Wednesday, September 17, 2014

Turnout matters- in the Marquette Law School Poll

My immediate reaction to seeing today’s poll from Marquette Law School was to give a giant “WTF?” The Walker campaign had been acting increasingly desperate in recent weeks as round after round of bad news hit almost daily, but this poll says Scotty was doing BETTER among likely voters than he was a month ago? On its face, it made no sense, but given that you’re dealing with small samples and the fact that Walker led by 3 points among registered voters in the previous poll, maybe there was something to it.

Then I noted a comment by Marquette Law School’s Charles Franklin, and it set off my BS detectors

That comment about "increased enthusiasm" made me want to look inside the numbers at the crosstabs, to see what he meant by that. And when I did, the answers behind Walker’s “improvement” became obvious. The poll sample added a ton of Republicans to the mix.

Take a look at the question of “Are you a Democrat, Republican, or Independent?” in the September Marquette Poll, and then compare it with the answers to the same question in August.

Dem, Republican, or Indy? Reg. voters
Aug 2014 Marq. Law Poll- 30.9 D, 27.0 R, 37.9 I
Sept. 2014 Marq. Law Poll- 27.9 D, 28.8 R, 40.7 I

So that sample went from D +3.9 to R +0.9. Yet at the same time, Burke IMPROVED by 3 points among registered voters, from -3 to a tie. The "likely voter" poll was even more GOP-leaning, at R +3.7 vs D +6 in August- no wonder why Burke lost 5 points. So there's a lot of your change right there- it's not Walker gaining support. And to pollster Charles Franklin's credit, he admits this party ID stat is a huge factor.
“It is unusual to see a 5-point net shift in partisan composition,” Franklin said. “People should be appropriately skeptical since it is always possible this sample is simply an outlier. However, the shift to more Republicans and fewer Democrats occurred across all regions of the state and most demographic groups, demonstrating that it was not a localized difference in response rates.”

In fact, if anything, Walker is losing some of his core backing. Previously, polls had had the Dem and GOP vote going around 90-5 for each party’s candidate, and Walker slightly favored among Independents (because a sizable amount of them are TeaBaggers). Well, there was a slight change in this breakdown for the September poll.

Sept. 2014 Marq. Law Poll by party
Democrats- Burke 93.2-3.4
Republicans- Walker 88.8-8.0
“Indys”- Walker 44.1-43.0

Scotty then got about an extra 1% from a scattering group, so that’s where you get the 46-46 Registered Voters result from. It shows a bit of erosion from Scotty’s own GOP. I think that's a result of some of them seeing the light, and realizing a crony capitalist that blows a gigantic hole in Wisconsin’s budget is no old-school, good government conservative.

Now with this in mind, let’s use those same Dem-GOP-Indy responses, and plug it into the August sample of registered voters in the Marquette Poll. The result? Burke leads by 4.6 points, 47.3- 42.7.

I understand that sometimes party IDs do switch and reflect people latching onto certain “teams” as they get more fired up for an election. It also reflects who is more or less likely to vote in a certain election (the voters more likely to drop out at a midterm are generally young, single, and/or minority- all three of which lean Dem). But let’s take a look at the gov exit polls from 2010 and the recall election of 2012 along with the 2012 presidential exit poll , and see what’s more likely to be the electorate in Wisconsin in November 2014.

Party ID, Wisconsin exit polls
2010 Gov- Dem 37, GOP 36 (D+1)
2012 recall- Dem 34, GOP 35 (R+1)
2012 presidential- Dem 37, GOP 32 (D+5)

And casual Dems are likely more fired up in 2014 than they were in 2010, or even 2012, as some wussed out on the idea of a recall (how’d that idea work out for you?). At the same time, the September Marquette Poll shows some GOPs probably aren’t as likely to “Stand with Walker.” So a D+4 electorate (as was done in the August Marquette Poll) seems to be a more likely outcome in November than an R+1 electorate (which was in the September poll). which would give the advantage to Burke.

That being said, maybe the Marquette Law Poll is closer to right this time and was off the last. Their record in 2012 was pretty good when their final polls came out for the recall and presidential election. And I’m also aware that this sounds pretty close to the “unskewing” phenomenom that GOPs fell victim to in 2012, where those dingbats beLIEved that Romney was going to win, all the way to Election Night (which made their crash all the more hilarious).

But these figures are striking, and help explain why Walker and his supporters are trying to so hard to suppress the vote with the increasingly insane ramifications from an all-GOP Appeals Court panel’s decision to reinstate voter ID for the November election. Walker and WisGOP can’t win a statewide election unless the Democratic electorate is shrunken or discouraged to the point that it’s near even on Election Day. If it’s a strong Dem turnout- Walker’s done.

Which is what makes it all the more important to educate and assist Wisconsinites with getting that ID (at least until the ruling is tossed due to not being workable), and in making them know that it’s the Republicans that don’t want them to vote. And that anger and outreach effort is why the voter ID decision could prove not to be a boon to Republicans (like the suburban slimeballs think it will be), but instead could be one of the final blows to their chances in the Fall.

So instead of despair, today’s Marquette Law Poll tells me IF WE GET OUT TO VOTE IN NOVEMBER,WE WILL WIN IN WISCONSIN. Period. So make it happen!

1 comment:

  1. Yep, so on the basis of the September likely voter sample, if we get an R+1 turnout like in the recall then Walker wins by 1.3 points; if we get a D+1 turnout like in 2010 then Burke wins by 0.5 points.

    Including the September poll, Marquette has averaged D+4.2 amongst RV's in their six polls this year.

    You're right to point out that people do change their party ID's over time. This is an issue in polling circles: clearly party ID is far and away the best predictor of how someone will vote, but how stable is it? If it can be assumed to be reasonably stable, then you can weight by it: Rasmussen did that and botched the 2012 election. MLSP weight by various census-driven demographics and GAB registration data alone.

    I made a prediction 12 months ago that on the basis of recall exit polls, migration and demographic statistics, Tom Barrett would beat Walker by 2 points in November 2014.

    This could be in part a reflection of national movement towards the GOP in the generic ballot, but Sam Wang's chart there also points to the most conceivable black swan event: Ted Cruz & Co. shutting down the government over Obamacare again just in time for the election (if that happens, then the State Assembly comes into play).

    This Marquette poll, at the end of the day, tells us the same thing that the last three have: it's a statistical dead heat of a race, and anyone on either side of the political divide who doesn't GOTV their hearts out has a good chance of regretting it.