September 28, 2012 the polls are all right
There’s been some news media blather about how the polls (which show a month-long growth in Obama support) are biased in favor of Obama.
So I looked at the last set of national polls.
The Rasmussen poll historically tends to favor Republicans a bit more than other polls. So it seems to me that if another major poll is quite different from Rasmussen, either Rasmussen is too pro-Republican or the other polls are too pro-Democratic.
Indeed, there is a difference. Rasmussen has 47-46 Obama, while the most pro-Obama poll (National Journal) has a 50-43 Obama lead.
But this is not a statistically significant difference. A national poll typically has about a 3 point margin of error. So if a poll shows Romney with 43 percent, he could have anywhere between 40 and 46. Similarly if a poll shows Obama with 47 he could have 44-50.
In other words, the difference between the most pro-Obama poll and the most pro-Romney poll could just be sampling error: the difference between the Obama high of 50 (National Journal) and the Obama low of 47 (Rasmussen) is not statistically significant. Ditto for the difference between Romney’s high of 46 and his low of 43. So in reality (if you can use the word “reality” to describe a poll a month before the election) Obama probably has somewhere in the high forties, and Romney somewhere in the mid-forties.
I wouldn’t feel particularly safe with that kind of lead if I was in the Obama camp.