The decision to launch the Wrap this week is in conjunction with the debut of a project that we here at DK Elections are enormously excited about: our Daily Kos Elections Poll Explorer, which launched on Monday.
From now until the first full week in October, we will feature these two data-driven efforts twice a week. You will find the Poll Explorer on Mondays and Thursdays, with the Polling Wrap coming on Tuesdays and Fridays. Once the election season really kicks off in earnest in the final month, you will see us ramp up the coverage here, with both features coming to you three days a week, on alternating days.
For the inaugural edition of the Wrap, we will explore something that came up in the comments of the inaugural Poll Explorer, which is a bit of a glut of Republican sponsored-data in the 2014 data thus far. Some asked if that might bias the polling model we debuted yesterday. Others offered, as has been offered before, that GOP pollsters generally suck.
While both of those things may well be valid points, beyond the jump I will explain why Democrats should also be at least a little bit concerned about that data glut.
But we'll address that in a bit. For now, though, on to the numbers, which is a listing of all the general election polls released to the public in the last five days (Aug 21-25):
AK-Sen (Rasmussen): Dan Sullivan (R) 47, Sen. Mark Begich (D) 45A few thoughts, as always, await you just past the jump ...
AR-Sen (Opinion Research for the AR Dem Party): Sen. Mark Pryor (D) 46, Tom Cotton (R) 41
GA-Sen (Landmark/Rosetta Stone): Michelle Nunn (D) 47, David Perdue (R) 40
MN-Sen (SurveyUSA): Sen. Al Franken (D) 51, Mike McFadden (R) 42
OR-Sen (Moore Information for Richardson): Sen. Jeff Merkley (D) 47, Monica Wehby (R) 38
WV-Sen (Rasmussen): Shelley Moore Capito (R) 50, Natalie Tennant (D) 33
WY-Sen (Rasmussen): Sen. Mike Enzi (R) 63, Charlie Hardy (D) 27
AR-Gov (Opinion Research for the AR Dem Party): Mike Ross (D) 44, Asa Hutchinson (R) 44
CT-Gov (Rasmussen): Tom Foley (R) 45, Gov. Dan Malloy (D) 38
GA-Gov (Landmark/Rosetta Stone): Jason Carter (D) 44, Gov. Nathan Deal (R) 40
MA-Gov (Social Sphere): Martha Coakley (D) 41, Charlie Baker (R) 34
MA-Gov (Social Sphere): Charlie Baker (R) 34, Steve Grossman (D) 32
MA-Gov (Social Sphere): Charlie Baker (R) 40, Don Berwick (D) 22
MD-Gov (OnMessage for Hogan—R): Anthony Brown (D) 45, Larry Hogan (R) 42
MN-Gov (SurveyUSA): Gov. Mark Dayton (D) 49, Jeff Johnson (R) 40
OR-Gov (Moore Information for Richardson): Gov. John Kitzhaber (D) 45, Dennis Richardson (R) 41
OR-Gov (OnMessage for the RGA): Gov. John Kitzhaber (D) 42, Dennis Richardson (R) 38
WY-Gov (Rasmussen): Gov. Matt Mead (R) 55, Pete Gosar (D) 34
IA-01 (The Polling Company for Blum): Pat Murphy (D) 40, Rod Blum (R) 35
WV-02 (Public Opinion Strategies for Mooney): Alex Mooney (R) 40, Nick Casey (D) 28, Others 13
One thing that jumps out in that rather modest cross-section of data is that much of the polling presented here was the work of partisan pollsters, and released by campaigns or other interested parties. That, in itself, is a fairly recent phenomenon, but not a particularly surprising one.
It is often easy for us to forget that political polling is pretty damned expensive. Therefore, we have seen media outlets really curtail the amount of polling they pay for and provide to their viewers and readers. With tightening budgets, these outlets have to be a bit more judicious about how to stretch those dollars. If I had to guess, we'll see more media-sponsored polling in late September and October, when the interest level ramps up a bit. But, for now, released internals are an increasingly common source for numbers.
One other thing we have noticed, represented in this particular data set and throughout the cycle, is that GOP pollsters have been a bit more likely to release data than Democratic pollsters. If you exclude PPP polls that were not for a specific client, you get 111 Democratic-released polls, versus 138 Republican-released polls (non-client polls from folks like Gravis and Vox Populi were also not counted.
Some of our readers reflected on this in the comments for the first edition of our Poll Explorer yesterday. They were concerned that all of that data from Republican pollsters, which seemed awfully bullish on Republican prospects, would bias a poll-based model. Which, in the short run, it certainly might, but let our own Drew Linzer and David Jarman explain why we use them, anyway:
Do we include partisan polls in the model? Yes, we do, if only because so many of the polls in the public record are partisan polls, as fewer media outlets pay to generate their own surveys these days. Unfortunately, that leaves more of the playing field open to leaked results from pollsters or campaigns who may or may not have an agenda. We compensate for this by subtracting 1.5 percentage points from candidates of the party or campaign that commissioned the poll. That 1.5 percent gets added to the opposing major party candidate's vote share, for a total swing of 3 points. The size of this correction is based on our calculation of the typical difference in survey error between partisan and non-partisan polls relative to states' underlying trendlines.I would take it one step further, for the purposes of trying to understand an election. In August of 2012, I addressed this subject in a piece for Sunday Kos. One of the things I noted, which has relevance here, is that in recent history looking at which party is feeling frisky about releasing internal data might tell us a little bit about the forthcoming election. Since 2006, we have only seen more Republican internals released than Democratic ones on a single occasion: 2010.
What about polls released by companies like Rasmussen, that have had partisan "house effects" in past elections? Rasmussen, for example, had a Republican tilt in 2012, and many observers assume they have a pro-Republican agenda. But rather than trying to guess at each firm's 2014 house effect (and run the risk of guessing wrong), our approach is to throw these polls on the pile with every other pollster. If there is an adequate amount of polling data, then there should be enough naturally-occurring error on all sides of the polling spectrum that a smoothed trendline will buff out any one pollster's idiosyncrasies.
Now, is that to say that this year is a 2010 redux. Nope. For one thing, the difference between Democratic-released polls and GOP-released polls is not huge. To put it in percentage terms, Republican affiliated groups have released 55 percent of the "internal polls" thus far, compared to 45 percent of said polls being dropped by Democratic affiliated groups.
For another, no one wins a game in the third quarter, which is roughly where this election cycle resides right now. There are miles to go before this election cycle is in the books, and a lot of polls to digest along the way. And you can count on Daily Kos Elections to help you make sense of them, in more ways than we ever have before.