General Question

girlofscience's avatar

Who are the people changing their minds?

Asked by girlofscience (7556points) October 5th, 2008
34 responses
“Great Question” (9points)

As a person closely following the presidential race, I am always surprised when the polls change. There are many websites on which you can observe poll figures, as they change by day. In many states, the polls have changed drastically. I’m wondering, “Who are the people changing their minds?”

Everyone I know has known for months which candidate he/she will be supporting on election day. I don’t know a single person who has changed his/her mind. While it certainly can happen, I was under the impression that such a shift in opinion was a rarity and did not reflect the gigantic percentage points that are consistently jumping in each of the polls.

Does anyone know why the polls results change so frequently? Are people really changing their minds, or are different people polled everyday? How many people are polled? How generalizable are the polls to what’s going to happen in the actual election? (I know how statistics work, but there are obviously a lot of other factors involved here.) Also, why do different polls produce such different results?

Observing members: 0
Composing members: 0


girlofscience's avatar

Haha, I always get lurve on my questions before I get any answers!

girlofscience's avatar

If anyone’s interested in additional reading about polls, I found this Yahoo Answers thread: How do election polls work?

JackAdams's avatar

EACH POLL IS DIFFERENT, because in each poll, different people are asked, and that skews the results.

girlofscience's avatar

@JackAdams: Ok, but the changes in poll results are spoken about as if people are changing their minds. When Obama soared in the polls around the time of the economic disaster, the media was broadcasting this as though people were changing their minds and going with Obama because they feel he’d handle the economy better. The media wasn’t like, “As a result of the economic crisis, the people who happen to be polled this week were more likely to be Obama supporters!” Instead, they implied that something actually changed… And I’m asking who those people are who changed their minds.

aidje's avatar

“When Obama soared in the polls around the time of the economic disaster…”

Just for the record, it’s not as if it has all blown over.

girlofscience's avatar

@aidje: Sorry, I didn’t mean to imply that I thought it was. I should have said, “when the economic disaster hit” or something.

aidje's avatar

@girlofscience Fair enough. Sorry for the interruption.

girlofscience's avatar

I’m still doing some reading about this. Just found this: “In 2004, the national exit poll revealed that about one out of every 10 voters made up their mind during the final three days before the election—and 5 percent said they didn’t decided until election day.”

Wow! I just want to know who these people are!

basp's avatar

that’s a goof question, girlofscience. I can’t imagine someone not knowing by now who they support.

marinelife's avatar

Here is some information on undecided voters, how they decide, when they decide, etc.


“Voters who remain uncommitted are more likely to be women (55%). 46% describe themselves as independents. 48% say they are moderates. 53% are age 45 or older. Some of these voters support one of the candidates, but their support is not yet firm.”

La_chica_gomela's avatar

Great answer, Marina! May I ask where you got the information? I’d like to read more!

marinelife's avatar

@La_Chica_Gomela The embedded link takes you to the article, which is about a CBS panel that is tracking these voters. Did the link work for you?

La_chica_gomela's avatar

LOL, somehow I didn’t notice the link!! sorry!

fireside's avatar

Here’s a site that tracks polls

It looks like Sarah Palin and McCain’s actions during the economic crisis cased a lot of potential voters to lose faith.

This link actually shows the respondent’s change from poll to poll, by question.

Another article
Nationally, more than 2 million Democrats have been added to the rolls in the 28 states that register voters according to party affiliation, according to the Associated Press. Republicans have lost nearly 344,000 thousand voters in the same states during the same period.

basp's avatar

I just realized that when I responded to you earlier I wrote “goof” when I really meant “GOOD”. So sorry for the mistake. I am using an iPhone and my typing is not so great.

La_chica_gomela's avatar

lol, basp. i didn’t noticed till now

galileogirl's avatar

On a lot of political polls it is not an either or question. They may ask “Are you very likely to vote for___________. somewhat likely to vote for_______________, somewhat unlikely to vote for_________________, very unlikely to vote for__________”.

