[besson3c]

Sorry, I'm just being silly

No mockery intended, we all were affected by all sorts of polls in one way or another, I'm just kind of happy that we won't be reading headlines like this for a while!

[stupendousman]

Originally Posted by besson3c 

Sorry, I'm just being silly

No mockery intended, we all were affected by all sorts of polls in one way or another, I'm just kind of happy that we won't be reading headlines like this for a while!

I'll be coming up with a poll "Hall of Shame" and "Hall of Fame" when the percentages pan out. I've see the final from 5-8% so far, depending on what numbers are being used.

...and getting it about right the day of the election doesn't count in my book and consistently over or understating support by more than say 4 percentage points pretty much makes a poll worthless IMO. While polls do have a "margin of error", being off is STILL ERROR. 4 points covers error in most of the better polls anyways.

If it's 8 percent, both IBD and CBS/NYT have some explaining to do. IBD jumped in about one day 5 points. I don't think that's credible. NYT/CBS had the race at about 13 most of the last month. It dropped down to 9 right at the bell too. Again, I don't think there was that big of a shift in support at the last minute. I'm inclined to believe that the polls at the ends of the spectrum adjusted their weighting at the last minute because they knew the numbers they were using before were way off.

[Kerrigan]

What is the tally right now, in percentages, of the popular vote? I just want to see if my prediction of a 3% Obama victory is accurate or not.

[Shaddim]

Looks like around 3.5%.

[Shaddim]

This got a laugh out of me.

[besson3c]

There are still a lot of uncounted votes in states like Washington and Oregon.

[stupendousman]

The numbers I've seen are:

OBAMA: 62,450,831
MCCAIN: 55,393,194

That's what CNN has. That's 52% to 47%. 5 points.

Here are the numbers the day before the election, from highest to lowest (according to realclearpolitics.com):

Gallup +11
Reuters/C-SPAN/Zogby +11
ABC News/Wash Post +9
CBS News +9
Marist +9
IBD/TIPP +8
NBC News/Wall St. Jrnl +8
FOX News +7
Ipsos/McClatchy +7
CNN/Opinion Research +7
Pew Research +6
Rasmussen Reports +6
---------------------------
Diageo/Hotline +5
Battleground (Lake)*+5
---------------------------
Battleground (Tarrance)* +2


If the numbers hold up, pretty much all of them but Diageo/Hotline, and the two Battleground polls overshot in favor of Obama, as I suspected. Zogby, Gallup, CBS and ABC come out with the most egg on on their faces.

Zogby was ALL OVER THE PLACE this year, but when he finally landed, he landed right on his face. Gallup for the most part of the last few months had the trend right (they averaged around 3-6), but the past week for some reason totally ramped up the numbers on Obama and way overshot. CBS and ABC should just be lucky they tweeked things so that they didn't still have the totally phony 13-14 point spread that they've had for the past month or two. Those turned out to be about 10 points off. They still ended up with the maximum error, and maximum embarrassment.

Then we had Pew that miraculously (and suddenly) had results that would appear to be close to the final numbers. Pew went from 15 points for Obama and in a matter of a couple of days dropped 9 points to get to a respectable 6 point prediction. I DON'T BUY IT.

Rassmussen hovered around 5 for the past several weeks, so their numbers seemed consistent. They were only one point off. Diageo/Hotline had been forecasting a little higher than their final numbers, but they weren't so far off the past couple of weeks and so their final tally seems consistent with the direction their polling had gone. Battleground has traditionally been VERY accurate over the last couple of Presidential elections. It seems as though that Lake gal knows what she is doing.

http://www.nixguy.com/?p=4724

I realize that technically, pollsters can CLAIM that they are "accurate" as long as their polls fall within their margin of error. Normally, that allows up to a six point difference in the point spread. Gallup only had 4 points available, and they even surpassed that. They were just flat wrong. Zogby could have 5.8 points difference and they surpassed that as well. As I predicted from the outset, the pollsters where likely getting it wrong on weighting and sample. As is almost always the case, the pollsters almost always err on the side of giving Democrats more than they end up getting in the long run. Will we remember this next election season? Probably not. We'll again be overshooting the Democrat chances by 5-6 points and assuring everyone that the numbers don't lie. The question is, are they doing it on purpose or are they just incompetent?

