Articles and Analysis


Morning Status Update for Monday 11/3

Topics: Status Update

If you hoping for (or dreading) indications of a clear shift in voter preferences on the surveys released in the last 24 hours, you will not find them here this morning. As of this writing, Barack Obama's margin has clicked up very slightly on our national trend estimate, while yesterday's statewide surveys show a very slight narrowing of that margin in a few key battleground states. However, our current classification of the states remains unchanged over the weekend: We still show 311 electoral votes for Obama, 142 for McCain with seven states representing 85 electoral votes still in our toss-up category.

One big caveat: We will no doubt see a heavy pace of new statewide and national survey releases later this morning. Some will be released by the time you read this. I am planning to do another update at the end of the day to summarize the final round of polls.

At the state level, we logged in 25 new surveys yesterday, of which 15 tracked previous polls by the same pollster released since October 15. Of these 15 tracking polls, 11 showed small nominal shifts toward McCain, 2 showed nominal shifts to Obama and 2 showed no change in margin.

081103 new1

The impact of these new surveys on our trend estimates is very slight. We do show slight shifts on the margins of our trend estimates to McCain in more states (8) than to Obama (4) but these changes mostly represent a few tenths of a percentage point. The average shift across all of the battleground states since Sunday is just 0.1% to McCain.

081103 trends1

Four new survey releases in Pennsylvania confirm that Obama's lead there has narrowed significantly over the last week, although four polls still show Obama leading by margins of 6 to 8 percentage points, with Obama receiving more than 50 percent of the vote. Even if the McCain campaign is counting on the typical "Bradley Effect" (in which most or all undecided voters "break" to the white candidate), Obama's 51.8% of the vote on our trend estimate indicates that McCain would still fall short.

In Virginia, two new automated surveys from SurveyUSA and PPP both show Obama leading (by margins of 4 and 6 percentage points respectively), although by slightly narrower margins for Obama since last week. Our trend estimate shows Obama leading by 6.3 points (50.6% to 44.3%), still a big enough margin to qualify as "leaning" Obama.

Three new surveys in Ohio show a similar pattern. The new automated telephone survey from PPP and the mail survey from the Columbus Dispatch shows Obama leading by 2 and 6 points respectively, while a new live interviewer survey from Mason-Dixon shows Obama trailing by two percentage points. The new surveys narrow to 5.6 points (49.9% to 44.3%), just enough to remain in the "lean" Obama category.

Ohio is one state where Obama's margin has been consistently narrower on automated IVR surveys (+1.5 percentage points as of this writing) than on live interviewer telephone surveys (+5.8 points). Here too, the difference is mostly in McCain's percentage of the vote. As of this writing, Obama receives 48.7% of the vote on our trend estimate if based only on automated surveys and 49.3% if based only on live interviewer surveys.

081103 national1

I will have more to say about the final round of national polls later, but for now let me underscore yesterday's warning about the danger's of cherry-picking. The pace of releases and the not entirely comparable field dates make apple-to-apple trend comparisons difficult, but the table above obviously shows shifts in McCain's direction on some surveys (as compared to their most recent sample with non-overlapping field dates), and shifts to Obama on others. With the exception of the outlying IBD/TIPP survey, Obama's leads by comfortable margins on all of these final or nearly final polls. Four years ago, George Bush held a average 1.5 percentage point lead on the final round of national polls. Obama's lead on these 12 surveys averages 7.3 points -- virtually the same margin (7.2) that he receives on our overall national trend estimate as of this writing.


Charles Franklin's graphic posted last night is probably the best way to look for late trends in the national tracking polls. As of last night, eight trackers were showing Obama's margin increasing and two decreasing, but as Franklin points out "three or four of the rises are quite small."



It looks highly likely that Obama will get elected President of the United States of America. The state polls indicate that Obama will win Ohio and Pennsylvania.

Game over Republicans.

Please answer a single payer universal health care poll at http://poll.democratz.org



What is the significance in the difference of the solid Obama/Biden and the toss up numbers between RCP EC count & Pollster EC maps for 11-2-2008?

RealClearPolitics Electoral College

Obama/Biden 278 - 238 Solid 40 Leaning

McCain/Palin 132 - 118 Solid 14 Leaning

Toss Up 128

RCP also now has PA listed as leaning Obama????

at Pollster.com small top right map on poll results pages

Obama/Biden 311 - 264 Solid 47 Leaning

McCain/Palin 132 - 129 Solid 13 Leaning

Toss Up 85



At a guess the current national chart set at the least sensitive has the trends the most accurate. What it shows is the margin more or less being established around 10/10, and since then Obama and McCain attracting new support (presumably from undecideds) at roughly the same rate.

On the more sensitive charts, you get to the same place but with more wiggles. As many have noted at this point, there has been the appearance of an odd sort of cyclical effect with the national trends narrowing than expanding a couple times. I'm not sure we really know why--it could be random, or there could be some sort of real cycle in the polling (perhaps related to things like weekend effects, variations in the poll mix, or so on). But in the end I think the least likely explanation is that the margin in voter preference has really shifted back and forth, and so that is why I think the least sensitive national chart may end up being the most accurate this time.




One basic point is that Pollster uses a regression method to calculate their state estimates, whereas RCP uses a recent average method. Pollster also uses a fairly rigorous statistical approach to defining "strong" and "lean" states. This is the core of their approach:

"What is the basis of the classification of each race?

Regardless of the number of polls, we calculate a 'confidence intervals' around the trend estimate based on the average sample size for the available polls in each state. These intervals reflect the uncertainty in the estimate due to random noise in the polling data.

If a race shows a lead that is outside the 95% confidence interval, then we classify this as a 'strong' lead. If the lead is between the 68% and 95% confidence intervals, then we classify it as a 'lean'. If the race is inside the 68% confidence interval, then we classify the race as 'too close to call.'"


