The median age of a voter in CA-52, according to the new SurveyUSA poll
As one might expect, the emotions of the election cycle get more tense as the calendar counts down closer to Election Day. That was never more evidence than this week, when a mid-week wave of rather pessimistic polls for the Democrats in a variety of races led to teeth-gnashing wailing among many in the blue-tinted commentariat, as well as more than a fair share of stomach-turning smugness from the right.
But the taunt that caught my attention the most, because it was so incredibly and absurdly errant, was the invocation of one of the dominant election-nerd stories of the 2012 cycle.
Whether the issue was the spate of Fox News polls out of all the Senate swing states, or some particularly interesting underlying details in the SurveyUSA poll out of San Diego County's swing district (CA-52), the second that a Democratic observer found a curious characteristic in the data, the accusation rained down from some cackling right-winger in the Twittersphere:
"Oh, look at the Dems trying to unskew the poll! Losers!"
It was then that I decided that a tutorial might be in order. Because, obviously, in the 23-month interim, we have forgotten what "unskewing" was, and why it was such an absurd little McNugget in the whole 2012 electoral experience.
So we will take a trip down memory lane, and we will look at what is, and what is not, "unskewing."
But, first, as we draw closer to the big day, the volume of polling continues to get downright nasty. In just three days (Oct 7-9), we added another 69 polls to the database. So check those out first, and then let's take a deeper dive into what is, and what is not, acceptable polling critique.
AK-Sen (CNN/ORC): Dan Sullivan (R) 50, Sen. Mark Begich (D) 44
AK-Sen (Fox News): Dan Sullivan (R) 44, Sen. Mark Begich (D) 40
AR-Sen (Fox News): Tom Cotton (R) 46, Sen. Mark Pryor (D) 39
AR-Sen (Opinion Research Associates): Sen. Mark Pryor (D) 45, Tom Cotton (R) 42
CO-Sen (Fox News): Cory Gardner (R) 43, Sen. Mark Udall (D) 37
GA-Sen (PPP--D): David Perdue (R) 45, Michelle Nunn (D) 43, Amanda Swafford (L) 5
GA-Sen (SurveyUSA): David Perdue (R) 46, Michelle Nunn (D) 45, Amanda Swafford (L) 3
IL-Sen (We Ask America--R): Sen. Dick Durbin (D) 51, Jim Oberweis (R) 38, Sharon Hanson (L) 4
IA-Sen (The Morey Group--R): Bruce Braley (D) 39, Joni Ernst (R) 38
KS-Sen (CNN/ORC): Sen. Pat Roberts (R) 49, Greg Orman (I) 48
KS-Sen (Fox News): Sen. Pat Roberts (R) 44, Greg Orman (I) 39
KS-Sen (SurveyUSA): Greg Orman (I) 47, Sen. Pat Roberts (R) 42
KY-Sen (Fox News): Sen. Mitch McConnell (R) 45, Alison Lundergan Grimes (D) 41
MA-Sen (MassINC): Sen. Ed Markey (D) 56, Brian Herr (R) 30
MI-Sen (Glengariff Group): Gary Peters (D) 44, Terri Land (R) 35
MI-Sen (Marketing Resource Group--R): Gary Peters (D) 47, Terri Land (R) 36
MI-Sen (Wenzel--R): Gary Peters (D) 47, Terri Land (R) 44
MN-Sen (Public Opinion Strategies--R): Sen. Al Franken (D) 46, Mike McFadden (R) 39
NC-Sen (The Morey Group--R): Sen. Kay Hagan (D) 40, Thom Tillis (R) 38, Sean Haugh (L) 2
NC-Sen (Public Opinion Strategies--R): Sen. Kay Hagan (D) 44, Thom Tillis (R) 42, Sean Haugh (L) 4
NC-Sen (Rasmussen): Sen. Kay Hagan (D) 48, Thom Tillis (R) 46
NC-Sen (Suffolk): Sen. Kay Hagan (D) 47, Thom Tillis (R) 45, Sean Haugh (L) 4
NH-Sen (New England College): Sen. Jeanne Shaheen (D) 49, Scott Brown (R) 46
NH-Sen (Univ. of New Hampshire): Sen. Jeanne Shaheen (D) 47, Scott Brown (R) 41
NJ-Sen (Stockton Polling Institute): Sen. Cory Booker (D) 48, Jeff Bell (R) 39
