A breathless article by Rebekah Metzler at USNews.com informs us that Sen. Mitch McConnell is in trouble in my native state because "his Democratic opponent, Alison Lundergan Grimes, leads him in his 2014 re-election race, according to a new poll."
Though this may make the hearts of Kentucky Democrats go pitty-pat, once you look past the opening paragraph the whole thing evaporates.
Ms. Grimes's exciting lead, it turns out, is one percentage point, 45 percent to 44 percent. The margin of error, quoted in the last paragraph, is plus or minus 2.8 percent. This is not a gossamer lead; it is a statistically nonexistent lead. Both candidates are floating somewhere between 41 percent and 48 percent, and no one can say with certainty which one is ahead.
Moreover, the poll of 1,210 voters was conducted on behalf of "the Democratic polling firm Public Policy Polling." Hmmm. Well, we don't know the confidence level of the poll, or exactly what population of "voters" was sampled, or how the questions were worded; but, even so, it seems just within the realm of possibility that Public Policy Polling managed to provide a result that the client was hoping to get.
Not that it means that much who might be ahead today. The polling was conducted in the latter part of July, sixteen months in advance of the election, at a time that it seems unlikely that a multitude of voters in the Commonwealth had focused any particular attention on this race.
It's not as if there were no way of avoiding the shoddiness of this article. At HeadsUp: The Blog, Fred Vultee has posted repeatedly about how to evaluate and write about opinion polls; his post from 2006 should be instructive. And the Associated Press Stylebook, however much it has been tweaked on these premises, has an excellent set of directives on writing about polls and surveys; it can be found on pages 214-16 of the 2013 edition.
Perhaps someone at USNews.com might pick it up and have a look.