The Polling Industrial Complex and The Media "Horse-race" Narrative


The Polling Industrial Complex and The Media "Horse-race" Narrative
by nancy a heitzeg


I am writing this in the immediate aftermath of the Joe Biden Beat-down of Lyin Paul Ryan, aka the VP Debates.

The Sunshine Patriots and Fair Weather White Guys got what they claimed they wanted all right -- the Democratic Party's #1 Enforcer had high-sticked, body checked, fact-checked and dropped the gloves, all with in the span of 5 minutes.

And for 90 minutes, He. Never. Let. Up.

Ryan was left with nothing but smirks, endless water sipping and spin, including talking himself into sending more troops into Afghanistan, scaring seniors with Social Security privatization and Medicare vouchers, and denying the right to choose with a new twist - now Position #671 on Abortion from the Romney/Ryan campaign.

A clear win -- the contrast was stark and Ryan was clearly out of his league. On all counts.

So were the Hacks happy now?

Well sort of. But tonight -- unlike last Wednesday's melt-down - they decided to be "professional". Their right leaning counter-parts who declared the bullying Romney a clear victor on "style" were now reduced to complaining that Biden was too assertive. And last week scowling bad, but now smiling is out.

And the insta-polls? A CBS snap poll of undecided voters (no demographic detail provided) had 50% declare Biden the winner. And CNN, whose sample of old white men from the South last week declared Romney the overwhelming winner, now reports that a sample with fairly even Republican Democrat and Indie splits ( no demographic at all) declares Ryan the Winner 48% - 44%.

And if the waters were not muddied enough, Nate Silver decides --within an hour, before the dust is even settled - to report these two "polls" as "evidence" of a Biden "hold". No clear winner - no probable poll impact.

As Joe Biden would say, "Malarkey!!"

But but but the Page Clicks and the Horse Race must go on.




The Media - Polls Feedback Loop

In this Election Cycle, it is increasingly clear that, for whatever reason, media narratives and polls are self re -enforcing. Media headlines report polls without context and/or analysis. Shabby pollsters ( and there are plenty this cycle) use polling cycles to drive the narrative, and long standing pollsters may find shifts in opinion that are shaped by media narratives. Media reports that back. Feedback-loop..

The so-called Romney Debate Surge is a case in point.

Kevin Drum of Mother Jones sums up this pattern with regard to Presidential Debate #!:
.. the polling I saw suggests that viewers polled during the debate thought it was about even; viewers polled right after the debate though Romney had won; and viewers polled a little later still thought it was a rout. I can't think of any good explanation for this aside from the effect of the talking heads right after the debate and the firestorm of liberal criticism that quickly turned into a feeding frenzy of outrage. And generally speaking, I'd say there's plenty of historical evidence that media coverage of presidential debates has much more impact than the debates themselves.
Media ( and of course, the GOP) has an interest in a horse-race and wasted no time in attempting to construct the first debate-- in spite of historical evidence to the contrary - as a "game-changer".

A Black Swan Moment.

Here's how the narrative and the polls collided ( or colluded ?? -- your choice):
  • "Liberal" media, both TV and blogland, throws a tantrum over President Obama's performance and gives Romney a pass on bullying racist dog-whistles and a blizzard of lies
  • CNN reports an "insta-poll" of viewers that has 69% believing Romney has won. Of course they fail to tell us that the sample is nearly entirely on over 55 white men from the South
  • Romney Win! becomes the headline, supported of course by the faux CNN poll and more liberal whining
  • Two right wing polling outfits announce a Romney bounce the very next day. Timing is everything.
  • The "non-partisan " PEW follows with another poll that shows Romney +4. This lead is almost entirely the result of PEW giving Republicans a +8 advantage in sampling, and as a result, a sample that is also disproportionately older whiter and male.
  • Gallup, as planned, changes from a Registered Voter Screen to a Likely Voter screen, which as is typical, blosters a Republican advantage
  • These sampling issues and methodological questions are buried by MSM headlines and the "horse-race narrative solidifies.
  • The rest herd on in.
The ultimate result is the illusion of massive Mittmentum.

To quote Joe Biden again, "Thats' a Bunch of Stuff"

A closer look at polling outfits, polling internals re sampling screens and demographics indicated that the "debate bump" is mots likely the result of shifting methodologies in combination with an energized GOP base .

Questions to Ask and A Word of Caution

This election season, the media - polling narrative seems more closely linked than ever in my memory. Just me?? Coincidence?? Post-Citizens United Dark Money?? Who knows??

Just pay closer attention..

In the midst of media-driven narratives and the polling chaos, here are some questions to ask about every reported poll:
  • Who is this outfit?? This season seems to have a cavalcade of questionable outfits who are reporting out results using dubious data.. See Gravis Marketing, Exhibit A.
  • Is this a Partisan Pollster? Some pollsters, like Rasmussen, are known to have a +% Republican sampling lean. Like clockwork. Others like the once non-partisan Suffolk, have blatantly and prematurely declared Romney victories in Florida and Virginia.
  • How is Party ID represented? Even respectable pollsters are tending towards over-sampling Republicans relative to their percentage of the electorate.
  • Who is sampled? Check the demographics -- are there representative samples of African Americans, Latino/as, all age groups, Women, various income and educational groups?
  • What is the Polling Cycle? Are these Daily Polls, "Insta-Polls" or Rolling Averages? Are the surveys being conducted over several days ? And Which days? The news cycle matters.
  • Are Registered Voters or Likely Voters Reported? This has a dramatic impact on polls results. There are at least 5+ more Democratic Registered Voters, while Republicans typically dominated Likely Voters screens by +2 or more.
  • Do any of these take the impact of Early Voting into account? Are they removing those who already voted from their polls? If yes, then how does that impact the resulting sample?
This election cycle also raises questions about the quality of various Poll Trackers and Electoral College Predictors. ( See 538 for an excellent recent discussion of poll tracker variations and more) Some trackers use state level data only to estimate electoral college victory -- others combine national and state polls. Take your pick.

Gaps have persisted all cycle between the National Popular vote estimates and the state level polling data. The past week has brought wild and rapid swings that seem driven by data that may fail to capture the reality of the ground.

This leads to more questions:
  • Are the poll tracking models relying on polling samples that under-represent key segments of the Electorate?
  • Do the models presume that "outliers" are randomly distributed in all directions and will cancel themselves out?? What if they don't? What if noise becomes the signal?
  • How effective are predictive models if the input data is foundationally flawed? Is it time for somebody to start more carefully examining the entrails of these polls?
  • Do these models rely too heavily on polls while missing other data re ground games, field offices, differences in targeted state campaigns, the impact of advertizing dollars etc?
  • Do these models inadequately account for other data such economic indicators, Presidential approval ratings and more?
We probably won't know the answers until after November 6, if then.

But we do know this and always have, that cliche is true:

The Only Poll that Matters is the One on Election Day.

So my advice as always..

Turn off the TV. Clear out the Noise.

Ignore the Polls if it makes you saner. It really doesn't matter if Obama is or Romney is up by 20+, we will need to GOTV.

Talk to everyone you know and those you don't. Go to the polls and take 10 with you

Because Everything is at Stake..



Originally Published at Critical Mass Progress