Pollster.com

Articles and Analysis

 

Polling's Perfect Storm

Topics: Age Weighting , Bradley/Wilder , Cell Phones , Likely Voters , National Journal

My NationalJournal.com column this week reviews the three big sources of worry among pollsters and others about the accuracy of polling this fall: the potential that pollsters are missing voters in "cell phone only" households, the potential that "likely voter models" may be missing certain types of voters and the worries about the Bradley-Wilder effect.

We have covered all of these topics before and my plan is to devote a column to each over the next three weeks because while everyone is speculating, we probably know more than we think we do empirically, about the potential for error from all three sources.

And since some will react to the title without clicking through, I blog my own spoiler. Here's the last paragraph of the column:

Meanwhile, we can probably dispense with the "Perfect Storm" analogy. In the movie of the same name, three different weather-related phenomena combined to produce a storm of exceptional severity. In this case, as Democratic strategist Joe Trippi pointed out in September, the potential polling foibles may work in opposite directions and "cancel each other out." A return of the Bradley-Wilder effect would work to McCain's benefit, while an underrepresentation of younger, African American or "cell-phone-only" voters will likely benefit Obama.

 

Comments
Gary Kilbride:

My instinct is the result will be tighter than the polling. I have no great rationale for that, at least nothing that would look decent in print.

I understand David Gergen said last night on CNN that he doesn't believe the polls. I wouldn't go that far, but in an open race I believe the outcome will naturally tighten, particularly given some of the variables like Democrats having a difficult time busting 50% in a national race, a black man with resume questions atop the ticket, and McCain with more strength among independents than a typical Republican in this climate.

I'll be content to ignore the cell phone factor because the posters who I respect the least on message boards continue to push it, including great confidence in 2004. We've seen Bradley-Wilder before so I believe it will show up in certain states.

No doubt the state by state relationship to the national margin will look weird in some cases in 2008, and not necessarily reliable going forward.

____________________

brambster:

@Gary

It's not like Obama represents something outside of the mainstream, and I have seen absolutely no evidence of a Bradley effect in either the primaries, or the Harold Ford and Deval Patrick races in 2006.

I don't doubt that there are voters out there that would vote for Hillary instead of Obama because she is white, but most of these people have been voting Republican for years in the first place. If there was a black Republican running against a white Democrat, I might give this effect some consideration, but I think that most people that have been voting for Democrats over the years are long past this.

It's my belief that for every one that wouldn't vote for a black candidate, there would be an extra black voter showing up to the polls, and less Republican votes among black voters. Harold Ford received 95% of the AA vote in Tennessee in 2006, and I believe that Obama will likely do around 93% nationally whereas Kerry only achieved 88% of the AA vote in 2004.

You are banking on people lying to pollsters for this effect to be come true. I think there's a good chance that the polls are underestimating AA turnout and the AA percentage for Obama, and that there may well be a reverse-Bradley effect in many places just like the primaries. Even in the special 1st CD election in Mississippi, Travis Childers benefited greatly from the increased AA turnout, and won by 7.5 points.

There is plenty of enthusiasm for Obama among the black community, and at the same time, there are very few whites that see Obama as being different just because of the color of his skin.

Beating Hillary was the hard part. That was the real race for the presidency. Obama is a very formidable candidate, certainly not because of the color of his skin, but because of who he is and the issues of this election that strongly favor Obama.

If there was a 4th part to the "perfect storm", I would add in the loser effect. People don't enjoy supporting or voting for a loser. If the polls predict a clear Obama win on election day, you will likely find fewer people going out of their way on election day to vote against Obama. You will probably also see a pile-on effect as people start to think that if he has a 10 point lead in the polls, he must not be that bad, and all those things people say about him might just be dirty politics. I believe this is in fact responsible for half of Obama's surge in the past month (Palin's very bad performances being the other half).

____________________

Pachete:

Bramster -

While i dont think AA % will be higher than the polling (I actually think it will be lower and a reverse Bradley effect will take place in terms of strait percentages). I absolutely agree that they are underestimateing turnout. Obamas registration drive has registered millions nationwide.