[villalobos]

Originally Posted by stupendousman 

The numbers I've seen are:

OBAMA: 62,450,831
MCCAIN: 55,393,194

That's what CNN has. That's 52% to 47%. 5 points.

Here are the numbers the day before the election, from highest to lowest (according to realclearpolitics.com):

fivethirtyeight.com

he did a pretty good job.

[SpaceMonkey]

Technically a poll can't be "wrong," only the conclusions drawn from it can. But whatevs. Enjoy your day!

[besson3c]

Firstly, there are still plenty of uncounted votes. CNN has the margin at 6% now. It's a little premature to get into national poll scoring, but this looks to be about exactly what the poll averages work out to be.

I'm not sure why you chose to obsess over individual polls, a variance in data is to be expected, especially now that cell phones are emerging as the standard and many people are dropping land lines. This is why we work with polling averages, which appear to have been accurate for the most part.

[besson3c]

Here's how pollster.com fared in the individual swing states (rounded off to nearest integer):

OH: predicated margin: 3, actual: 4
PA: predicted: 7, actual: 10
FL: predicted: 2, actual: 3
VA: predicted: 6, actual: 5
IN: predicted McCain by 1, actual Obama by 1
NC: predicted Obama by a hair, Obama by a slightly bigger hair
MO: predicted Obama by 1, McCain won by a hair
CO: predicted: 8, actual: 7
NM: predicted: 9, actual: 15
NV: predicted: 7, actual: 12
IA: predicted: 13, actual: 9


I left out some states like ND and MT, since I don't think there was ever a whole lot of data in these small states.

[besson3c]

CNN says that 96% of overall precincts have reported.

[tie]

I never really cared about the individual polls, since I used fivethirtyeight to aggregate them. Talking about egg on faces, though, it should be noted that polls affect the outcome of what they were trying to predict. For example, if every poll had predicted a 13% Obama lead, then I can guarantee that the actual lead would have been less---because fewer Obama supporters would have bothered to vote. I have no idea what happened this time, though.

[stupendousman]

Individual polls skew the averages. If you've got 2 or 3 that have consistently (and incorrectly) predicted a 12-15 point advantage that did not exist, that does have an effect on the average. Realclearpolitics final average was about 8 points. You take away the really bad, outlier polls and the average is a lot closer. That's even with some of the worst of the outliers hedging their bets and apparently tightening up their weighting. That's why I ditched them for the "Stupendousman Poll of Polls". I think the last one I posted was about a 4 point average.

Also, as TIE mentioned, the reporting of polls can have an effect on turnout, but it can happen both ways. People can see a "landslide" coming and won't bother since their vote "won't count".

I think that it's important to find out why several of the polls were so wrong, to ensure that next election they aren't doing the same thing and unduly influencing things. They need their noses rubbed into the smelly messes they've made and those in charge need replaced if their organizations want any kind of credibility in the long run.

[besson3c]

Originally Posted by stupendousman 

Individual polls skew the averages. If you've got 2 or 3 that have consistently (and incorrectly) predicted a 12-15 point advantage that did not exist, that does have an effect on the average. Realclearpolitics final average was about 8 points. You take away the really bad, outlier polls and the average is a lot closer. That's even with some of the worst of the outliers hedging their bets and apparently tightening up their weighting. That's why I ditched them for the "Stupendousman Poll of Polls". I think the last one I posted was about a 4 point average.

Also, as TIE mentioned, the reporting of polls can have an effect on turnout, but it can happen both ways. People can see a "landslide" coming and won't bother since their vote "won't count".

I think that it's important to find out why several of the polls were so wrong, to ensure that next election they aren't doing the same thing and unduly influencing things. They need their noses rubbed into the smelly messes they've made and those in charge need replaced if their organizations want any kind of credibility in the long run.

That's exactly why I've been advocating Pollster.com, because they discard outliers and analyze trend lines. Now do you agree with your good buddy besson3c?