Then they read a statement about one of the candidates and and ask “After hearing the statement are you very likely to vote for” etc. So it is possible to change your mind DURING a poll.

OMG it’s time to start screening my calls again.

EmpressPixie's avatar

I’m just popping in to bring up, a website for political polls. They have a lot of information on how the polls are done, what they mean, and other things. It’s a great site for tracking this stuff and he also often explains “big shifts” that are really polling problems or when there appears to be missing data or whatever else is going on.

dalepetrie's avatar

Yes, fivethirtyeight is my favorite site, been following it religiously for months.

girlofscience's avatar

Fantastic site, EmpressPixie!! Thanks!

It really looks like McCain doesn’t stand a chance…


dalepetrie's avatar

fivethirtyeight founders are doing a road trip, if you look back at some former posts, they’ve been driving to swing states and reporting on the ground game. They’ve consistently been coming up with Obama just annhilating McCain in the ground game, and their analysis by and large is that this isn’t even reflected in the polls.

Among the things they’ve pointed out recently is not only hopes that all the swing states we’ve been talking about will turn blue this time, but North Carolina is now a tossup; Missouri is slightly leaning McCain, but there’s almost no McCain ground game there; Indiana is experiencing a grassroots effort unlike anything seen since 1972 when Birch Bayh, a Democrat, was swept into the Senate in a very red district; the one electoral vote attached to Omaha, NE (Nebraska splits it’s 5 electoral votes by district, one of two states where it’s possible for each candidate to get some EVs) is actually considered competitive based on Obama’s ground game and the fact that candidates NEVER go there (because it always goes red), but McCain has dispatched Palin to speak there; and MOST surprisingly, because of the swell of African American registrations, and because African Americans vote 95/5 in favor of Obama in Georgia, even though the state favors McCain by over 7% in polling, if African Americans represent 31–32% of the voting populace instead of the 25–26% that pollsters are assuming, it would tip to Obama and shock EVERYONE, and that early voting is showing 40% of voters are African American. In other words, if you just look at what 538 shows right now in terms of who is ahead in the polling in each state, and then take Missouri and Indiana, both of which are very close right now and turn them blue, give Obama that 1 EV from Nebraska instead of giving all 5 to McCain, and flip Georgia blue, he wins with 391 electoral votes to McCain’s 147!

Another thing they’ve pointed out recently is that if you were to just look at the states where Obama is more than 6% ahead of McCain, even if Obama lost every state where he’s 6% ahead or less, Obama would STILL beat McCain. Basically McCain’s only hope right now is to come up with a game changer along the lines of when he picked his running mate. Or as they sometimes say in politics, if Obama is caught with a dead girl or a live boy. And I’m starting to wonder if THAT would matter at this point.

wundayatta's avatar

Polls, of course, get more accurate as predictors, the closer you get to the election. Even a month out, it is still a risky game to make predictions, but the picture, as Dale described, is getting much clearer.

I heard a talk by a political scientist last year who was looking at the accuracy of the polls compared to the Iowa Election Market. has anyone looked at IEM lately? He found that if you apply a correction to the polls that is based on previous historical relationships between polls a certain number of days out from the election, you can actually get a much more accurate prediction for the election.

Of course, to do this, you need a lot of data, and you need to run some tedious statistical analyses, and I don’t know if he’s doing that this year. If he is, I’m sure it’s out there on the net.

My personal sense is that a month out, it’s all but over. Although several things can go wrong. The polls might be under-estimating the Republican support, as they did in 2000. They might be over-estimating Obama’s support, because of the lying factor (people say they’ll vote for a black candidate when they really won’t because they don’t want to apear prejudiced). They might be under-estimating the Obama vote, because of his registration efforts.