I also honestly beleive the Bradley effect will not be an issue the race with Hillary provided possible racists with a cover for John McCain (They will stick to McCain has experience) and lets not forget McCain is a "maverick".

Come election day i think the democratic voter registration drive will be a net plus for obama and completely overshadow any bradley effect taking place.

____________________

brambster:

Just to be clear about this. I do not believe the Bradley Effect exists, or at least does not exist today.

The effect doesn't describe racism, it describes people that tell pollsters that they will vote for a black candidate when they in reality aren't prepared to do so and end up voting otherwise.

There will always be situations where a small number of people question their own intentions and change when they go to the polls, but I would argue that the same thing exists when it comes to women candidates, Jewish candidates, and even Catholic candidates in certain places. The Bradley Effect though assumes a big swing from polled to actual votes, and I don't see that happening. I think it is quite possible that the Bradley Effect was actually created to explain bad polling.

So I think the polls today are not being influenced by a Bradley Effect, but instead, they are likely underestimating the turnout and percentages among the AA vote for Obama, i.e. a Reverse Bradley Effect (which again would only be bad polling which underestimated the turnout and support).

____________________

brambster:

Oh, and one more thing that is related.

I expect undecideds in the last few days to break strongly for McCain. The undecideds are likely becomming more and more conservative during Obama's surge, and people making up their minds now aren't that likely to change.

This group of undecideds will prefer the more established candidate, just as they frequently broke for Hillary.

Some have claimed that the Bradley Effect was in reality the movement of the undecideds to Deukmejian during that election, which was also helped by effective campaigning on the part of Regan.

Obviously an October surprise could change things, but people that are still undecided at this point are clearly out of touch and more likely to vote on comfort than change.

____________________

Anyone care to predict the turnout percentage this year? And notwithstanding early voting now underway in so many states, can we expect chaos at the polls due to heavy turnout for a historic race? If so, where especially?

____________________

ThatSeattleGuy:

Mark: Assuming I'm allowed to comment on something in this thread OTHER than the Bradley effect... :)

You mention the oft-discussed issue of cell-phone-only voters. Please, if you could, address the issue of double-counting as well? It just occurred to me because I got an IVR presidential race poll on my cell phone today (first time!), and so I'm wondering...if the same pollster calls me on my landline at home, wouldn't I be double-counted? Or is that so statistically unlikely as to matter? If not, wouldn't it over-represent certain demographics (the ones with cell phones and land lines, whomever we may be...)?

thanks /TSG/

____________________

The Bradley-Wilder effect is a fascinating subject. I'll look forward to reading about it here, as well as the other subjects.

A key piece of new evidence comes in the form of a recent study of 133 gubernatorial and U.S. Senate races. The study showed that on average, the Bradley-Wilder effect has largely disappeared since the mid-1990s. You can read about it in my piece on The Disappearing Bradley Effect.

Sam Wang
Princeton Election Consortium

____________________

RE: Bradley-Wilder

I'd be more worried about it (as an Obama supporter) if it had shown itself during the primaries. There were notable occasions where the polls were way off going into an election (NH, NC, IN to name a few), but there didn't seem to any particular pattern. If all the polls had shown a pro-Obama bias compared to the end results, that would be cause for concern, but the results of the primaries (of which there were many!) didn't seem to show any. Either the Bradley-Wilder effect isn't relevant in this case, or else, like Mark said, it's being canceled out by missing cell-only voters and bad likely voter models. Either way, there's no reason to think that the polls will be significantly off as a result.

____________________

Vicente Duque:

Dear Friends of this page :

You have absolutely and completely forgotten two very important words in this election and page "latino" and "hispanic".

It may be possible that a Big Turnout of Latinos appears as an important factor on November 4.

I would bet for a Big Latino Landslide for Obama in the Southwest : New Mexico, Nevada, Colorado and probably a smaller one in Florida.

Latinos could also have some relevance in North Carolina and Virgina to tip the scales.

There are few reasons for poor latinos, feeling big economic pain in Nevada, Colorado or New Mexico to vote Republican.

Dont Forget Immigration Fears, Tears and Dears.