Ironic that your best-case-scenario poll of polls and the RCP polls averaged together equals about what it actually came out to be

I'm still waiting for the rest of the results particularly in those Western states, but so far it looks like Pollster's national trend line was off by 2 (Obama: 8, actual 6). I wouldn't be surprised if the remainder of results bump the actual up to 7 though.

[zerostar]

Originally Posted by villalobos 

fivethirtyeight.com
he did a pretty good job.

Yes Nate was pretty spot on!

his estimates: 348.6 and popular 52.3 to 46.2

[besson3c]

Pretty close! I think most would have agreed that states like NC, IN, and MO could have flipped either way, so I would say anything in the vicinity of 350 or so was a success.

[besson3c]

Sure enough, Obama's victory margin is up to 7% now, according to CNN, with still more to come (97% reporting)

[stupendousman]

Originally Posted by besson3c 

Sure enough, Obama's victory margin is up to 7% now, according to CNN, with still more to come (97% reporting)

Back to 6.

If it drops back to 5 or goes up to 7, It won't matter for those lame pollsters who got it so wrong that a point or two in either direction won't likely help the perceived accuracy of their polls.

[besson3c]

stupendousman: I guess I just don't understand what is fueling your argument then? I don't think that anybody is disagreeing with you that some pollsters were wildly inaccurate, and that some seemed to have a definite lean. Is that your point?

[stupendousman]

Originally Posted by besson3c 

stupendousman: I guess I just don't understand what is fueling your argument then? I don't think that anybody is disagreeing with you that some pollsters were wildly inaccurate, and that some seemed to have a definite lean. Is that your point?

I think I made my point pretty clear. Just about ALL of the polls overshot Democrat support, and this is a pretty consistent thing over the past years. These polls are what the media uses as "news" to tell the country what your peers think. When Gallup or the New York Times reports a phony double digit lead, fence sitters take note. If I don't yet know who to vote for, but most people are voting for Obama, why wouldn't that influence my vote? We all know that people often times make decisions based on popularity, not on what is logically sound.

Polls which got it totally wrong at both ends (though there are none that got it as wrong for McCain as those at the high end did for Obama - not surprisingly) should receive scorn for unprofessionalism due to their inability to correctly analyze the data they were tasked with interpreting. It would be different if several prominent polls hadn't gotten it pretty close and had been hovering around those numbers for most of the election. Pew, CBS/NYT and ABC had it a doulbe digit blow-out for most of the post convention time period. They should have been getting the same sort of data the other pollsters where, unless they were simply screwing up the sample. If they got the same basic sample, then they were screwing up the weighting. The question is why they'd do that when other pollsters did not. Was it done for partisan reasons, or due to incompetence? Neither are acceptable reasons to mislead the public.

I guess my point is that polls who consistently where predicting results that were 5-8 points higher than the end result probably shouldn't be taken seriously next election cycle. Especially someone like NYT/CBS which at least in 96 did the same exact thing. I don't think we should be promoting polls as "news" when it's clear that many are simply engaged in fantasy. I pointed out again and again when people claimed it was a double-digit lead (even Fox News reported that) that this was something that the pollsters in question should have to answer for. If they don't, what's to stop them from doing the same thing next election cycle? Not everyone makes the effort to go to websites to look for aggregate poll averaging. They wait to hear the news on their boob tube, which normally tells them the latest outrageously inaccurate poll before they tell us how the Democrat candidate makes our legs tingly and use doctored video to show us that the Republican candidate has scary religious beliefs.

[subego]

Originally Posted by stupendousman 

how the Democrat candidate makes our legs tingly and use doctored video to show us that the Republican candidate has scary religious beliefs.

It was doctored?

I should note however, the Democratic tingly media did show some tape which could lead someone to the conclusion the Democratic candidate has scary religious beliefs too.

Edit: me singling out this point does not imply a lack of validity to your other points, which are well stated.

