Parsing The Political Polls & Predicting The Winners: Report From 20 Paws Ranch
How is it that with one week to go, Barack Obama has a slight lead in about half of the national polls and Mitt Romney has a slight lead in the other half of the national polls? How is this possible when Obama is going to get most of the African-American and other minority votes, and a big share of the independent women and young adult/college graduate vote?
My view is that Obama voters are generally underrepresented in polls, and nail biting aside, Obama will win the popular vote on November 6 by a few percentage points and win the Electoral College in a landslide.
What is your view? Do you think that the polls are accurate? If not, why? And what is your prediction for the popular and Electoral College vote results?
“Report From 20 Paws Ranch,” which is the name of his mountain hideaway, appears on Mondays.
Share This
Hear, hear… And, IMO, we’ll take back the lower house too!
Things have begun to sound a bit “FoxNewsish” everywhere lately and since I voted two weeks ago there’s nothing left to do but enjoy the BS… ah, I mean squirming.
I’ll read more than comment till the votes are counted but I’m taking notes and will be delighted to discuss what happened, after the results are in, with those who simply can’t understand.
The national polls are utterly useless. They show trends, but they may not be the same trends that are happening in the swing states. We have had some really odd state polling over the past couple of days, but not enough to effect the averages enough to be of concern. Ohio, Virginia, Iowa, Nevada are the states that matter and they look good for Obama.
I also voted early, SteveK, and while the experience was not quite akin to 2008 it felt quite good. One needn’t ask for whom I voted.
I forgot CO
interestingly, Obama could lose Ohio, win Virginia, and still have a shot. Romney, not so much.
I really wanted Governor Huntsman to get the GOP nomination. You know, a REAL moderate Republican. Alas, we have the candidates we have, so I’m hoping President Obama wins.
I *hope* Nate Silver’s 538 predictions work out. Though, I’m giving Virginia to Mr Romney and keeping CO a toss up for now. However, if the other states go the way Mr Silver predicts, then President Obama is still over the 270 threshold.
If you look at the state polls on the 538 blog, some are still very close. Iowa, Nevada, and all important Ohio -and others- are all within the margin of error. Sure they are looking good for the President but I’m not putting anything past fate/cheating/bad-luck/poor-turnout/late Oct surprise/Murphy’s Law to tip the scale into a Romney win.
Also interesting to watch Intrade shares over the last several days. http://www.intrade.com/v4/misc/scoreboard/
The percentage chance hasn’t changed much, but the share price for an Obama win has gone up over 40-cents since I’ve been watching.
As for CO, the Romney camp in Utah (if not other states) is shipping campaign workers form here to Colorado to beef up the GOTV for Romney there.
I have been trying to take Nate Silver’s advice and not look at the polls everyday. If I did, it would make me a nervous wreck and I would never get any work done. I just hope Obama takes the win. Not just because I’m a dem, but because Romney scares the bejesus out of me.
I’ve had a feeling for a while that the Obama numbers are under-represented. Totally non-scientific. It came to me sitting at home and watching the caller ID as my phone rang for the 12 millionth time, showing “unknown” or somesuch. I’m not answering any of those calls. But I wondered how many Romney backers, eager to show support, answer every call on the chance that it’s a pollster calling? And then there’s cell phones, which for the most part don’t get called. What’s the political demography of those who only have cell phones?
I don’t understand polls so I can’t really support them. How do they figure in the phones that don’t get answered and cells, as Harry points out. Do more Reps or more Dems answer the calls and when they do, do not participate in the survey. What about people that hate surveys, like my wife, and me who likes to help the caller out since I remember the ass____ when I did some cold calling. I know, we need a “scientific” study/survey to get those answers.
DDuck, put simply, How? With statistics.
Nate Silver has done many articles about how the polls are conducted and compares and contracts some from each other.
yoopermoose, oddly enough Nate Silver doesn’t follow his own advise
the prediction models are based on past polling and performances… so real data against real election results. While I dont agree with Silver’s weighting, his methods are pretty good. I think he is missing the idea that certain pollster manipulate only certain polls when it is political relevant and then fall back the opposite direction in polls that are not so relevant at the time. So in the end, they look like a wash mathematically, even though there is serious manipulation going on.
Harry:
Cell phone users such as myself are underrepresented pretty across the board, but the two or three polls that do reach out to cell phone users tend to trend slightly more positive for Obama. That is anecdotal at this point. After a couple or so more election polling cycles we’ll have a better idea.
Great, I’ll stick to tea leaves.
Watch the early voting numbers.
Shaun:
It will really be interesting to see how the final results tally with the various polls. White the campaigns say they are employing all the “new” technologies to reach voters, you would expect polls to truly take a scientific, studied approach; but trends like these will really test their methods. I wonder how many people with cell phones only are like some of my younger coworkers, who basically only use it for text messaging and twitter, and wouldn’t answer a call even if it came? And will enough of that group even show up to vote and make a difference in the accuracy of the polls?
Interesting premise. It’s entirely unclear what underrepresented looks like. I’ll take a really simple example of a single poll in a single state, although clearly it’s a lot more complex than this looked at nationally.
So we’ll start with the latest PPP poll from Ohio. The poll has the President up by 4 so it’s one of the more favorable recent polls for the President in Ohio. Now, the good news (and the reason I chose the PPP poll) is that they make the makeup of the sample easy to find. So let’s take a look.
We’ll start with race – 83% white, 12% African American, 5% other. What should we compare it to? Well, the easiest thing to look at is the 2008 exit polling data. These data, I would assert, are relatively favorable to the President versus what we’d expect to see in 2012 but they provide a benchmark. The exit polls are 83/11/6 in Ohio so just about in line with what PPP has this time around. There’s no evidence of underrepresentation here.
How about gender? Well, PPP has it 54/46, slightly higher female than the 52/48 in 2008 and favorable to the President, given how much better, relatively, he does with women.
Going through the rest…the young are significantly undersampled versus 2008 (by 5 percentage points), liberals are heavily oversampled and moderates are heavily undersampled. Finally, self-identified Republicans and Democrats are both oversampled in about equal numbers while independents are undersampled by about 10 percentage points. Undersampling independents benefits the President since Romney is generally up among independents (by 4 points in this poll).
Long way of saying, it’s pretty hard to find evidence in the poll of undersampling of the President’s core constituencies. The only place you can find it is among the young. That said, most poll watchers believe youth turnout will be a lot lower in 2012 than it was in 2008.
And that’s just one poll.
My more general thoughts are that most polls that show the President ahead are showing an electorate equal to or more Democratic than was the case in 2008. Most polls that show Romney ahead are showing an electorate about half way between 2008 and 2004 in terms of party ID. In the end, you sense of who is going to win should be a lot about your sense of who is going to vote in 2012 relative to 2008. My sense therefore is that Romney is likely to win the popular vote. As to the electoral college, it’s anyone’s guess. Too much data to process unless you make your life doing it.