The May Poll released by Marquette University and Charles Franklin has received attention from both sides of the aisle. Scott Walker and the Republican Party are touting the 6 point lead as proof that voters know “it’s working.” Tom Barrett and the Democratic Party are responding by attacking the poll, and the pollster himself, alleging intentional conservative bias. The truth is neither of the above. As a poll is a snapshot in time, the May poll, when observed in the context of the previous four months (January – April polls) give us a clear picture of how close this race really is – and what Barrett needs to do to win.
First, a word on the methodology of the Franklin polls. Accusations of intentionally biased polls on the part of Charles Franklin are unfounded. Badger Democracy has interviewed at length two independent experts in the area of polling and statistics (one at UW-Madison, one at The University of Illinois-Chicago). Both agreed that the polling methods used at Marquette are sound and unbiased – with full disclosure of the results and any bias through random sampling. So let’s take this poll as it stands, and look at what it means.
One issue the polling raises is how the population distribution is calculated and “post-stratified” (or weighted to reflect the population of the state based on regional populations). As seen in the May Poll Methodology, the population distribution is adjusted to reflect census data. As is the case in January Poll, February Poll,March Poll, and April – the end result is very close to the real population distribution. While this has been consistent for all the Marquette Polls, the potential issue is how the population distribution differs from the actual voter distribution.
According to Government Accountability Board 2010 General Election statistics, the difference is significant. In the 2010 Fall election; Milwaukee accounted for 16% of the statewide vote, Suburban Milwaukee 12%, Dane County 10%, Green Bay/Appleton 7%, with the rest of the state 55%. The population models result in suburban Milwaukee (Waukesha, Ozaukee, Washington counties) being over weighted by about 15%, and the statewide vote being under weighted by as much as 30%. The upstate, non-urban vote is largely under-represented. The highly conservative collar counties of Milwaukee are over-represented. That is critical to those who are involved in campaign strategy.
The current May poll also results in a 5% higher response from Republican identifying voters, with 5% lower independents. This has not been the case in every Marquette poll. January leaned 4% GOP, February was almost even GOP/Dem, March leaned slightly Dem (3%), April about 3% GOP (the difference between 5% higher GOP and 2% Dem, lower Independent). In general, it is fair to say the poll, over the past 5 months, has leaned slightly GOP, with May being an anomaly – for several reasons.
While every other poll from January-April have adjusted for age, the May poll did not. The age brackets from 18-29 and 30-44 were under represented (based on census data) by 8 and 11% respectively. The age range of 60+ was over represented by 16%. The reason – the under 30 age group broke significantly higher not only for Scott Walker (55 Walker – 44 Barrett), but for Mitt Romney (51 Romney – 44 Obama). This is nearly a 20 point flip from all other polls, and as Professor Franklin acknowledged in an email to Badger Democracy, “a random fluke.” This, coupled with the over 60 (generally GOP voting) age group being over represented, shows an anomaly in the random sample – this group was just overly conservative.
In fact, the picture of the respondent that influenced this vote can be seen by the fact that married respondents were over weighted by 6%, while all single groups were under weighted by about 20%. The “voter” that influenced this poll was most likely the younger, married, white male that lives in a collar county of Milwaukee – leaning Republican.
Polls give us a picture in time of a group of voters. In random sampling, just as flipping a coin, you can have a period of time when you get five or six “heads” in a row. If you flip the coin enough, the result will “regress to the mean” of about 50%. The May poll is an example of that reality. Taking all five polls into account, and the trend, the picture is clear.
This will be a very close election, decided by two points or less. It is all about turnout. Walker and the GOP are relying on big money and a huge propaganda machine – walking the line of legality. Dems and progressives are relying on grassroots, boots on the ground. If progressives achieve our biggest turnout ever, Walker loses. That is the bottom line.
The proof? The final question in the poll shows where this GOP – leaning demographic has gone in two years:
Did you vote for the Republican Scott Walker or for the Democrat Tom Barrett in November
2010 or did you vote for someone else?
Scott Walker 301 52%
Tom Barrett 238 41%
Within the margin of error for this poll of 3.8% (which resulted in a 50-44 lead for Walker). The respondents in this poll did not change their vote in the past two years. The path to victory gets out every vote for every person that HAS changed their mind. The cumulative results of all five polls show that a slightly GOP leaning poll gives the slim victory to Walker. A slightly Democratic leaning poll gives the slim victory to Tom Barrett.
The path to victory is clear. Get out the vote – every union household, rural, minority, progressive, women, teachers, and democratic voter in the state must get to the polls on June 5th. If that is accomplished – Scott Walker will not be Governor on June 6th.