[besson3c]

Originally Posted by stupendousman 

I think I made my point pretty clear. Just about ALL of the polls overshot Democrat support, and this is a pretty consistent thing over the past years. These polls are what the media uses as "news" to tell the country what your peers think. When Gallup or the New York Times reports a phony double digit lead, fence sitters take note. If I don't yet know who to vote for, but most people are voting for Obama, why wouldn't that influence my vote? We all know that people often times make decisions based on popularity, not on what is logically sound.

Polls which got it totally wrong at both ends (though there are none that got it as wrong for McCain as those at the high end did for Obama - not surprisingly) should receive scorn for unprofessionalism due to their inability to correctly analyze the data they were tasked with interpreting. It would be different if several prominent polls hadn't gotten it pretty close and had been hovering around those numbers for most of the election. Pew, CBS/NYT and ABC had it a doulbe digit blow-out for most of the post convention time period. They should have been getting the same sort of data the other pollsters where, unless they were simply screwing up the sample. If they got the same basic sample, then they were screwing up the weighting. The question is why they'd do that when other pollsters did not. Was it done for partisan reasons, or due to incompetence? Neither are acceptable reasons to mislead the public.

I guess my point is that polls who consistently where predicting results that were 5-8 points higher than the end result probably shouldn't be taken seriously next election cycle. Especially someone like NYT/CBS which at least in 96 did the same exact thing. I don't think we should be promoting polls as "news" when it's clear that many are simply engaged in fantasy. I pointed out again and again when people claimed it was a double-digit lead (even Fox News reported that) that this was something that the pollsters in question should have to answer for. If they don't, what's to stop them from doing the same thing next election cycle? Not everyone makes the effort to go to websites to look for aggregate poll averaging. They wait to hear the news on their boob tube, which normally tells them the latest outrageously inaccurate poll before they tell us how the Democrat candidate makes our legs tingly and use doctored video to show us that the Republican candidate has scary religious beliefs.

There are several problems with your argument:

1) Who's to say that the polling problems were simply methodology and not some hidden agenda, and that next election they will make adjustments that ultimately result in them being more accurate?

2) Cell phones are a variable that polls have difficulty accounting for. All but one of the polls that showed bigger Obama leads than reality included cell phones in their sample.

3) All polls like your beloved LV Gallup poll took last election into account in terms of determining voter turnout, but turn out this election was nothing like last. What should they have done differently?

[subego]

Originally Posted by besson3c 

2) Cell phones are a variable that polls have difficulty accounting for.

How?

[besson3c]

Originally Posted by subego 

How?

Should they be used? How are the numbers obtained? How should they be weighted? Should cell phones be called on the weekends when many have free minutes? Are people less likely to take the poll during the week on their cell phone? How do you get a balanced demographic on the cell phone when there are probably a greater number of younger people using them? How do you know how to balance the cell phone only vs. land line only ratio of users?

[besson3c]

http://www.fivethirtyeight.com/2008/...continued.html

It's too bad that the little chart isn't displaying for me, but on Nov. 2 the pollsters who used cell phones had Obama up by 9.4 on average nationally, while the ones that didn't had him up by 5.1. He ended up being up around 7 (and counting, 98% reporting), which is just about the average of these two averages.

Go figure.

[stupendousman]

Originally Posted by besson3c 

1) Who's to say that the polling problems were simply methodology and not some hidden agenda, and that next election they will make adjustments that ultimately result in them being more accurate?

Why should polling institutions that act on "hidden agendas" ever be trusted? Especially when there are polling institutions that seem to get it fairly close every election.

Cell phones are a variable that polls have difficulty accounting for. All but one of the polls that showed bigger Obama leads than reality included cell phones in their sample.

So, the answer would be not to poll cell phones?

I think that the problem is that pollsters were looking for reasons to justify weighing more heavily for Democrats than was reasonable, given past precedent. They were hearing the "buzz" and being unduly influenced by that. Cell phone users skew younger and more minorities have cell phone only service. Using those callers would help their pre-conceived notions about potential turn-out.

3) All polls like your beloved LV Gallup poll took last election into account in terms of determining voter turnout, but turn out this election was nothing like last. What should they have done differently?

