ELECTIONATE JUNE 27, 2012
-
Read Later
READ LATERAvailable only to subscribers. SUBSCRIBE TODAY
-
Listen
ARTICLE AUDIO
- Font Size
So, you think Obama leads by 8 percentage points in the swing states, as suggested by last night’s NBC/WSJ survey? Before you jump on the bandwagon, understand what that entails: a blowout.
In 2008, Obama carried NBC/WSJ’s twelve swing states by 7.7 percentage points. A result like last night’s poll would require a repeat performance, even as most polls show Obama’s standing substantially worse than four years ago.
Obama’s big 7.7 percent advantage was driven by decisive victories in several states, including big states like Michigan and Pennsylvania. Without such large margins of victory, Obama’s narrow wins in populous Ohio, Florida, and North Carolina would have swamped Obama’s impressive showings in tiny Nevada or New Mexico.
If Obama leads by 8 percentage points in the swing states, where is Obama running up the score? According to recent polls, Obama doesn’t lead by 10 percentage points in any state, except New Mexico, which represented a meager 2 percent of votes cast by these 12 states in 2008. In fact, not one poll over the last month has shown Obama leading by more than 10 points in any of the other 11 states. Obama is not leading in Florida, Ohio, or North Carolina by such a large margin that it obviates the need for Obama to run up the score on more favorable turf.
Perhaps new polls over the next few weeks will show Obama with a growing advantage in the battlegrounds without commensurate gains nationally. In the meantime, defer to dozens of state polls and history over a swing state sub-sample with a large margin of error from a single national poll.
1 comments
Okay, but even with the small sample size and consequent large margin of error, an 8 point advantage is still statistically significant. You're discounting the result because with a 6.2 MOE, Obama's real advantage in the swing states might be as little as 1.8, and yet you neglect to mention that it is literally just as likely (that is, not very likely at all) that Obama's real advantage is a whopping 14.2! Now, it might be the case that those 249 swing state voters were not properly distributed across the 12 states under consideration, i.e. a potential problem with a subset analysis such as this one is that inasmuch as the sampling procedure was designed to model voting preference across the whole of the USA and not just the swing states, through pure random chance, it could happen that a low population state that strongly favors one candidate or the other might contribute more than its share of subjects to the poll. But that's a separate issue to the one you've raised about sample size. Ideally a poll designed specifically to answer the swing-state margin question would need to be somewhat bigger--though if your starting hypothesis was that Obama's advantage there was indeed 8% your poll wouldn't need to be that much bigger. Perhaps more importantly, though, it would be stratified by state, i.e. however many subjects you decided you needed in your poll would be gathered from each state in numbers proportional to that state's population/the total population of all 12 states.
- AaronW
June 28, 2012 at 10:00am