Inside the Obama-Romney Rematch Numbers: CNN's Deception and the Return of the PUMAs

So a new CNN poll released over the weekend gave conservatives quite the tingle up the legs: were a hypothetical rematch of the 2012 election to be held between July 18 and 20, 2014, CNN's poll said, Romney would smoke Obama by 9 points, 53 to 44 percent. Romney is, however, trailing a hypothetical matchup with Hillary Clinton by 13 points.

Interestingly enough, CNN failed to disclose most of the the crosstabs that matter - i.e. while they broke down support by party, region, age, and race, they did not disclose what portion of their polled electorate belonged to each category, except for party.

The poll is nothing more than fantasy - a bad conservative fantasy at that. Not only is the likelihood of Mitt Romney facing President Obama again in an election roughly the same as rainbow and glitter farting flying unicorns lighting up the skies over Boston, there is no campaign and no actual scrutiny on Romney. The President is daily demonized by the press - because problems created half a world away by the war policies of his predecessor is somehow now the black guy's fault (after all, isn't everything?).

But the criticism of the poll based on circumstances and fantasy, while accurate, does not capture the full absurdity of the poll. The bottom line of the poll is that CNN literally cannot add. Here is what they say at the beginning of their crosstabs:

A total of 1,012 adults were interviewed by telephone nationwide by live interviewers calling both landline and cell phones. All respondents were asked questions concerning basic demographics, and the entire sample was weighted to reflect national Census figures for gender, race, age, education, region of country, and telephone usage. Among the entire sample, 32% described themselves as Democrats, 24% described themselves as Republicans, and 44% described themselves as Independents or members of another party.
— CNN/ORC Poll

Got that? 32% Democrats, 24% Republicans, and 44% everyone else, out of a total universe of 1,012 adults. That's an 8-point edge for Democrats, in keeping with Dem registration and presidential election turnout advantages. Is there any other part of the poll that accounts for actual counts of Democrats and Republicans. Why, helpfully so!



Let's recap: GOP share: 278 is 27.4% of 1,012. Dem share: 303 is 29.9%. Hrmph. Perhaps CNN can explain how 24% who describe themselves as Republicans suddenly become 27.4% of the sample size (and if leaners are taken into account, a whopping 46.4%). The Democratic sample in the actual numbers, on the other hand, is actually less than 30% (with leaners, 44.3%).

So, CNN is telling us that their sample has an 8-point edge for Democrats (roughly in line with Democrats' registration advantage), but their actual numbers reflect a rough 2.5-point edge. In other words, CNN's actual sample is skewed away 5.5 points in favor of the GOP from their initial claim about the sample. Counting leaners, GOP has a 10 point swing from the initial claim (GOP +2 as opposed to original claim of Dem +8).

Furthermore, the reader will notice an uncanny resemblance between the portion of the sample that is D+ (Dems and leaning D) and President Obama's support against Romney: both at exactly 44%. Funny how that works out. It is as if CNN is claiming that not only would their Republican-leaning independents pick Romney, but so would every other independent who do not lean either way (9% of the sample). So, um, what makes those voters non-leaning?

CNN's inability to do simple math leads to a few bizarre conclusions within their numbers:

It gives Romney a 52-45 win among all women - which is about as ludicrous as Sarah Palin making sense. The idea that American women, as a group, prefer the candidate in favor of repealing Roe v. Wade and explicitly opposed to women's pay equity to a pro-choice president who first signed legislation was the Lily Ledbetter Fair Pay Act is not just an affront to the intelligence of women but is not even in the realm of possibilities.

That evidence of manipulation repeats itself when one looks at CNN's polling on the ethnicity breakdown as well. It gives Romney a 63-34 edge among white voters - even with the super racist Republican party, one generally only finds that types of disparity in the deep south or red pockets of other states - and Obama is up among non-whites by 67-30. By way of comparison, Barack Obama won 80% of the non-white vote, and 40% of the white vote in 2012. In 2012, even Republicans admitted that their ceiling among the white vote would be 60%, and the idea that minorities have suddenly grown cozy with the party fully dedicated to white supremacy only lends credence to the idea that CNN polled more than its share of uncle Toms.

But but. Then how come Hillary Clinton is beating Romney so handily in the same poll? According to the poll's crosstabs, the advantage comes from a combination of higher Democratic support and a major swing (2 out of 3) among non-leaning independents polled (assuming you hold Dem-leaning independents constant). Essentially 2 out of three all the non-leaners that would choose Romney over Obama would have to switch to Clinton over Romney (the rest can be chalked up to sampling error and the fact that presidents in office actually have to do things rather than simply bloviate).

It will come as no surprise to TPV readers that the Left has its own racist corner, which the CNN poll seems to be capitalizing on. It is almost as if in order to get their Democratic, liberal and Dem-leaning samples, they found and reached out to as many PUMA Democrats as possible - i.e. self-described disgruntled fringe Democrats from the 2008 primaries who never could stand this president.

Think about this. What can be responsible for ALL of the following?

  • ALL non-leaners would support Romney if the election were held today.
  • 8 percent of Democrats surveyed won't vote for Obama but would gladly cast a ballot for Hillary Clinton (91% vs 83%).
  • Two-thirds of non-leaners in Romney's corner against Obama would support Clinton if she and Romney were facing off.

There's a term for these Clinton-leaning but anti-Obama Democrats and independents: PUMAs (which stands for 'Party Unity My Ass'). The people who could not stand Barack Obama's nomination in 2008 and have ever since been throwing monkey wrenches into his presidency from the Left. Many of these people were fervent Clinton supporters in 2008, and left the party over her loss. PUMAs were heavily... white women - which would account for both Romney's sudden advantage among women in the poll as well as Hillary Clinton's subsequent edge.

In essence, CNN's "non-leaners" are made up of one part Republicans and two part PUMAs.

This is not to say that Hillary Clinton supporters from 2008 by and large became PUMAs - they did not. This is to say that PUMAs were a small contingent of Clinton supporters in 2008, and CNN seems to have found them for this poll.

But it should caution potential Clinton 2016 operatives - this poll makes a Clinton 2016 victory look far easier than it actually will be. Counting on PUMAs won't be enough. While I will support Secretary Clinton should she become the Democratic nominee with no hesitation, I fear the CNN poll is overstating current Democratic support for the former Secretary of State.

It's not like CNN cares. They flatly said one thing about party distribution in their sample and literally used a different distribution in the actual sample, hoping no one notices - and frankly, no one has, until now. The TV as well as online punditry have thus far failed to note the curious fact that if PUMAs account for the vast majority of "non-leaning independents" and anti-Obama Democrats that alone explains the numbers even after they have been cooked.

There are lies, damned lies, and statistics. This doesn't mean statistics is junk science. But it does mean that it is incredibly easy to manipulate numbers in statistics. When properly done, statistics are an invaluable tool for the truth. When manipulated or cooked, it is a dangerous weapon in the hands of the manipulator as well as the manipulated.

Like what you read? Chip in, keep us going.

The Real Michele Bachmann Is So Much Worse Than the Fake One

Looking Back: An American History Lesson from the year 2034