The Obama Coalition of Young Voters, Blacks, Latinos, PHDs, Masters, and BSs ( not BS please ) will win, see how :

http://milenials.blogspot.com/

http://tossUpStates.blogspot.com/

Vicente Duque

____________________

paulte:

I think the Bradley effect will show up mostly in the undecided vote. Prior to the Palin pick, the polls were on average Obama - McCain (46% - 42%). This left 12% undecided. I think most of the undecided vote (say 8% will go for McCain). This would make the race even. Another thing no one seems to be talking about is the continuing decline in the stock market. Granted the financial crisis is the main reason but could it also be a sign that people believe the polls and think that Obama will win?

As a conservative and religious white person myself I would not vote for Obama for many reasons and one of the reasons is that he is black. I don't think a black person can represent me in the White House. The President is the head of state and America is still a predominately white country and should have a representative person as President.

Most white people will not come out and say this but many feel this way.

____________________

nomorerepubs:

What I'm seeing in the McCain camp is all this out and out hatred for and smearing of Obama. Did you see him on the stump today in Wisconsin? He is so bad, they have Sarah there to break up the worsening performance of McCain.

Obama is on a roll, the polls in the states are going more and more blue, this will feed on itself and right now all this red meat slime comming from McCain is just to keep the repubs in it. As the polls get worse and worse for McCain some repubs will be saying to themselves its over, I'm not even gonna vote.

This is what McCain is fighting, he's losing and the deeper the hole gets the worse his base feels and the worse they feel the fewer will support him and come election day Obama could win by margins even higher than he currently enjoys.

____________________

sweetspot:

@ paulte:

That's a sad comment from someone who seems fairly thoughtful. The idea that a black man cannot represent you in the White House is odd. Is skin color the defining factor of a human? As a religious person, you and Obama have more in common than I do (I am agnostic), and as a man, you and Obama have more in common than he does with 51% of the population who are female. By your arguments the president should be a woman, right?
But here's the error of your argument: the president is not our representative. For that you have Senators and Representatives in Congress. The president has a different role to play. S/he should be a thoughtful person who can lead the country to do what's best for it.
Who best fits that description?

____________________

FlyOnTneWall:

MB:

Your conclusion, I have to say, makes no sense.

Your own column points out (and Sam Wang verifies empirically) that we have no evidence for a Bradley effect in recent years. It makes for great copy, and feeds into the fears of Democrats and the ambitions of Republicans, but there's just no reason to think it's playing a role.

The other two factors you discuss are very, very different. Pew has now quantified for us the impact of cell-phones, and it's worse than we thought - it not only leads to undersampling, but the people being ignored are more liberal than the ones being counted. And it's a complimentary effect to your other concern, bad modeling.

So what you've got are two reasons to think that the polls are likely underestimating the strength of support for Obama. To which one could reasonably add a third - we've seen plenty of statistical evidence for a reverse Bradley Wilder effect this year, particularly among black voters, but also among the broader electorate. So there's your perfect storm.

Sometimes, as nice as it is to conclude that everything balances out and we should just throw our hands in the air and wait, the evidence strongly supports a contrary conclusion.

____________________

Cookie100:

What people are missing about the Bradley effect, it took place well over 20 years ago.
Since 1980, 83 million Americans have been born, in a country that is multi racial and ethnic, kids do not see skin color.
I say kids, cause I'm 55, so, 28 year olds are kids to me.
Obama is a kid to me!
This is the future of our country, not another pasty old white war monger.
You want war with Iran and Russia, McCain's your guy, but HE'S GOING TO START THE DRAFT,
I REMEMBER THE DRAFT, a lot of this country does not. I also remember when it was illegal to have an abortion,
folks, do you really want to go backwards into that world?
Palin and McCain will outlaw abortion and start the draft, it's your world, so do what you want, but please, hire the guy that's going to fix my 401K!!!

____________________



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.

MAP - US, AL, AK, AZ, AR, CA, CO, CT, DE, FL, GA, HI, ID, IL, IN, IA, KS, KY, LA, ME, MD, MA, MI, MN, MS, MO, MT, NE, NV, NH, NJ, NM, NY, NC, ND, OH, OK, OR, PA, RI, SC, SD, TN, TX, UT, VT, VA, WA, WV, WI, WY, PR