You are using an outdated browser.
Please upgrade your browser
and improve your visit to our site.
Skip Navigation

Stop The Battleground Subsamples!

Last week, NBC/WSJ blessed the political community with a battleground state subsample showing Obama leading by 8 percentage points in twelve critical states. Predictably, NBC/WSJ’s finding revived the spring tale of Obama’s decided advantage in the Electoral College, which would allow him to decisively win a close national contest. As I mentioned at the time, a small sample and contradictory data combined to cast great doubt on that conclusion.

This week, CNN Opinion Research followed NBC/WSJ’s footsteps with a similarly misguided polling adventure. Just like the NBC/WSJ poll, CNN found Obama leading Romney by 3 percentage points nationally, 49-46, and decided to produce a battleground subsample. But unlike NBC/WSJ, CNN’s battleground subsample produced the exact opposite result: Romney leads by 8 in CNN’s battleground subsample, while Obama leads by 8 in NBC/WSJ’s.

What explains the gap? First, the two battleground subsamples are not identical. CNN relied on a broader definition of the battlegrounds by including Indiana, Missouri, and Arizona—three relatively red states currently deemed unworthy of ad spending by either campaign. As a result, CNN’s battleground subsample should take on a slightly redder hue than the NBC/WSJ sample. But those three states still don't account for most of the gap between the NBC/WSJ and CNN battleground polls. In the last election, the 15 CNN battlegrounds were slightly more Republican than the country as a whole, but only by 2 percentage points—not the massive 16 percentage point gap between NBC/WSJ and CNN, or the 11 percentage point gap between the battlegrounds and the country.

To CNN’s credit, they also relied on a much larger sample, which reduces the margin of error. But the CNN battleground subsample has a bigger problem—its result is implausible. If the battleground states were really 11 percentage points to the right of the national mean, they wouldn't be battlegrounds. For comparison, the CNN sample found Obama losing in the South—a decidedly non-battleground region—by 10 percentage points. At the same time, Obama and Romney were in a dead heat in the West and Midwest. Since almost the entirety of the Midwest is part of the CNN battleground, and Obama is competing in three southern states that are indisputably far friendlier to him than the rest of the South, it becomes impossible to reconcile Obama's standing in the battlegrounds with the CNN regional breakdowns.

So what's the big deal? The utility of battleground subsamples is marginal at best; at worst, the polls risk counterproductive media coverage. Since different battleground polls sample different battleground states, the inevitable comparisons between the polls are meaningless. And the subsamples include fewer respondents—and are thus less accurate—than a national poll, let alone the dozens of state and national polls conducted every month. But the biggest strike against them is that they're just unnecessary: In 2008, the battlegrounds voted similarly to the country, and there are dozens of state polls showing the battlegrounds behaving similarly to the rest of the country in this election as well. With that history, it would be unwise to give any weight to battleground subsamples—whether they show Obama or Romney in the lead.