The Gallup LV poll actually was trending right along with the actual end result until the last couple of days. It went from 5 - 11 points in just a couple of days of polling, which suggests they were basing their weighting on some newly discovered data which turned out to be bogus. It's odd that both their tradition and "expanded" LV both showed the same, wrong results. What's even odder is that their "registered voter" poll showed only a two point difference from their LV polls. It would seem that Gallup just took everyone's word for it that they'd vote if they said they would, despite never having voted before in their lives.

What should the pollsters done differently? I don't know. I don't know the internals of most of the final polls and don't know where the data went wrong for the ones that were far off. I'd previously pointed out though that estimates of a 16 point Democrat advantage was pure fantasy. I myself figured there's be up a 6 point Dem. advantage, based on past precedent and the great job Obama did (ignoring that it required tactics that resulted in vote fraud) in getting out the vote. Rasmussen somehow figured out how to do it right, and they and the Battleground poll normally seem to do the Presidential election polls so they are pretty close every time. Not so CBS/NYT, Gallup, ABC, etc. There are people getting it pretty close year after year. Those are the guys whose results should be emphasized.

But, one thing this does point to is why "registered" voter polls are essentially meaningless. Gallup told us that right before the election, a sample of registered voters with NO weighting gave a 13 point edge to Obama. Of course, that was double the amount that he actually got. This is why several months ago I pointed out that such polls were pretty much as good as flipping a coin into the air and that even LV polls usually gave way too much to the Democrat candidate.

[besson3c]

Originally Posted by stupendousman 

Why should polling institutions that act on "hidden agendas" ever be trusted? Especially when there are polling institutions that seem to get it fairly close every election.

You need to make your case that there is, in fact, a hidden agenda. In the past pollers ilke CBS/ABC/NYTimes have been much more reliable. In the past others have been more reliable and the ones that were reliable this election weren't.

If you think that a particular poller has been consistently inaccurate across different elections, make your case. To me, it just looks like many struggled finding a workable methodology with results that could be repeated on follow-up polling, but this was a massive challenge for all of them all things considered, polling is not science. However, taking all of the data that was available and working with averages or trend-lines seemed to yield pretty accurate results, despite this alleged Democratic bias that you are claiming. Why is that? Why is it that both the Fox News and Daily Kos polls ended up being accurate? If anybody would have a bias, I would think it would be either of them, no? If there was a Democratic bias across the board, why did the overall averages still yield accurate results?

stupendousman, you really need to collect your evidence and have your data ready to defend your claim before you make it, and you need to research this with an open mind rather than simply looking for something that backs up your pre-existing beliefs.

So, the answer would be not to poll cell phones?

I think that the problem is that pollsters were looking for reasons to justify weighing more heavily for Democrats than was reasonable, given past precedent. They were hearing the "buzz" and being unduly influenced by that. Cell phone users skew younger and more minorities have cell phone only service. Using those callers would help their pre-conceived notions about potential turn-out.

Again, you will have to make your case of an intentional Democratic skew. You haven't given us enough data to sink our teeth into yet. I hope this thread doesn't end up being like the Al Gore thread...

What would you do about cell phones?

What should the pollsters done differently? I don't know. I don't know the internals of most of the final polls and don't know where the data went wrong for the ones that were far off. I'd previously pointed out though that estimates of a 16 point Democrat advantage was pure fantasy. I myself figured there's be up a 6 point Dem. advantage, based on past precedent and the great job Obama did (ignoring that it required tactics that resulted in vote fraud) in getting out the vote. Rasmussen somehow figured out how to do it right, and they and the Battleground poll normally seem to do the Presidential election polls so they are pretty close every time. Not so CBS/NYT, Gallup, ABC, etc. There are people getting it pretty close year after year. Those are the guys whose results should be emphasized.

Rasmussen had Obama at 4 on the final day, and Battleground had him at 3, no? So, they were off by at least 3 and 4 points, respectfully. Why do you feel that these pollsters outperformed some of the others?

[stupendousman]

Originally Posted by besson3c 

You need to make your case that there is, in fact, a hidden agenda. In the past pollers ilke CBS/ABC/NYTimes have been much more reliable. In the past others have been more reliable and the ones that were reliable this election weren't.

