STRANGE CURRENCIES

Life and the job market have established that I'm not worth a whole lot and I generally lack useful knowledge or skills, but I know a thing or two about polling – at least enough to recognize something funny going on. And the disconnect between the current dominant media narrative and some of the numbers we're seeing looks an awful lot like shenanigans.

First, looking at the generic Congressionals you'd hardly know that for the past year the media have breathlessly covered TEA PARTY!!11!! and the impending GOP revolution:

Hmm. Now, generics are among the least useful polls, mostly because of Fenno's Paradox – people disapprove of Congress but keep re-electing their own Congressman. More broadly, the phenomenon means that expressing generic preferences for a party doesn't tell us who these poll respondents will actually vote for in their own district. With 95% of incumbents being re-elected in recent years, it's more likely than not that a person's generic preference and actual vote are poorly correlated. Nonetheless, if there was some sort of "revolution" brewing, I have to imagine that we'd see slightly more favorable numbers for the GOP. We're five months out and they're losing to the generic Democrats.

Second, a lot of the numbers coming out of Rasmussen are supporting the theory that they are going the way of Zogby and making the purposeful transition from legitimate pollsters to GOP Propaganda Services. This year they have been the only agency – like, literally the only one – consistently showing Rubio in the lead in Florida. Even when other agencies were showing Crist winning a three-way matchup, Rasmussen had Rubio up 20. That makes absolutely no sense. Variance among polls is expected but not on the order of 20% unless one of the parties involved is seriously off the scent. Check this out (all Rasmussen):

5/3/10 500 LV
Rubio 34%
Meeks 17%
Crist 38%
Undecided 11%

13 days later:

5/16/10 500 LV
Rubio 39%
Meeks 18%
Crist 31%
Undecided 12%

And while that was happening:

Rubio Favorability

04/21/10 Rasmussen
Favorable 52%
Unfavorable 37%

5/16/20
Favorable 46%
Unfavorable 43%

And…

Crist Favorability

04/21/10 Rasmussen
Favorable 55%
Unfavorable 40%

5/16/20
Favorable 57%
Unfavorable 41%

So his favorables fell 6% as he was taking the lead away from Crist in the head-to-head (to head) matchup. I've been to two county fairs and a live Carrot Top show, and this is the dumbest thing I've ever seen.

The problem inherent to Rasmussen (and every other agency, to some degree) is their bizarre, proprietary "likely voter" model. You can basically make a poll result look however you want by carefully parsing the definition of a LV. Rasmussen's LV model was disastrously wrong in 2008, as it was structured around the assumption that only old, white Republicans actually show up. I'm willing to bet that they're doing something similar here. If the electorate is defined as geriatric teabaggers, Rubio's going to look pretty good. Maybe the gamble will pay off for Rasmussen – midterm turnout is unpredictable and their guess may be as good as any regarding who is actually going to show up for this thing. Maybe it will be nothing but teabaggers. Maybe it won't.

I remain 100% convinced that this will be a year of Republican gains, but the numbers trickling in as the election heats up are wildly inconsistent with a sweeping Republican victory. Anti-incumbency might be the closest thing to a theme this year (just ask Bob Bennett, Arlen Specter, or Blanche Lincoln). Whatever happens, the teabaggers will declare victory but their Glenn Beck approved candidates have done horribly thus far. Defeating Bennett in the Utah primary was probably the first victory they can claim, and even that was probably unrelated to his opponents' Teabag credentials.

As expected, we're starting to see a little bit of a pullback from the GOP high water mark after the Scott Brown MA special election victory, after which gasbags were predicting a GOP takeover of both chambers of Congress. Predictions are getting a little more muted and a number of Democratic Senate candidates are not quite as dead as previously claimed. They'd actually be in great shape if they could re-energize the base that came out in droves in 2008. The odds of that happening without a major legislative victory apart from a confusing, watered-down health care bill are quite slim.

Be Sociable, Share!
Tags: ,

12 Responses to “STRANGE CURRENCIES”

  1. party with tina Says:

    Really, you can only make a poll look one way or another if you take it's context away. I doubt that pollsters are faking results in order to rally voters, that would be the job of political opinion broadcasters and writers with agendas, no? Benefit of the doubt…

  2. beau Says:

    @party with tina – That's a lovely straw man you've built yourself there. Now, if you only had a brain…

  3. bb in GA Says:

    Ultimately to stay on the stage (and get $ for your other work) you have to maintain credibility by being mostly right AT THE END in that polling biz, correct?

    If you are juicing the stats in a direction that you (or your Masters) favor, when the Big Tuesday approaches (the last X weeks) you have to do a slide job to close the gap between your favs and reality.

    If you don't, you become dog meat with everybody ripping you on the TV etc. while your revenues head elsewhere.

    Which of the major pollsters do you think is/are willing to jump on the grenade for the GOP at the end?

    //bb

  4. Brandon Says:

    Maybe it's the poli sci dork in me, but I sure do loves me some "Ed and the Polls" posts.

  5. ladiesbane Says:

    Do polls sway voters? This is not a rhetorical question. Whether a strong lead encourages some idiots to vote for whoever looks like the winner, or discourages the projected losers (or mobilizes them to action), does following polls make a difference to the outcome of the elections?

  6. Ed Says:

    LB, I am currently in the middle of conducting an experiment on that question with a colleague at University of Nebraska. I'm really looking forward to seeing our results and I'll be sure to talk about them.

  7. Ed Says:

    bb, Rasmussen isn't necessarily looking to fall on a grenade for their GOP buddies or anything, but they define the likely turnout in a way that, in my opinion, biases their results toward the right-leaning candidates.

    In 2008 the election proved their turnout assumptions to be incorrect, while in 2004 they were pretty accurate. The way they do things is essentially a dice roll, because they're making a prediction based on a prediction (the outcome based on who they assume is and isn't going to show up). It usually makes them look brilliant or really bad.

  8. ProudLiberal Says:

    @Brandon. I concur. When Ed talks contextual politics and theories, it gives me the tingles. I'm a poli sci dork, and when a man starts talking poli sci, he could be dog's ass ugly, and I'd still find him irresistibly attractive.

    On to bidness: Ed, we read a couple studies in grad school like, five years ago that seemed to imply that biased polling can confirm how one was leaning anyway. So, if you were undecided, but leaning toward the winner, you were more likely to head that direction. If polling showed you were undecided, but leaning toward the loser (in the poll), then you may be inclined to not vote at all. Mind you, the study wasn't on that set of phenomena specifically. It was more like an environmental factor that could or could not exist, period. I wish I could remember the author and the actual study, but, sadly, I was drunk through most of grad school.

  9. Entomologista Says:

    I will not be voting for that choad Ben Nelson, that's for sure.

  10. party with tina Says:

    are you not voting for ben nelson because he's a choad, or is he a simply a choad, whom you will not vote for?

  11. Don Says:

    All I got out of this was that you went to a Carrot Top show. Everything just went blank right there.

  12. Patrick Says:

    If you look at the pollster numbers now, they show Republicans tying with Democrats in May 2010. I'm not sure if the results were incomplete or they performed some sort of revision…