Wednesday, June 14, 2017

More District 6

I know we're watching stories on the London fire and the GOP shooting today, and rightfully so given the tragedies, but the 6th Congressional District race still matters, at least politically. The Democrat, John Ossoff, had enjoyed a couple of 7-percentage-point leads in recent polls, but a new one has it as a dead heat (freaking out Dems, heartening Republicans). It's really important to peek under the hood on this one. Indeed, the folks who conducted the poll included this warning:

I love that, cut corners recklessly, which kinda defines of a lot of political reporting, especially on cable news. Anyway, the text on the linked page above goes to some length to explain why the sudden change may or may not matter. It's worth reading, the problem is it's hard to incorporate such caveats into political stories, especially on television -- a medium ill suited to nuance.

Among those respondents who say they're already voted, Ossoff has a 19 percentage point advantage over Karen Handel. Among those who have not yet returned an early ballot, Handel has a 14 percentage point advantage. What's that tell you? It's gonna be damn close. Dems tend to do well with early voters and do well in high-turnout races.

Then there's this fascinating warning in the poll methodology that you should read:
3 weeks ago, a SurveyUSA poll for WXIA-TV had Ossoff 7 points atop Handel. Some of the change in outcome poll-on-poll may reflect sampling vagaries and not reflect actual movement in the contest. Today's survey has fewer high-school educated respondents and fewer lower-income respondents than did SurveyUSA's sample 3-weeks ago. Both high-school educated voters and lower-income voters back Ossoff. Today's sample is older than SurveyUSA's release 3 weeks ago. Older voters back Handel. If you see these 3 differences as bugs not features, you can argue that Ossoff is today no worse off than he was 3 weeks ago.
So we're talking about the vagaries of sampling here, the kind that make it tough to call a congressional race in the first place. Interesting weird factoid, if I'm reading this right. Keep in mind GOP candidates do better in landline surveys, Dems in cell phone surveys, given the ages of respondents who rely on each. In the previous survey, Handel won on in the landline sample 52-46 and Ossoff won on the cell sample 57-34. In the newer survey, Handel again won the landline 52-46 but Ossoff only won the cell sample 48-42. What's that mean? No idea. If I had more time I'd dig and figure it out, but it just could be statistical noise.

No comments: