Saturday, October 13, 2012

Speaking of Pollsters ...

... this looks fishy. The Behavior Research Center has Obama ahead of Romney in Arizona 44-42. There are lots of fishy things in this poll and so perhaps it serves as a good jumping off point to discuss what kinds of things jump out at me as not quite right when I'm looking at a poll:

1) Is the poll an outlier (especially compared with pollsters I think are good)? Yeah, to say the least. PPP did a poll in Arizona BEFORE the first debate and had Romney up by 9.

2) Have I ever heard of the pollster? No, I haven't. Every pollster gets some media attention the first time sometime but when there's no track record to say this poll should be believed, it should be read with a skeptical eye.

3) Are there numbers below that topline that seem counterintuitive? Yes. The gender gap is very narrow here for one and the poll has Obama and Romney tied among men. The poll also has Obama ahead 77-10 among Latinos. There's other things that don't feel right but those are the biggest ones.

4) Are the campaigns behaving in a way that is consistent with the polls? You can figure out that Ohio is close in both campaigns' internal polls by the way the two campaigns are behaving (running a million commercials and campaigning there). We don't see the two campaigns behaving as if Arizona is in play.

5) Is it a live-interviewer poll with cell phones or is it a robo-poll or is it an online poll? The order in which I listed them is in descending order of quality and this is an online poll.

So, this poll has all the hallmarks of a poll that I would not tend to believe. It doesn't mean it can't be right. It just means ... it probably isn't right.

Now, compare all that with this PPP poll in Ohio that was just released that has Obama up by 5:

1) Is the poll an outlier? Not too much of an outlier. CNN had Obama ahead by 4 a few days ago and NBC/WSJ had Obama up by 6. Others have it closer and PPP does have a slight Democratic lean in their polls. But this is not a shocking number.

2) Have I heard of PPP? Yeah, and they have a generally good track record.

3) Are there numbers below the topline that seem counterintuitive? Mostly no. For instance, Romney has a 46/51 fav/unfav. That sounds about right, especially in Ohio. Gender gap? Yeah. Obama leads 54-42 among women and trails 50-48 among men. I'd guess the Obama might be a tad more behind in Ohio but this is not a crazy number.

4) Are the campaigns behaving in a way that is consistent with the poll? Oh yeah, to say the least.

5) Live-interviewer or robo-poll or online poll? Robo-poll. Not ideal but at least it isn't online.

This poll mostly passes the smell test. There are a few things I scratch my head on. PPP says 19% of voters have already voted and those voters went 76-24 for Obama. That's hard to believe. If it is true, Romney is in a DEEP hole in Ohio and NBC/WST had a similar finding in their poll though not quite as big a margin. But look, I think it is clear that Obama leads by a little in Ohio (even Rasmussen has Obama ahead by 1 there) and this poll is not far off if it is off.


Anonymous said...

Latino Decisions had a poll of Arizona and the President led 80/20 with latinos. If the latino voters turn out in large numbers (30% of Arizona is latino) due to the anger in the community re: new Arizona immigration laws, this could be a surprise on Nov 6th.

Larry Becker said...

Yes, I thought about the possibility that SB 1070 might explain a larger margin among Latinos than in other states and Nate Silver also points out that this poll did some Spanish language interviewing and that this has been shown to do a better job of getting at the Latino vote numbers. But, that said, this was not 80/20. It was 77/10. Deadly for Romney if true. But hard to believe.