CNN Is Just Making Up Poll Numbers Now
Okay, I'm not certain they are literally making up poll results, but the poll CNN and British market research firm ORC International released Thursday afternoon is so screwy and raises so many questions that they might as well be doing it intentionally. If CNN is already resorting to these kinds of tricks before the conventions have even started, it's going to be a very long campaign.
First let me say, with less than three months to go in this campaign, can we please stop polling the political views of "adults", rather than "registered voters." We really should be moving soon to a "likely voter" screen, but I'll take "registered voters" for now. Knowing the political views of unregistered voters is worthless at the height of an election campaign and serves no real purpose but to give the Democrats an advantage. Nate Silver at the New York Times has estimated that polls of adults are biased towards Democrats by around 7 points.
CNN does at least test an Obama-Romney match-up among registered voters. It trumpets across its news page that, among registered voters, Obama is leading Romney 52-45. Looking deeper into the poll, however, we learn that these numbers include "leaners." In other words 52% of registered voters either support Obama or are "leaning" towards him. Now, there is nothing wrong with documenting voters who are "leaning" towards a candidate. But, you know, you have to document it! Knowing the percentage of voters who support a candidate and those who are currently "leaning" towards that candidate is what we call relevant information.
For example, if 50% of voters say they support Obama while 2% are leaning towards him that tells us one thing. If, however, 45% support him and 7% are leaning, that is kind of something else entirely, right? If you are going to combine two different numbers to get a final result, you kind of have to show your math. CNN obviously has this information, so why didn't they publish it?
CNN takes another questionable turn when they produce a sub-sample of Republicans for a poll on potential VP picks. They report their sample is 419 Republicans and Independents leaning Republican. Wait, what? How many of each, exactly? You can see how that information might make a difference in the results. Independents who lean Republican are not exactly the party base. If, however, they make up a majority of CNN's sample, it might not accurately reflect the views of the base GOP. Again, CNN, you have to show your math.
At least this sub-sample gives us a small glimpse of the partisan make-up of the poll. Naturally, CNN doesn't provide any information on this directly. According to the sub-sample, though, Republicans and GOP-leaning Independents make up about 45% of CNN's overall sample of registered voters. (Hey, that's Romney's vote share!) So, around 55% are Democrat or Democrat-leaning independents. By this, not only is the poll a heavily skewed D+10, but Obama is underperforming at only 52%.
Maybe there is some large number of Independent voters in the poll who lean neither way, but we don't know, because CNN won't tell us. At this late stage in the campaign, if you don't show your work on a poll, then I'm just going to conclude you're making it up.
After all, at the end of June, CNN and ORC International conducted a poll and found that economic optimism was "skyrocketing." How is that result holding up?
Follow me on twitter.