Friday, October 4, 2013

Bad Poll? Blame and the Shutdown

There's a story on my local newspaper's site today about that age-old question -- assigning blame -- this time about the federal government shutdown. But this is a Georgia question. The hed:

Poll: Most Georgians
blame Democrats
for government shutdown

I'm not writing to blame one side or another; others do that better than I. No, I'm here to raise (yet again) questions about how a poll is reported. And don't blame just me. My colleague Lee Becker also noted the odd methodological quirk I'll focus on below, and he knows a lot more about running surveys than I do.

Let's look at the first two grafs:
    ATLANTA — A plurality of Georgians blames Democrats for the shutdown of the federal government and favor repeal of Obamacare, according to a new poll of registered voters.

    The survey of 1,000 active voters was conducted Tuesday and has a 3 percent margin of error.
Okay, it is damn near impossible to run a quality survey of a thousand people in one day unless it's a robo-poll. You know these polls, they're those annoying computer-driven calls that due to the vagaries of who is willing to participate tend to skew older, and Republican. Important methdological note -- they can't do cell phones, making their sample really skewed away from younger people.

Hell, one of the two orgs that did this poll specializes in robo-polling. Says so at the bottom of their site's front page. Details about their robo-calling service is here. So you'd expect the poll to skew to the right, meaning it should be more likely to blame Democrats, especially in a red state like Georgia.

But lets look at something else, the response alternatives, which in technical terms means the choices a survey respondent is given on the phone by that annoying computer voice.
Here's the telling graf:
    It showed that 33 percent of those questioned blame President Barack Obama for the shutdown, 13 percent blame the Democrats who control the Senate – a combined 46 percent. On the other hand, 39 percent fault the Republicans who control the House of Representatives, and 14 percent consider both sides equally culpable.
Notice something? Of the choices offered for blame, two of the four are Democrats, one is Republican, the other is equal blame.

As we say in the journalism side of academy: what the fuck?

Then if you collapse two of the four back into a single category you're surprised Democrats (win?) at being blamed? Maybe they are to blame, but that's not my point here. My point is we're better served by good polling and, dammit, good journalism about the polling. Report polls with skepticism, especially if the polling org used a method that tends to skew one way or maybe has a partisan axe to grind. Landmark Communications, the other sponsor? See below:
Landmark has worked for over 100 Republican state legislators since our firm's inception in 1991. We are proud of the part we played in helping build the new Georgia Republican majority, working to help elect many good Republicans to districts where no Republican had won before. 
C'mon. That doesn't mean they don't do good polling. It does mean they come from a specific partisan tilt, something that deserves mention in an honest news story. He does lots of work for news orgs too, the guy who runs Landmark, so give him his due.

Note there are few methodological details in the story about how the poll was conducted. Not there are no damn links to the sites where you can read the actual poll questions, see the methodology. No, you gotta go look yourself for it. If you find them, let me know.


Dr. Becker, being both smarter than I and not buzzing on pain meds, found some methodological details of the survey I missed. Key points:
  • The survey used ACA or Obamacare, not Affordable Care Act. Yes, this can make a subtle yet significant difference in the results.
  • No sign the poll randomly rotated the possible answers. This is standard practice on professional polls that some would randomly receive Obama to blame first, some might get him second, or third, and so on.
  • He noted the federal rules against autodialers for robo-type calls and the AAPOR standards for this.
  • I'd do more, but time for another pain med. Had work done on my vocal cords.

1 comment:

Lee Becker said...

Some data are here:

No indication that the response categories were rotated randomly.

Question 2 on the Affordable Care Act never uses that name.

The data marginals suggest young people are underrepresented.