Usually “likely” voters are polled, because who cares about the opinions of people who don’t vote? But likely voters are often defined as those who have voted in the last election (presidential, or congressional, depending on pollster), so it misses those who have just registered.

As to “undecideds”? They baffle me. But I believe it is true that some decide at the last minute. Usually they swing the way other folks are swinging.

In understanding why poll results change, it helps to understand statistics. Each poll is based on a sample of the overall population. When you randomly sample, there’s always a chance you happened to select a sample that does not accurately reflect the entire population. Pollster’s rarely report their estimates of error. We just get the straight poll results, without confidence intervals (how far off the estimates might be).

If you draw ten samples from the same population, you expect them to say different things. If you average the polls together, you well hone in on the population preference. That’s why I like looking at the Poll of polls. It’s also fun to look at the Iowa Election Market.

The “winner take all” market is showing that people have to pay around 75 cents per vote for Obama, and you can have McCain votes for only 25 cents. In the vote share market, it seems that folks believe Obama will get around 54% of the total vote, compared to McCain’s 46%. Hmmm. I should go take Obama at that price!

dalepetrie's avatar


Basically what you are describing about applying historical relationships to polls based on days out is exactly what 538 does.

As for the polls underestimating Republican support, what the polls didn’t (and couldn’t) pick up on is the 96 hour push that the Republicans do, for 4 days leading up to the election they have a MASSIVE get out the vote drive where they canvas and try to knock on a million doors (might even be more). Polls can’t pick that up. But conversely, polls are having a hard time keeping up with GOTV (get out the vote) efforts going on RIGHT NOW on the Dem side. The problem always was that that last minute GOTV effort the Republicans put forth far surpassed the efforts the Dems were able to pull together last minute like that. Dems have been competing in this select handful of states and allowing the elections to swing on Florida, Ohio and Pennsylvania. With Howard Dean running the show now, he instituted the 50 state effort, so that Dems would be competitive everywhere. The drawn out primary campaign which brought the Dems to all 50 states allowed them to set up field offices everywhere, and Obama with all the money he has, has been able to use those offices to do GOTV in every state, and only as the election gets closer is he dropping efforts in states where it’s not competitive and ramping up efforts in states where it is. But his support is so broad that whereas right now the Republicans have offices open, they are barely staffed, and they’ll rely on that last minute push (kind of like cramming right before the big test), whereas Dems have been doing a push every bit as aggressive as what the Republicans will do in the last 4 days. In other words, in 4 days, the Republicans managed in the last two elections to swing the electorate by a couple percentage points. Imagine what the Dems will have been able to do with 3 months solid of similar efforts. I suspect you will see some states we didn’t even think had a chance of turning blue fall into Obama’s column by suprise (Georgia for example).

As for the lying, aka the Bradley effect, the guys at 538 have pretty much put several nails in that coffin. They have even demonstrated how this year, there is a reverse Bradley effect at play, where there are indeed whites who live in more conservative/rural areas who don’t want it getting out in their community that they’re voting for the black guy, and if anything, you are going to see Obama’s numbers better than polled. They even demonstrated how that effect had really fallen by the wayside a few months back by looking at data from the primaries. Basically the only place it exists at all is in some of the most racist states in the deep south, and even there it’s minimal. Even up here in Minnesota, I heard a guy on the radio, a retired barber say he wouldn’t vote for a “nigger”...most people have either by and large overcome their racism, or have embraced it, very few who are racists these days make any effort to hide it.

Also, be careful of electronic markets. Another thing 538 pointed out was that over at Intrade, something very strange was going on…basically about $400,000 is traded every day, which to some idiot degenerate gambler is small potatoes, and as such they’ve been able to manipulate the markets. So where you’re seeing 75% chance on IEM, and betfair over in London is showing 1.3 to 1 odds for Obama and 4.3 to one for McCain, those seem fairly consistent with 538’s 88.5% chance of an Obama victory, but Intrade only shows Obama w/ a 69% chance, even though their electoral vote predictor shows Obama at 353. And yesterday he closed at 65%, even though yesterday was Obama’s single best polling day of the year.

