.


An interesting phenomenon indeed. On the other hand, it should be noted that today, for example, both the TIPP and Zogby tracking polls had Kerry pull himself back up into a tie with Bush, despite the mid-week sample.

right on!

4 more days!, 4 more days!, 4 more days!, 4 more days!, 4 more days!!!

I think the Republicans are "buying" polls so that these various meta-analyses will be skewed toward them.

In 2000 it was Rasmussen. Every right wing talk show host was quoting the Rasmussen poll whenever Gore seemed to be going up in the polls. In a country where the greatest landslides in anyone's memory are 60% - 40%, Rasmussen predicted a nine point Bush victory in the popular vote, off by 10 points. That's half the difference between Reagan and Mondale.

One would think that such a performance would force Scott Rasmussen out of the business and into something else, like a nice family bakery where he could sell rolls instead of polls. But no, he's still sticking around, and apparently there are people who actually take him seriously. Unbelievable.

With Rasmussen an obvious property of the Republican party since 2000, this year's GOP hireling is the Gallup organization. With no evidence whatsoever, Gallup has decided that the electorate this year will consist of 40% Republicans and 33% Democrats, and that their polls will be arranged to reflect that notion. So any Gallup poll has a built-in 7% slant towards Bush, any way you look at it.

What's the purpose of this charade? Well, partly to take advantage of the people who will vote for whom they think will be the winner, because they think that makes them one. Even more nefarious, however, is the fact that in our statistic dominated age, all these meta analyses look so...so...Geek Chic. All these reports based on a variety of polls-hey, they're bound to even out, right?

No. Not if some of them are geared to reflect a desired result, they're not. You throw Rasmussen and Gallup in the mix,and you are going to have a meta mess.

At this point, it is too close to call, of course, either in the popular vote or the Electoral College, but be aware of two things.

A) Younger voters tend to favor Kerry, especially since the Draft Question reared it's head, (where has it been all this time?) The polls don't record people who only have cell phones, a situation which describes a large group of young people today. People who lean toward Kerry are not showing up on published polls, and the polls are razor close. Cannot be comforting for Bush.

B) There is a tendency for voters to "break" towards the challenger at the last minute when he runs against an incumbent. True,there don't seem to be many undecideds left, but one or two percentage points in a race this close are like gold.

That's two things Kerry has going for him,and the race seems evenly balanced even without those two factors counted in. Looks like Kerry has the advantage here.

Hamburger, I did read as well today about how one prominent pollster complained that, while in the old days two out of three people called would answer the poll questions, nowadays just one out of three people they call is willing to answer. Now you can compensate a lot with weightings, but when the sample becomes that small, it does raise some serious questions.

On the other hand, the pollsters' collective track record has been pretty consistent in recent elections. The NCPP overview has the average margin error between 2% and 4% in all the last four Presidential elections. And last time round they did as good as they ever did since 1976. Based on that, one would think the average of all the last polls shouldn't be off more than 2-3%, 4% at most. Of course, thats well enough to tip the elections either way.

I see that Rasmussen is being more sophisticated this year than in 2000. Instead of waiting for a place where Bush breaks ahead a little so that Rasmussen can show Bush breaking away by a wide margin, like he did last time, Rasmussen has waited until the final few days to come out with his "evidence" that undecideds will break towards the incumbent this year.

Once again,the aim is the same as last time-to create a sense of momentum for Bush. I mean, going into the final five days tied, having some well documented factors going against Bush is not what the GOP wants, is it?

So Rasmussen comes up with "evidence" at the last minute to show that the incumbent will benefit from the break this time.

To quote Princeton's Sam Wang, a meta-analyst himself, (italics mine): "Recently Scott Rasmussen reported data that he says supports the notion that late-deciding voters prefer Bush. The survey was done from 136 late-deciding voters, far too few to reach statistical significance. This is a message poll aimed at driving the discussion in his preferred direction. Also, the survey assumes that the voters who decided during the survey period are similar in characteristics to those who wait until the last minute, possibly until they are standing in the voting booth. This is untested."
http://synapse.princeton.edu/~sam/pollcalc.html#EVmap

For Rasmussen's "evidence", go here:
http://www.rasmussenreports.com/VoterDecisionTime.htm


The GOP buys polls, but Zogby, the most accurate pollster the last two elections, gets labelled with the "bias" tag. Unbelievable.

The "evidence" the Gallup poll bases its numbers on is, quite simply, the number of Republicans and Democrats they encounter in their likely voter sample. No more and no less. The striking (and controversial) point about their polls is namely that, unlike other pollsters, they do not "arrange" their numbers to meet certain proportions of party ID. That's what they're being criticized for.

