Wednesday, October 1, 2014

Walker wins! If the voters are old white guys

Think turnout won’t be a big part of this November's election for governor in Wisconsin? Take a look at 3 different polls that have been released in the last 2 days. at the “Likely voter” sample of the Marquette Law Poll, the “registered voter” sample of the Marquette Law Poll, and the “registered voter” sample of a poll by Gravis Marketing. All of these polls were taken in the last week, but look at how minor changes in the sample make for different outcomes.

First, let’s look at the toplines in the Burke-Walker Governor’s race:

Marquette Likely- Walker 50.5%- Burke 44.7%
Marquette Reg.- Walker 46.4%- Burke 45.1%
Gravis- Burke 50- Walker 45%

Quite a wide disparity in itself- No poll is within 4.5% of each other. So what’s going on here? Going inside the numbers will tell the story.

Party ID (with Leaners in MU Poll)-
Marquette Likely- Republican +4.6%
Marquette Reg.- Democrat +0.4%
Gravis- Democrat +1%

Interesting- let’s check out some other demos.

Marquette Likely- 18-49 41.4%, 50-64 31.2% 65+ 27.4%
Marquette Reg.- 18-49 46.7%, 50-64 28.5%, 65+ 24.8%
Gravis 18-49 48%, 50-64 31%, 65+ 20%

Race (of those giving an answer)
Marquette Likely- White 90.5%, African-American 4.5%, Hispanic/Latino 3.0%
Marquette Reg- White 89.2%, African-American 5.8%, Hispanic/Latino 4.8%
Gravis- White 84.9%, African-American 6.5%, Hispanic/Latino 3.2%

As an interesting addition to this- UW-Madison just came out with a report today on the rapidly increasing Latino population throughout Wisconsin, so go with the Marquette registered voter model’s larger Hispanic numbers on this one.

Education level
Marquette Likely- Bachelor’s+ 37.9%, AA/Some College 32.7%, HS diploma or less 28.7%
Marquette Reg- Bachelor’s+ 32.1%, AA/Some College 32.4%, HS diploma or less 34.4%
Gravis- Bachelor’s+ 54%, AA/Some College 30%, HS diploma or less 17%

There’s a definite link between the difference in demos in this survey, and the outcome. The Marquette Law “likely voter” list has the oldest, whitest most Republican electorate, and Walker does best in that one. The Gravis sample has a younger, more racially diverse Dem electorate, and Burke is winning in that one. The registered voter one is mostly in between, with a slight edge to Walker’s electorate, and it’s basically even.

The 5% difference in party ID explains a lot of the 4.5% difference in Scott Walker's lead between the "likely" and "registered" polls from Marquette. What's interesting is if you go inside the cross-tabs of the Marquette Poll you’ll see the difference in poll sample bear out between "registered" and "likely" voters across numerous categories. I’m also going to add gender, as the responses are interestingly different, even though the percentage of respondents is mostly the same in all three polls (51-53% female)

Age 18-44 Walker 47.5-46.2 LV, Burke 47.5-40.7 RV
Age 45+ Walker 52.1-44.2 LV, Walker 50.2-43.8 RV
Male- Walker 61.9-33.9 LV (!), Walker 54.2-39.1 RV
Female- Burke 54.5-40.2 LV, Burke 50.4-39.5 RV
Bachelor’s+ Burke 48.3-44.9 LV, Burke 47.0-44.4 RV
AA/Some College- Walker 54.3-41.1 LV, Walker 50.9-41.0 RV
HS diploma or less- Walker 52.9-45.2 LV, Burke 48.1-43.9 RV
White- Walker 52.9-42.6 LV,Walker 49.9-41.9 RV
African-American- Burke 91.3-0 LV (!), 91.3-0 RV (!)
Hispanic/Latino- Walker 55.0-42.3 LV, Walker 48.4-41.0 RV(?)
Other- Burke 60.7-39.3 LV, Burke 72.4-27.6 RV

Look at the change you get from “likely” voters to “registered” voters. Walker loses support across the board except among women, and those with bachelor’s degrees or above (and he still loses both demographics). These crosstabs and differences in sample make a clear point. If the electorate in November is overwhelmingly older, white, male, and with more Republicans coming to the polls, then Scott Walker will probably win. If it’s younger, more diverse, with more women, and with Dems turning out, then Burke will likely win.

Which means TURNOUT! TURNOUT! TURNOUT! The GOPs have been cranking up their fauxtrage machine in recent weeks, and it is reflected in their advantage in the Marquette Poll’s (too tight?) likely voter screen. The Marquette Poll was taken at a relative low point in Burke’s campaign, with Dan Bice and the GOP-aganda machine pumping the “copied job plan” story. The question is, can the Dems respond in kind, and get their voters fired up to go to the polls (and what the hell are they waiting for)? With Burke’s strong ad that just came out answering that non-story, the rising anger over voter ID from Dem-leaning groups, and Walker's campaign contributors allies at WMC denigrating the idea of equal pay for women, I’m thinking that “enthusiasm gap” is gonna drop real quickly- if it ever really existed in the first place.

With that in mind, it makes sense to me to use the “registered voters” numbers from the Marquette and Gravis polls if you want any idea where things stand, because I’d bet that’s going to be more like the group casting ballots in November than the “GOP +4.6” that was in the Likely voter Marquette poll. This seems to be anyone’s game with a month to go.


  1. Just for kicks, I went and took the exit poll stats from 2010 and used Marquette's registered voter results to see what would happen if 2014 turnout was exactly like it in that Republican wave year when Walker won by 5.8 points:

    * If the age distribution were just like 2010's, Walker will win by 1.2 points (I had to jam the highest bracket into a singular 45+ one to maintain equivalency - trying to interpolate four categories makes it a Walker win by 1.8).
    * If the ideological distribution (conservative/moderate/liberal) turnout is just like 2010's, Burke will win by 5.2 points.
    * If the D/I/R party id distribution is just like 2010's, Burke will win by 0.8 points.

    As big a Republican wave as 2010? Nationally, that was an R+7 year, while the parties are generally polling equally this time out. In Wisconsin, 2010 was a D+1 year, other pollsters are putting this year at D+2 to D+5 but Marquette's (tighter than most) likely voter screen suggests R+1.5 (w/o leaners). It's a very big outlier on that point. More Republican than the recall when 150,000 of Barrett's 2010 voters stayed at home? That is hard to credit.

    But it does mean that Team Burke has to aim to turn out all those supporters who aren't already 'absolutely certain' to vote, the 'very likely' and '50-50' voters.

  2. Beware any Gravis poll results. They seem to have been exposed as a scam operation (fake numbers). See details here--

  3. I agree that they aren't very reliable at Gravis (to be kind), but they usually lie for Republicans, which made me wonder why those numbers were released.

    The point is that these numbers are all over the board, and change based on turnout. The difference between the Marquette Registered and Likely voter polls speaks volumes.