Thursday, September 20, 2012

The Omnivore Interviews Public Policy Polling!

Polling occupies a major portion of the mind-share here at the offices of The Omnivore (it may not be entirely accurate, but imagine a massive, light-filled open-plan office with lots of creativity-inducing empty space and, most importantly, the staffers wearing pants). As such, we wanted to get a few words with one of our favorite pollsters: Public Policy Polling!

Public Policy Polling (PPP) is a robo-calling player who is generally identified as working for Democrats and has teamed up with the Service Employees International Union (SEIU)--another left-identified outfit. This isn't to say they don't have a reputation for accuracy. Let's check the Republican leaning National Review Online from 2010:
Overall, PPP has a reputation for reliability; in 2008, the Wall Street Journal looked at various pollsters’ results and ranked PPP among the best. Its last poll in the New Jersey governor’s race had Republican Chris Christie winning by six points (he won by four), its last poll in the Virginia governor’s race had Bob McDonnell winning by 14 (he won by 18), and its last poll in the Massachusetts special Senate election had Scott Brown by five (on the nose).
On the other hand, they have a serious reputation in conservative circles for "over-sampling Dems." This means that when they go and robo-poll a bunch of people they ask their party identification there's usually a LOT more Democrats in the list-of-people-called than Republicans. So of course they find that the preferred candidate is the donkey!

As PPP has gotten a lot of push-back, they have taken to re-tweeting negative tweets sent to them on Twitter! It's difficult for The Omnivore to describe it's love for this: words haven't been invented--but a word was invented it would probably be 'Dramaluzariffic.' It's really a good thing that word wasn't invented.

So we tweeted PPP and asked them to do a 15 min phone interview with us on the topic of polling conspiracy theories. They didn't respond--but ... we know tweets sometimes "get lost in the mail." So we sent them a nice email with the topics and some of the questions we'd ask. Nothing. No response. We know that emails sometimes get, erm, lost by the server.

So we pivoted ... we picked up the phone and called Clint Eastwood: "Hey, Clint," we said, "We're having trouble reaching PPP--what should we do?" "They're a robo-calling polling firm, you idiots," the former mayor of Carmel told us. "You pick up the phone and call them."

So we did ...
IMPORTANT: Phone Is Unplugged!

The Omnivore: Hello PPP, thanks for taking the time to speak with us today.

PPP: It's great to talk to you.

The Omnivore: Okay, first things first. You over-sample the hell out of Democrats. What's up with that?

PPP: Let's be clear: firstly, when we conduct a poll we make sure we survey more than enough people to get an accurate sample size--that allows us to randomly reject individual surveys from the sample if certain demographics are over-represented. The random selection removes any potential bias with that step alone. Then we mathematically weigh the results to the actual demographics. Our results are transparent and accurate. Don't focus on the D+12 number--keep in mind that the actual end-state results mathematically reflect the entire population we polled.

The Omnivore: Yeah--but how do you know what the "actual demographics" are for, say, Virginia?

PPP: Well, there are a lot of ways to do that and they're kind of proprietary.

The Omnivore: So you can't tell us?

PPP: It's not on our website anywhere you can find it.

The Omnivore: But you do NOT use party affiliation. Isn't that like, all important?

PPP: There's some conventional wisdom that way--but we don't think so. We weight for race, age, and gender based on a combination of census numbers, voter participation in past elections, and poll response. The mathematics make this pretty accurate.

The Omnivore: Can you explain the ... uh ... math to us?

PPP: How long do you have? And how smart are you?

The Omnivore: Uh ... Never mind--but look: so on, like, Sunday Todd Akin makes his 'Legitimate Rape' comments and the whole world blows up--but suddenly there's PPP with a poll showing him with a lead--even after the comments over Claire McCaskill. It has a +11 Republican sample. Do, uh, any of your other polls have a +11 Republican sample?

PPP: Very few, if any, admittedly. Maybe more Republicans were picking up the phone that day?

The Omnivore: Maybe. But then Akin clings to his victory in the primaries and his lead in your poll like he clings to guns and religion and stays in possibly costing the GOP a senate seat. Coincidence?

PPP: You think we cooked up a poll just to make Akin stay in?

The Omnivore: Lot of people do. Isn't it true that lefty-Daily Kos guy sued Rasmussen for "fixing polls" to 'create a narrative'? Didn't the Daily Kos lawsuit allege that polls could have an impact on voter behavior leading to making commissioning a fixed poll sort like ... say ... an advertising campaign? But one that gets priced in to every polling aggregator on the planet?

PPP: Well, in that case Nate Silver took a look at their numbers and the math didn't add up. They did get an AAPOR rebuke. A lot of people take a look at our stuff and we've never been sanctioned. Many people don't understand polling--the math is sometimes counter-intuitive so it leads to conspiracy theory. Especially when people don't like the results.

But look at it this way: who would ever think that a hard core Republican like Todd Akin would trust a 'fixed' PPP poll? I mean, if we were going to fake a poll, wouldn't we want to fake one that had a chance of being taken seriously? And if we did fake the numbers, wouldn't we have something like, I don't know, D+3? Something that "readers would buy?"

The Omnivore: Good point, I guess. What Republican can be expected to read an R+11 poll and go SEEMS LEGIT?

PPP: In Todd Akin's case: 'SEEMS LEGITIMATE.'

The Omnivore: Leave the humor to us. No, wait--aren't you guys the ones who asked if Mitt Romney should get some of the credit for killing Bin Laden?

PPP: We did! About 15% of Ohio gave Mitt credit for that! We wanted to see if some people would credit Obama for anything.

The Omnivore: You also found that Charlie Sheen would beat Sarah Palin. What do you say to charges that you set up these crazy questions just to embarrass Republicans?

PPP: Hey--we report, you decide.

The Omnivore: Ha. Ha. Okay, here's a question for you: your CEO Dean Debham in a fist fight with Scott Rasmussen--who wins?

PPP: We'll go with D+17 for the knock-out.

No comments:

Post a Comment