SD-Sen (SurveyUSA): Mike Rounds (R) 35, Larry Pressler (I) 32, Rick Weiland (D) 28
VA-Sen (Christopher Newport Univ.): Sen. Mark Warner (D) 51, Ed Gillespie (R) 39
VA-Sen (Univ. of Mary Washington): Sen. Mark Warner (D) 47, Ed Gillespie (R) 37, Robert Sarvis (L) 6
AK-Gov (Fox News): Gov. Sean Parnell (R) 42, Bill Walker (I) 37
AR-Gov (Fox News): Asa Hutchinson (R) 46, Mike Ross (D) 37
AR-Gov (Opinion Research Associates--D): Mike Ross (D) 45, Asa Hutchinson (R) 41
CO-Gov (Fox News): Gov. John Hickenlooper (D) 42, Bob Beauprez (R) 42
CT-Gov (Quinnipiac): Gov. Dan Malloy (D) 43, Tom Foley (R) 43, Joe Visconti (I) 9
FL-Gov (Øptimus--R): Charlie Crist (D) 40, Gov. Rick Scott (R) 39, Adrian Wyllie (L) 13
FL-Gov (PPP--D): Charlie Crist (D) 45, Gov. Rick Scott (R) 43, Adrian Wyllie (L) 8
FL-Gov (SurveyUSA): Charlie Crist (D) 44, Gov. Rick Scott (R) 42, Adrian Wyllie (L) 6
FL-Gov (Univ. of North Florida): Charlie Crist (D) 43, Gov. Rick Scott (R) 38, Adrian Wyllie (L) 10
GA-Gov (PPP--D): Gov. Nathan Deal (R) 46, Jason Carter (D) 41, Andrew Hunt (L) 4
GA-Gov (SurveyUSA): Gov. Nathan Deal (R) 46, Jason Carter (D) 44, Andrew Hunt (L) 4
IA-Gov (Loras College): Gov. Terry Branstad (R) 56, Jack Hatch (D) 33
IL-Gov (We Ask America--R): Gov. Pat Quinn (D) 44, Bruce Rauner (R) 40
KS-Gov (CNN/ORC): Gov. Sam Brownback (R) 49, Paul Davis (D) 49
KS-Gov (Fox News): Gov. Sam Brownback (R) 46, Paul Davis (D) 40
KS-Gov (SurveyUSA): Paul Davis (D) 47, Gov. Sam Brownback (R) 42
MA-Gov (Emerson College Polling Society): Charlie Baker (R) 45, Martha Coakley (D) 40, Others 8
MA-Gov (MassINC): Martha Coakley (D) 41, Charlie Baker (R) 39, Others 5
MA-Gov (SocialSphere/Boston Globe): Martha Coakley (D) 39, Charlie Baker (R) 34, Others 7
ME-Gov (Greenberg Quinlan Rosner--D): Michael Michaud (D) 43, Gov. Paul LePage (R) 39, Eliot Cutler (I) 15
ME-Gov (Pan Atlantic/SMS Group): Gov. Paul LePage (R) 39, Michael Michaud (D) 34, Eliot Cutler (I) 20
MI-Gov (Marketing Resource Group--R): Gov. Rick Snyder (R) 46, Mark Schauer (D) 41
MI-Gov (Glengariff Group): Gov. Rick Snyder (R) 45, Mark Schauer (D) 37
NH-Gov (New England College): Gov. Maggie Hassan (D) 51, Walt Havenstein (R) 41
NY-Gov (Quinnipiac): Gov. Andrew Cuomo (D) 51, Rob Astorino (R) 31, Howie Hawkins (G) 9
PA-Gov (Quinnipiac): Tom Wolf (D) 55, Gov. Tom Corbett (R) 37
SD-Gov (SurveyUSA): Gov. Dennis Daugaard (R) 59, Susan Wismer (D) 30, Michael Myers (I) 7
CA-52 (SurveyUSA): Carl DeMaio (R) 48, Rep. Scott Peters (D) 45
IA-01 (The Polling Company--R): Rod Blum (R) 40, Pat Murphy (D) 39
IA-03 (Greenberg Quinlan Rosner--D): Staci Appel (D) 49, David Young (R) 42
KS-02 (Anzalone Liszt Grove--D): Rep. Lynn Jenkins (R) 48, Margie Wakefield (D) 43
MA-06 (Garin Hart Yang--D): Seth Moulton (D) 43, Richard Tisei (R) 33, Chris Stockwell (I) 11
MA-09 (Emerson College Polling Society): John Chapman (R) 45, Rep. Bill Keating (D) 40
NH-01 (New England College): Rep. Carol Shea-Porter (D) 47, Frank Guinta (R) 44
NH-01 (Univ. of New Hampshire): Rep. Carol Shea-Porter (D) 42, Frank Guinta (R) 39
NH-02 (New England College): Rep. Anne Kuster (D) 50, Marilinda Garcia (R) 38
NH-02 (Univ. of New Hampshire): Marilinda Garcia (R) 41, Rep. Anne Kuster (D) 37
SD-AL (SurveyUSA): Rep. Kristi Noem (R) 55, Corinna Robinson (D) 37
UT-04 (FM3--D): Mia Love (R) 47, Doug Owens (D) 44
UT-04 (Y2 Analytics--R): Mia Love (R) 47, Doug Owens (D) 28
In this set of data, there is a SurveyUSA poll for the
San Diego Union Tribune and a local television outlet (KGTV) with a fairly benign topline result, but some absolutely batshit crazy details once you go under the hood.