Bottom line, you want to make some money, put whatever you can on Obama over at Intrade. But if you want to predict the winner, these betting markets can be fooled.

wundayatta's avatar

@dale, I got into IEM for the Philadelphia Mayoral Election, which, although it predicted the winner, had too few trades to be reliable. I’m not planning to play with IEM. I don’t even remember my password.

What is 538? And is there a link here?

dalepetrie's avatar

yeah, daloon, read a couple of my earlier posts, it tells all about it. They track all the polls and assign them reliability ratings and try to come up with a reasonable prediction based on numbers and not assumptions. I trust them far more than any other site I’ve found, it’s like my bible for this election.

jvgr's avatar

Some polls are intentionally biased to show the candidate they support is doing better than they are.

mea05key's avatar

I read then news recently that there is this something caled the bradley effect which apparently says that the ‘whites’ might change their mind on voting obama because he his black. Obama himself realise that racism issue is present and he is unable to comfortably lead the presidential race eventhough the polls say so.

dalepetrie's avatar

mea05key -

The Bradley Effect is bunk. It assumes that because Tom Bradley, and African American who ran for Governor of California in 1982 lost the race even though he was ahead in the polls, that any African American would have the same problem. The problem is, it’s 26 years later and this so-called effect has not been demonstrated in over a decade in any election anywhere. What happened there wasn’t that whites “changed their minds” per se, more that they were embarrassed or afraid to admit to pollsters that they weren’t going to vote for the black candidate for fear they’d be seen as racist.

It’s very dangerous to believe in this bunk, because if the Republicans try to steal this election too, they can explain away the difference between the polls and the acutal results using the Bradley Effect.

Also note that something called the “reverse” Bradley Effect is VERY real. This is where in conservative white districts, white people who want to vote for Obama are afraid of the repercussions and say they’re voting for McCain but are actually going to vote for Obama.

La_chica_gomela's avatar

Dalepetrie, you said that there has been no evidence in favor of the Bradley effect in 26 years, thus it’s bunk. Can you share some evidence that supports the “reverse-Bradley” effect?

dalepetrie's avatar

Well, I didn’t say there hadn’t been evidence in 26 years, I said there hasn’t been evidence in a decade, but that this effect was named after something that happened 26 years ago. It has happened since…in 1989 it happened to a guy named Wilder in North Carolina, so there they call it the “Wilder Effect”. But the last time this effect was shown anywhere was over a decade ago. As for the reverse effect, that was demonstrated in the primaries. There are articles on it out there, I have to search for them, but I will do so, it just may take me a little time, so bear with me.

mea05key's avatar

I thought they actually take into account the effect of Bradley by incooperating the marginal error into polls. no?

dalepetrie's avatar

mea05key…I’m glad you posted to this thread…I came up with that article I wanted to cite but couldn’t remember which question it was on.

Anyway, marginal error is not “incorporated into” polls, it is a function of the polls. It’s a statistical concept. If you are trying to predict how an entire population will act, you take a smaller, representative sample to represent that population, if you pick a truly random sample from a smaller but representative subset of this population, theoretically their responses will mirror the responses of the entire population within a certain confidence level or ‘margin of error’. They can say this is an approximation, and depending on how large the sample size is, they can be confident that they are right within plus or minus so many percentage points. That doesn’t take into account any theoretical effects, that’s just the normal, statistical margin of error that will exist when you use 1,000 people to predict the behavior of 140 million people.

Having said that, I’ll once again encourage EVERYONE to read for election polling and analysis (and predictions) that are far better than anywhere else I’ve found, and here is the link to the article they posted which references everything you need to know about how the Bradley Effect is (mostly…and entirely in areas that will matter) a thing of the past.

Answer this question




to answer.

Mobile | Desktop

Send Feedback