“Dem” Lying Polls: Has Romney Been Truman This Whole Time?
Date posted: Monday, October 29, 2012
The realization that our polling data might be completely inaccurate.
On Friday, I sat my wife down to have a serious talk. “I’m about to tell you something I haven’t told anyone.” “What?” she asked. “What’s going on?” I took a deep breath, ran my hand through my hair, and sputtered, “I think Romney’s going to win, and I don’t know what to do.”
Earlier that day, my political world had made sense. I knew every swing state’s value, where the polls stood, where the polls were, and even where they were likely going. I fully expected to follow the polls, pick every state correctly on Election Night, watch the president re-elected, and bask in the splendor of prescience. I had the numbers on my side. I knew the truth about this election.
But then I stumbled upon an entirely new truth.
That conversation with my wife was approximately 60 hours ago. Luckily, she takes bad news like a champ.1 She humored my misery then wisely moved on to other things. I, however, had a much more turbulent weekend.2
Mind you, my moment of stress and sorrow had little to do with ideology. Last week, while tremendous Construction columnists like Anthony Resnick, Ben Hoffman, and Stephen Kurczy made clear their trepidation that the “Etch A Sketch candidate” could actually win, I continued my Spock-like reliance on logic: the polls showed that President Obama still had the edge in the Electoral College. The president’s firewall in Ohio would hold.
On Friday, however, when unexpected discoveries about misleading polls led me to think Romney might win (more on that soon), I started doubting everything about the prediction process. The thought of a Romney Administration didn’t bother me; it was the thought of my fallibility as a pundit. After all, I had correctly predicted every election since 1988.3 I had a reputation to uphold.
For any regular reader of this column, you must be wondering why I suddenly vacillated. I’ve regularly pointed out Obama’s continued electoral advantage. Last week I predicted that while Romney would win the popular vote, Obama’s noticeable leads in Pennsylvania, Ohio, Michigan, Wisconsin, Iowa, and Nevada were enough to eclipse 270.
So, what changed? Well, it actually has nothing to do with new polls or any kind of “Mittmentum” from the last week. It has to do with the very indicators upon which I had been building my predictions—the polls themselves.
Quick aside: I wrote my Masters’ Thesis on the Election of 1948. Its iconic image—that of a victorious Truman holding the Chicago Tribune’s erroneous “Dewey Defeats Truman” headline—represents that, by about 10 to 15 points, the pollsters were dead wrong about the election. “Truman delightfully displayed the headline,” my intro concludes, “as Republicans and the media scrambled to identify what went wrong.”
I fear we are careening toward a similar hangover. What went wrong, it was later identified by nearly every major American publication, was an overreliance on polls. They became the scapegoat for all the pundits.
I went from thinking “Romney is winning this thing!” to “Yeah, we really don’t have a clue here.”
The polls of the first half of the 20th century bordered on pseudoscientific.4 Certainly the relative lack of technology diminished pollsters’ reach.5 We like to think that we’ve since learned from their mistakes. Certainly the evolution of technology inundates pollsters with unprecedented amount of macro- and micro-information about the country and voters. One would think that with all this new data, we’d have polling down to an exact science. And maybe for a while—during the 1980s and 90s—we did.
More recent technology, however, is counterproductive. While just about everyone now uses a telephone, many people solely use cell phones or have caller ID on their landlines. In both cases, fewer people pick up unfamiliar phone numbers, especially poll-weary citizens in the battleground states. This surely skews the poll numbers. Some pollsters still only survey listed landlines, but about one-third of American households are, according to a government survey, “cell phone only.” The Huffington Post recently declared that half of all battleground voters are exclusively cellular. How cells skew polling, though, is anyone’s guess. I’ve read arguments for both sides. The left thinks they’re not being counted, the right thinks it’s them. As always, it depends on where you get your information.
My frantic Friday consisted of finding what could be wrong with all these polls. The cell phone and caller ID stuff was just the beginning. I started looking at how many Democrats and Republicans the pollsters were polling. I found that Democrats are consistently oversampled in battleground polls. With these revelations, we finally arrive at my Friday meltdown.
Take Ohio surveys, as almost everyone agrees it is the most important state to watch. Obama looked strongest in an October 23 Time poll of the Buckeye State, which showed him up five points. However, Time polled 290 Democrats to only 220 Republicans. That’s nearly 31 percent more Democrats polled than Republicans, and Obama won the poll by only 5 points. Can Democrats even call that a victory?
The second most favorable Obama survey—CNN/Opinion Research’s poll from October 25—says the President leads by 4. However, 35 percent of those polled identified themselves as Democrats, 32 percent as Republican. (More on this one soon.)