Two points to make here. One: a number of polls (Zogby, TIPP, the ABC/WaPo tracking polls) weight the results they get from their polling sample by party ID. Eg, if they pull in a sample of 1000 that includes 400 Republicans and only 300 Democrats when, back in 2000, Dems and Reps each represent about a third of the voters according to the exit polls, they will weight these numbers to make them conform to those 2000 proportions. That is, they'll give each of those 400 Republicans less weight than each of the 300 Democrats, so in the end both subsets weigh in on the final result roughly the way they did in 2000.

Gallup does not do this. Gallup does not believe in weighting by party ID. And there is a good case to be made against such weighting: after all, it assumes the proportions of self-described Democrats and Republicans in the electorate will be the same this year as last time. This got pollsters in a lot of trouble in 2002, when suddenly a lot more Republicans turned up at the polls - or a lot more voters at the polls described themselves as Republicans - than past record would have suggested.

So in response to your post the point here is that Gallup has not "decided" that the electorate consists of 40% Republicans. In fact, it is one of the pollsters that does not add a decision of their own on what proportion of party IDs can realistically be expected to turn up at the vote into the mix when calculating likely voter results. If they got 40% Republicans, its because thats what they happened to find in their likely voter sample.

Which brings us to the next question and a more reasonable explanation of why Gallup has polled so Republican-favourable this year: what is up with their likely voter sample, if it yields such counterinstinctually high proportions of Republicans?

First off, it needs to be said in re: to your post, that of all pollsters, Gallup has by far been the most open and transparent about its methodology. Mystery Pollster (a Democratic analyst, by the way) writes:


In the same post, Mystery Pollster provides an excellent summary of what their methodology actually is, so no reason for me to repeat. After a follow-up post summarising some main points of critique levelled at it, he then explains the main problem. He doublechecks the assertion that Gallup has come up with strikingly Bush-friendly numbers this year (it has), and then analyses what the reason behind it could be. No, its not that Gallup has been "bought" by the Republicans. But the Gallup poll might well be "simply screening out too many voters who do not typically vote in presidential elections". First-time voters, for example. Self-reported first-time voters were just 6% of the Gallup poll's likely voter sample, while they constituted 9-10% of the LV sample in rival polls - and of the turnout in 2000. And if anything, their proportion now should be even higher than in 2000. This is significant because both opinion polls now and the 2000 results suggest that the Democratic candidate enjoys a huge advantage among first-time voters, netting a 10-20% lead among them.

Basically, the Gallup poll is so rigorous in its elaborate system of weeding out likely voters, that it appears to risk grossly overreporting the ever-reliable, ever-informed voters, and underrepresenting the less sure-footed ones. In turn this has led it to underreport the low-income and minority vote - Gallup LV samples have had fewer blacks and hispanics than other pollsters'. Other polls, like Fox and DemCorps, would have compensated such discrepancies a little still by their application of regional weightings, to make the sample reflect the regional turnout numbers from the previous election. But Gallup does not apply those either.

The difference between Gallup and other polls shows up so conspicuously this time because this year's elections involve such an increased interest among new and irregular voters. This has made other polls stretch up their classification of likely voters, encompassing a greater share of registered voters than normally. A NYTimes poll for FL released today for example (see Daly) defined 96% of registered voters as "likely to vote". That would be a huge turnout. And there's the rub, because Gallup does apply the one weighting: on turnout. They recalculate their LV sample as much as they need to to make it reflect a 55% turnout. So while other polls have increased their definition of who is likely to vote on the basis of the answers they got, Gallup kept counting only the 55% of respondents who best fit their criteria of being likely to vote. Excluding those new or irregular voters that boost Kerry's numbers in other polls.

Gallup has listened to the criticisms and will apparently raise their cutoff for likely voters on their last survey, coming up this weekend, to 60%. Let's see if that will indeed show up a different proportion of Reps and Dems, and what effect that in turn will have on their horserace numbers ...

In a final link, this is MysteryPollster's uncharacteristically impassioned response, as one Democrat to another, to the MoveOn anti-Gallup ads.

In other breaking news, KUTV reports that Utah might be safe for Bush after all, though his lead is still stuck at a precarious 68% to 23%.

(Sorry, feeling a little flippant after all that dry analysis).

I think its sorta cute that some folks figure the polls must have been "bought" since they don't favor The Democrats. Also amusing are the opinions voiced here by some that either Rasmussen or Zogby have a built-in Bush Bias. There is plenty of humor in politics.

There is also always the possibility that the minority of Bush-supporters on A2K is not representative of the Bush-supporters who have voted or will vote on Tuesday and that a majority of Americans, especially voters, do prefer Bush.

There is no question that at least on the Politics threads the preponerance of participants is other than Bush-Freindly. There are a couple threads around here somewhere which discuss who has what leaning, and how many of whichever there are. And there are some folks who obviously would prefer folks not of their own particular persuasion not be permitted to participate AT ALL.