To be honest, all I remember for sure is that CBS/NYT both this year and in 1996 grossly over-estimated Democrat chances. If at least half the time they are getting it majorly wrong, there has to be a reason. I remember 96 because it was another year where a good number of polls got it really wrong. I think that if I really wanted to do the research, I could show that during elections where the Democrat wins, the polls are off in favor of the Democrat way more than those where the Republican was the front-runner.

Remember, my claim was that the polls THIS YEAR where using bogus assumptions based on how they were weighing the numbers. I said that I was sure that there were going to be several pollsters at least on the high end which where going to end up being embarrassed. I was right.

However, taking all of the data that was available and working with averages or trend-lines seemed to yield pretty accurate results, despite this alleged Democratic bias that you are claiming.

I said that some polls got it right. The averages typically where STILL skewed in favor of Democrats, though much closer to the final tally than some of the way crazy outliers. What I know is what I said - the pollsters where using figures based on assumptions that could not be made based on any past precedents. Turns out, this election wasn't all that much different than the last...

CNN Political Ticker: All politics, all the time Blog Archive - Report: ‘08 turnout same as or only slightly higher than ‘04 � - Blogs from CNN.com

Rasmussen had Obama at 4 on the final day, and Battleground had him at 3, no? So, they were off by at least 3 and 4 points, respectfully. Why do you feel that these pollsters outperformed some of the others?

Rasmussen Reports 11/01 - 11/03 3000 LV 2.0 52 46 Obama +6
Rasmussen Reports: The most comprehensive public opinion coverage ever provided for a presidential election.
GWU/Battleground 10/29 - 11/02 800 LV 3.5 50 44 Obama +6

Those are the last two polls released by them that I could find.. I understand though that Tarrance and Lake each did submit a second set of polls based on their individual predictions (Tarrance had it at +2 and Lake +5)

[Gee-Man]

Stupendousman, you should be big enough to admit that the one polling site that you scoffed at and dismissed as being "biased" turned out to be the most dead-on accurate in predicting the final result:

fivethirtyeight.com.

I don't expect you to become an Obama supporter, but the numbers he had don't lie. Nate Silver may be a Democrat, but he is and was scrupulous with his number-crunching and was not inclined to bias the results one way or the other, and his accuracy shows. You have to admit this at the very least.

[tie]

538 was amazingly accurate. I think that it is great; it seems the Bradley Effect didn't really manifest itself.

[stupendousman]

Originally Posted by Gee-Man 

Stupendousman, you should be big enough to admit that the one polling site that you scoffed at and dismissed as being "biased" turned out to be the most dead-on accurate in predicting the final result:

fivethirtyeight.com.

To be honest, I scoffed and dismissed it because it WAS biased. The editorial content anyways. It's hard to get past that when you are looking for facts.

Could you show me a chart with his popular vote (since that's what we're talking about in this thread) predictions for more than just Nov. 4? I'd like to see how consistent he was. I'll give him a pat on the back if he wasn't predicting a 9-14 point Obama victory at some point after the convention. That was Pew's problem. They where all over the place and tightened up just right before the ballots where cast. You aren't going to be able to convince me that in about a day or two's time, Obama jumped maybe up to 8 points in the polls.

[kobi]

Originally Posted by stupendousman 

To be honest, I scoffed and dismissed it because it WAS biased. The editorial content anyways. It's hard to get past that when you are looking for facts.

Could you show me a chart with his popular vote (since that's what we're talking about in this thread) predictions for more than just Nov. 4? I'd like to see how consistent he was. I'll give him a pat on the back if he wasn't predicting a 9-14 point Obama victory at some point after the convention. That was Pew's problem. They where all over the place and tightened up just right before the ballots where cast. You aren't going to be able to convince me that in about a day or two's time, Obama jumped maybe up to 8 points in the polls.

Nate Silver and his team at fivethirtyeight.com did a great job. You can't beat the math, plain and simple.

Obama didn't jump up 8 pts in the last two days, he held a double digit lead for a least a month before the general on most polls.

The Supendousman poll filter just didn't want to believe it.