The third most favorable Obama poll—his 3-point SurveyUSA lead—had 39 percent of its participants as Democrats, 32 percent as Republican. Balance out the parties’ samples and it’s an even poll; it’s maybe even a Romney lead.
What’s more is in each of those Ohio polls, Independents solidly prefer Romney. In the SurveyUSA poll, they chose Romney 47 to 39 percent. In the Time poll that showed Obama’s 5-point lead, Independents went for Romney 53 to 38 percent. That’s massive. Does anyone think, if the Ohio Independents are skewed that much for Romney, there could be that many more Democrats than Republicans to outweigh his advantage with the swing voters? No chance.6
Here’s what I’m fairly certain of: Romney is currently in the lead.
All of these polls combined corrupt a Real Clear Politics average that says Obama has a 2-point Ohio lead. Fix those three poorly sampled polls, and his lead is gone. And, if you trust this Republican memo highlighting significant GOP progress in Early Voting and Absentee Ballot numbers compared to 2008, you’d think the Republicans have an enormous advantage in the only real voting that’s taken place.7.
As the weekend developed, though, I went from thinking “Romney is winning this thing!” to “Yeah, we really don’t have a clue here.” For example, that CNN poll that showed Obama’s four-point Ohio lead also said that, with Independents, Obama held a five-point lead. How does Time show Romney up 15 with Independents on October 23 then CNN shows Obama up 5 with the same group on October 25? That’s a 20-point swing in 48 hours. I assure you this does not reveal a spontaneous clamor for re-electing an incumbent we’ve known for four years. What it does reveal is that these polls just can’t be trusted.
So you can see why I’m going a little crazy. There is a strong chance these polls are misleading. Their erraticism suggests that, at best, they are circling around the truth. If you ask Gallup or Rasmussen right now, we’re in for a Romney triumph. Meanwhile, if you average all the polls, Obama is equally advantaged in the Electoral College.
Here, I suppose, is the most important sentence of this entire column: If the polls are accurate, Obama is the clear favorite. If they aren’t, though, we really can’t say with any kind of certainty who’s even in the lead. Even the great Nate Silver is running the data of these polls to make his predictions. However, the models only work if the polling is accurate. Philosophy 101 tells you that if your premise is flawed, any conclusion you draw will not be correct.
You might be wondering, then: is there anything we do know? There’s not much, but here’s what I’m fairly certain of: Romney is currently in the lead. Even by the Democrat-friendly polls, he still is the favorite to win the national popular vote, so I would be surprised if he doesn’t.8 Thus, since 53 of 56 elections have aligned the popular vote winner with the Electoral College winner, Romney is probably in the lead. Probably.
And you know what? I think the Obama camp knows it. CNN ran an article on Saturday about how it’s Obama taking the aggressive swings these days, while Romney stays on message. If anyone knows what’s actually going on, it’s the campaigns themselves (though, of course, they’ll shamelessly spin it any way they’d like on the talk shows). Ultimately, this election has turned unpredictable. The polls, after all, might be right, and it could still be Obama’s election to lose, but I think there are just too many variables with this polling data.
In 1948, only in retrospect did everyone blame an overreliance on polling data. I think it’s going to happen again. I don’t know who’s going to win next Tuesday. At some point, though, we’ll be looking at these polls and trying to figure out what went wrong.
And maybe that’s the only truth I know for certain.
- Most prominently when a judge pronounced us man and wife.
- Which reminds me, this column will make no requisite political analogies to Frankenstorm. Starting now.
- As a five-year-old, I may have chosen George HW Bush because I liked his glasses. But still.
- An infamous 1936 Literary Digest poll asked its readers to predict the election, and they resoundingly said Alf Landon would upend President Roosevelt in a rout. They turned out to be dead wrong. Turns out the Literary Digest had a largely Republican subscriber base. And why not? Who had disposable income for a magazine subscription at the height of the Great Depression? Those who weren’t depending on Roosevelt’s social programs.
- Phones were not ubiquitous, for starters. Party identification was also more fluid, as FDR’s re-polarization of political bases took a few decades to solidify. The result was some pretty awful sampling by pollsters like George Gallup and Elmo Roper. Both miserably failed in 1948. It was as if they were taking shots in the dark.
- Author’s note: Literally minutes ago this Monday morning, another Ohio poll came out from Public Policy Polling. It shows Obama up 51 to 47. The sample, however, was 43 percent Democrats, 35 percent Republicans. Is it that hard to get better samples! Oh, Independents went for Romney 50-47.
- I highly recommend that memo to Republicans if you want talking points and to Democrats if you want a heart attack.
- The most favorable Obama national poll, IBD/TIPP tracking, has seen his 6-point lead fall to 1.3. You can bet they’ve been consistently overestimating his success, so he’s probably losing.