That poll was the survey of California's 52nd district, where Democrat Scott Peters is seeking his second term in a (largely) suburban swath of San Diego. The poll had Republican challenger Carl DeMaio, who narrowly lost a mayoral election there in 2012, with a three-point edge in the race. While that is a little rosier than the previous SUSA polling in CA-52, it's not as much of a "WTF?" poll as ... say ...a college pollster making the case that a race in Massachusetts on no one's radar screen is somehow going to flip to the Republicans by 5 points.
But when going through a deeper dive of the numbers, several things stood out. For one thing, SUSA had Peters losing, but winning white voters. On the surface, that seems a little out there: if true, it would seem as if Peters would be the only Democrat in America who could carry white voters while losing an election.
But where some real questions arise are in sample composition. This was, without question, a fairly old sample. Only 17 percent of the sample were under the age of 50. By way of comparison, in 2010, 35 percent of the electorate statewide were under the age of 44 statewide. A cursory glance at the district's data tells us that there are actually more residents 18-44 in the district than there are residents 45 or older. Of course, population does not equal the electorate, and we anticipate an older electorate at all times, and especially in lightly attended midterm elections.
But is it really possible that a 55/45 split between "younger residents" and "older residents" would really become a 17/83 spread between "younger voters" and "older voters"?
My answer to that question is a phrase I have used many times in poll analysis over the last several years here at Daily Kos Elections: it is possible, but that sure doesn't make it likely.
The bottom line is that queries about the racial voting patterns and the age of the sample are, indeed, legitimate questions.
What they are not is "unskewing".
You can read my exhaustive takedown on "unskewing" (written almost exactly two years ago) right here). To simplify: the whole phenomenon flogged by the GOP in 2012 was based on the errant assumption that virtually all of the polls were wrong. Whats more, the assumption was that they were wrong for one simple reason: they were assuming that "too many" Democrats were going to show up at the polls.
In a previous edition of the Polling Wrap, we looked at how making assumptions about the partisan identification of the electorate can be a dangerous path upon which to question a poll. Arguably, the only time it becomes worth discussing is when a polling sample has a partisan composition that has not been seen in any recent cycle.
For example, if a California poll shows Neel Kashkari in striking distance of California's Democratic Gov. Jerry Brown, and shows a sample where there are more Republicans than Democrats, that is a poll that it would probably be wise to dismiss (Democrats, even in lousy 2010, enjoyed a double-digit partisan ID edge).
But that's not what those questioning the CA-52 poll were doing. They were questioning things that are not as fluid. Racially, it would take quite the research effort to find a poll where a Democrat carried white voters and trailed in the trial heat. In other words, the skeptics of the SUSA poll in San Diego were merely pointing out that SUSA was seeing in their poll a racial vote pattern that has simply not happened in recent electoral history. Not just in San Diego, but anywhere. That alone strains credulity, on the surface, as did the outsized number of older voters. These are legitimate questions.
Poll skeptics should keep in mind, though, that when you dig into the guts of a poll, you are often dealing with very small numbers. Consider: if 16 percent of the sample in CA-52 is nonwhite, that is a sample of around 85 voters or so. A quick look at a margin of error calculator puts the MoE for nonwhites in the poll at just over 10 percent. That's basically no better than a kiddie pool of data, and a wise observer should be careful not to draw too many conclusions.
Also, wise critics of polls should realize, as Republicans failed to do in 2012 (and Democrats in 2010, for that matter), that one can find a sample composition quirk, or curious demographic performance, in every single poll. But, if you are questioning every pessimistic poll that comes out?
At the risk of sounding redundant: it is possible that you are right, and every poll that does not favor your side is errant.
But that sure doesn't make it likely.