In a February 25-27, Star Tribune poll that was reported on March 6, “Fifty-three percent of Minnesotans say the state statute banning same-sex unions should stand. Only 38 percent say legislators should overturn the law this year, while 9 percent are undecided.” The is poll has generated significant discussion, with opponents of same-sex marriage contending that its results validate their argument that the rejection of the marriage amendment last fall does not translate into support for same-sex marriage in Minnesota. Supporters of same-sex marriage contend that the poll was flawed, pointing to the language or wording of the questions.
However, there is a different issue here–How accurate was the poll? In the last few years the Star Tribune polls have been way off, such as predicting Dayton to win big over Emmer in 2010. Yet at the same time in 2012 the last Star Tribune poll was about accurate in predicting the Rick Nolan and marriage amendment results. So what should we think about this poll?
The paper reports that the survey included 800 respondents and had a margin of error of +/- 3.5 percent. But additionally the poll also had 80% of respondents coming from landline and 20% cell phone, and the partisan split was Democrats 35%, Republicans 27%, Independents/Other 30%.
Consider first the landline/cell phone issue. The best estimates now are that close to 40% of Minnesotans are cellphone exclusive or primary users. Cellphone exclusive users tend to be younger, landline older, but as cell phone use grows, this trend decreases. Why is this breakdown important? The best evidence is that among those age 30 or younger, over 70% favor same-sex marriages. Thus, if the Star Tribune only had 20% cell phone users–less than half of what the actual usage is in Minnesota–then it potentially under-countered support for same-sex marriage in the state. This means that one could potentially adjust the 53/38 split on the issue by 6 or so points. By my estimate, a survey of 40% cell phone users might change the poll results to 47% wishing to continue to support a ban on same-sex marriage while 44% wishing to repeal the ban. Suddenly the poll looks tighter and within the margin of error.
Yet the partisan breakdown is interesting. It lists 35% DFL, 27% GOP and 30% Independent/other. Compare this to the 2012 exit poll listing 39% DFL, 31% Republican, and 30% Independent/ other. The Star Tribune poll perhaps slightly underpolls Democrats but at 35% I would argue that this is close to where DFL support is in the state. However, at 27%, the poll is lower than the 2012 exit poll and below what I think is an accurate read on Republican identification in the state. I put it at around 31-32%–about where the 2012 exit poll locates it.
The partisan breakdown may perhaps both undercount DFL and GOP voters in the state. With the exception of those under age 30, partisanship is a good predictor support for or against same-sex marriage. It is possible that the poll, were it to adjust for partisanship, could yield even two or three more points either way for maintaining or repealing than ban. Thus, one scenario might mean that up to 47% might support repealing the ban (adding this correct to my other correction), or conversely, we are again back to about 50% supportive of keeping the ban in place. It is just too difficult to make predictions here.
However, there are three bottom line points I want to make here. First, no one single poll tells the story about public opinion on an issue. Second, methodology matters and the mix of cell phone users and partisans can impact the accuracy of the poll. That is why nerds like me read the footnotes first. Third, the Star Tribune poll may not accurately reflect where the public is on same-sex marriage and that it is possible, but not certain, that the real sense of public opinion may be closer than the poll suggests once margins of error, cell phone usage, and partisanship are factored in.