[stupendousman]

Originally Posted by kobi 

Obama didn't jump up 8 pts in the last two days, he held a double digit lead for a least a month before the general on most polls

No, he didn't. Most of the polls showed between a 4 and 8 point spread for the month before the election. A couple showed it less than that, and several did show it double digits and they were wrong.

In the Pew Poll, he went from something like 15 points up down to 6 points in a day or two. Not a real credible scenario. Especially when all the other polls showed no such movement.

Can you help me with that information I did ask for though?

The Supendousman poll filter just didn't want to believe it.

I think the last time I pulled out the filter, I had it a 4-5 point race and that was a week or two ago. Pretty close I think. Much closer than Gallup, Zogby, Newsweek and the NYT/CBS poll.

[Wiskedjak]

Originally Posted by stupendousman 

No, he didn't. Most of the polls showed between a 4 and 8 point spread for the month before the election. A couple showed it less than that, and several did show it double digits and they were wrong.

For most of the polls, but what about the one that had it right?

[tie]

Originally Posted by stupendousman 

Could you show me a chart with his popular vote (since that's what we're talking about in this thread) predictions for more than just Nov. 4? I'd like to see how consistent he was. I'll give him a pat on the back if he wasn't predicting a 9-14 point Obama victory at some point after the convention.

This exact chart is on the top right of his webpage, "super tracker." He's been projecting a 6 point Obama victory for the last month, and this was correct. He has predictions for each state along the right side of the webpage. I only checked one or two, and they were right on, but that could just have been luck. I haven't checked more.

I am not sure why you have a problem with his bias. I found his site to be the best aggregator of polling information by far. How did you get your info? Did you just go to every polling site one at a time? If you don't want to believe his analysis, that is fine. But on the other hand, he explained in detail the math behind his analyses. It should still be taken with a grain of salt, but it was a fantastically useful site.

Edit: He was quite wrong on at least one state. He projected an 11.1% McCain win in Arkansas, and in fact it was a 20% win. On the other hand, even the most extreme poll only projected a 13% McCain win, so it isn't his fault as an aggregator. Any idea what happened in Arkansas?

[Gee-Man]

Originally Posted by stupendousman 

To be honest, I scoffed and dismissed it because it WAS biased. The editorial content anyways. It's hard to get past that when you are looking for facts.

Nobody asked you to accept the editorials. This was about the NUMBERS, which you assumed were wrong solely because the person who assembled them is a Democrat. For someone who constantly accuses others of logical fallacies when trying to make his arguments, you seem to have forgotten the biggest one of them all - Ad Hominem.

Could you show me a chart with his popular vote (since that's what we're talking about in this thread) predictions for more than just Nov. 4? I'd like to see how consistent he was.

You can find the info yourself. fivethirtyeight has numbers going all the way back to the beginning of this election. As an aggregator/analyst, his numbers have moved, but only when circumstances dictate that they move (i.e. major events in the race such as the conventions and news favoring one side or the other).

I'll give him a pat on the back if he wasn't predicting a 9-14 point Obama victory at some point after the convention.

Which convention? The Democratic convention? That's a ridiculous standard, and you know it. Polls change over the course of a campaign, that's the point of daily polls. People change their minds, there's no such thing as a poll that will stay locked to one number from the conventions until election day.

If the election were held the day after the DNC, I'm sure it's quite possible that Obama would have won by a 9-14 point victory. If the election were held the day after the RNC, it's quite possible that McCain would have won. The polls reflected the changes in the electorate, plain and simple.

That was Pew's problem. They where all over the place and tightened up just right before the ballots where cast. You aren't going to be able to convince me that in about a day or two's time, Obama jumped maybe up to 8 points in the polls.

Some polls jumped dramatically. fivethirtyeight is an aggregator, he doesn't do polling himself, he simply crunched the numbers, and did an excellent job of it.

Really, it's not that hard. I know it's hard for you, but admitting that a liberal did something right, even a tiny little thing like predicting the final results of the election accurately, won't give you liberal cooties. Your conservative credentials don't disappear immediately. Try it, just this once. Seriously. You'll need the practice during the Obama Administration.