With the Star Tribune and MPR on the sidelines for now, only two major polling organizations have covered Minnesota’s marriage amendment: SurveyUSA and Public Policy Polling. For the first time this election season, both showed “Vote Yes” forces leading.
Bad news for gay-marriage supporters? Yes. But the nature of those leads were very different: SUSA has the anti-gay-marriage amendment claiming a majority even without undecideds, 50-43, while PPP posited a nailbiter, 48-47.
Simply put, PPP showed much more unanimity among Democrats and a far wider gender gap.
Here’s how Democrats responded to each survey:
SUSA: 62-30 No
PPP: 78-16 No
Put another way, Dems opposed the amendment 2:1 in the SUSA poll, and nearly 5:1 in PPP’s.
Gender gap, “Yes” voters:
SUSA: Men 51, Women 48
PPP: Men 55, Women 41
Age also showed differences, but the two pollsters use different intervals, mucking up comparisons:
SUSA, 18-34-year-olds: 50-39 Yes
PPP, 18-29 year-olds: 50-44 No
Gay marriage foes think that PPP’s results seem truer. The general perception is the younger you are, the more likely you’ll oppose the amendment. SUSA actually shows 35-64-year-olds opposing the amendment more than 18-34s.
Both polls include 65+ age group, and their results are pretty close: 56-39 Yes for SUSA, 53-40 Yes for PPP.
What explains the differences?
The polls were conducted within days of each other (Sept. 6-9 for SUSA, Sept. 10-11 for PPP), making that an unlikely source of variation.
Another source we can write off is the question itself: for the first time, both pollsters asked the actual ballot question.
This is a big change for SUSA, which began asking the question for client KSTP-TV before the Legislature settled on language. SUSA stuck to such language as late as this July; I questioned that, and to their credit, SUSA & KSTP changed things.
I think SUSA’s old wording made it easier to answer “yes” to the marriage amendment question. Therefore, I’d caution gay-marriage-watchers from inferring a drop in support from SUSA’s July results (52-37 Yes) to the latest 50-43 support.
The polls do differ methodologically.
Both are “robo-polls” (voters with land lines respond to recorded prompts). However, SUSA also salts in a mobile text survey for those with cell phones, tablets and other electronic devices.
Texters accounted for 23 percent of SUSA’s respondents. They opposed the amendment 44-43, while landliners approved it 52-43.
If you believe young voters are more likely to eschew land lines, and the amendment, those numbers make sense. But remember, SUSA showed the youngest voters supporting the amendment more than their parents.
To further confuse things, the poll that relied purely on land lines (PPP’s), showed more marriage-amendment opposition.
(PPP believes weighting for age replicates cell-phone calling. Legally, robo-pollsters can’t auto-dial cell phones; they have to use more expensive human operators or workarounds like SUSA’s text approach.)
The polls also have small partisan differences: PPP’s Democrat/Republican split was 35/32, while SUSA’s was 37/30. Again, counter-intuitively, the more Democratic poll (SUSA’s) showed more gay marriage support, even though both polls showed Dems opposing the amendment better than 2:1.
Who’ll actually vote?
At this point in the campaign, pollsters want to capture who will actually vote in two months. So the “voters screen” can be a major source of difference.
SUSA uses a single question: are you certain, likely or unlikely to vote? They only use those who answer “certain.”
PPP asks people if they voted in 2006, 2008 or 2010. If respondents say “yes” to any of those years, they’re in. However, unlike PPP’s June survey, respondents are first told that if they don’t plan to vote this year, they should hang up.
I don’t know if that explains a rise in support from that June verdict, which showed opponents winning 49-43. So much time has passed it’s possible sentiment has changed, even if SUSA’s numbers are “moving” in the opposite direction.
It’s also important to note that normal statistical variation — the act of placing your statistical dipper into the vast ocean of voters — can produce differences, and that such variation is greater for subgroups such as gender, age, etc. compared above.