In contrast, RCP uses a dividing line approach: apparently a candidate has to be +10 in their averages to qualify the state as "strong" (incidentally, I think this changed recently from +9), and +5 to qualify as "leaning".

Personally I think the Pollster methods are obviously preferable, both in the way they estimate states and how they classify them. RCP's methods are better than relying on intuition and individual polls, but otherwise they just don't use the sorts of relevant statistical tools that Pollster does, and there is a good amount of arbitrariness inherent in their approach.


Vicente Duque:

Prices on Monday, November 3, 7 AM EST, to buy Betting Options ( shares ) for the Candidates. You win $100 at INTRADE if your candidate wins :

The Entire USA : Obama 89.8, McCain 11.3

Toss Up States for Obama with Last Prices :

NewMexico (5EVs) ...Obama 89.0, McCain 11.0
Colorado (9EVs) ..... Obama 89.0, McCain 14.0
Nevada (5EVs) ....... Obama 82.0, McCain 14.0
Florida (27EVs) ...... Obama 73.8 McCain 30.1
Ohio (20EVs) ......... Obama 80.9, McCain 21.4

Iowa (7EVs) .............. Obama 93.5, McCain 6.5
New Hampshire (4EVs) . Obama 86.0, McCain 14.0
Pennsylvania (21EVs) ... Obama 88.6, McCain 12.0
Virginia (13EVs) ......... Obama 87.1, McCain 18.9

North Carolina (15EVs) .. Obama 63.5 McCain 37.0

Toss Up States for McCain :

Missouri (11EVs) ......... Obama 45.1, McCain 52.5
Indiana (11EVs) ........... Obama 39.0, McCain 59.0
Georgia (15EVs) .......... Obama 30.0, McCain 75.0
Montana (3EVs) ........... Obama 30.0, McCain 70.0

Don't pray for Pennsylvania or Virginia, they are secure. Pray for North Carolina and Missouri.

All others are pretty secure.



Vicente Duque


KMart Dad:

I know there are those who don't believe the old maxim that the state polls lag the national polls, but it sure seems like this has been the case this election season. This time last week the national polls were tightening a bit, but the state polls were looking better for Obama. Now, the national polls show an expanding Obama lead while some of the state polls show a tightening. It all converges on Tues, when the state polls catch up with these national polls, and Obama wins by a comfortable margin, and the race is called by 11 p.m. Eastern.


Just a slight correction: the Bradley effect refers to voters who tell pollsters they support the black candidate and then vote for the white candidate. It is not about the behavior of undecided voters; if they were to break heavily for the white candidate, that would be a different phenomenon.


Okay, I think it's time for folks to give their predictions. Mine is here:


Or, quite simply:

Obama: 53%
McCain: 45%
Other: 2%


marian dieter:

Down to the wire - in just 18 hours we'll be watching the results pour in -- can't believe it's almost here and looking good for Obama. Thanks for all the analysis over the last few months - Pollster.com rocks and is addicting.



Thanks for the great analysis of this election. The movement of public opinion in this election has been amazing. I believe that your techniques are the best I have seen. I wonder if a multiple technique using a historical reliability index of a pollster as an additional y factor would be valuable.


There are quite a number of new state polls with data on preferences of early voters. We've added nine to our comprehensive collection of more than 75 polls with early voting data. See: http://xcurmudgeon.blogspot.com/2008/10/early-voting-exit-polls-and-demographic.html

Perhaps the most interesting new one is a PPP poll from Montana, where 38% of respondents had already voted, breaking for Obama by 61%-35%.



I've been visiting this site more and more regularly as we approached this historic election, and I just wanted to take a moment to say Thank You for all your hard work! I also appreciate everyone else who has participated in this community. I've enjoyed everyone's insight. Good luck tomorrow in whatever races you're following!




I have to say the above is such a wonderful and authorative explanation with various snapshots of where the polls are today. Thank you,I am stunned with the excellence of your work.



Speculation at this point, but I've also noticed the the cyclical expansion and contraction in the polling trends. One possibility could be a cell phone effect.

Discussion over on "538" suggests cell phone users might tend to answer more on weekends and that group may have a +3 lean to Obama.

Only certain pollsters include cell phones in their surveys, so a combination of timing of a poll release and a weekend response bump might be one possibility for the cycle.




I am new to this poll following stuff, but have looked at several sites. By far, this one is the best, most informative of them all. Thanks so much for the info and the hard work that it must take to compile it.



I've been using this site for the past month and half in my AP US Gov't and Politics class in Olive Branch, Miss. It has been invaluable! Thanks so much for your efforts! My students come into class each morning and the first thing they do is check the board for changes. For the brief period Mississippi was listed as "leaning" McCain, pollster was the talk of the school!


Anna in Florida:

I've been spending a lot of time on your site this election. Thank you for such a marvelous job in pulling all this data together.

Another post about Intrade prompted me to post about the UK site, Betfair. Their bettors also overwhelming predict an Obama win nationally, and in most battleground states, including Florida, Ohio, Pennsylvania, Virginia. It gives McCain Indiana, George, N.D., Georgia, Montana and his home state, Arizona.

North Carolina looks favorable but not a home run for Obama; in Missouri, it's tied. Very similar to Intrade.

Betfair correctly predicted all 50 states correct last election, including the toss-ups such as Ohio.

Americans can't bet but you can look at all the trends. Based on the betting projections, it's like the EV would be 364 Obama, remainder to McCain. It's interesting how closely the bets have matched the polling trends.



Post a comment

Please be patient while your comment posts - sometimes it takes a minute or two. To check your comment, please wait 60 seconds and click your browser's refresh button. Note that comments with three or more hyperlinks will be held for approval.