Nate Silver on his blog FiveThirtyEight examined the avalanche of polls conducted for the 2010 US midterm election and came to the somewhat shocking conclusion that certain pollsters produced biased results that favored the GOP. Specifically, he fingered the Rasmussen group. Their polls (Fox, Pulse Opinion Research/Rasmussen) produced an average bias of almost +4% for the GOP. Of the 105 polls conducted by Rasmussen in the three weeks prior to the midterm, 55 overestimated the popularity of the Republican candidate by at least 3%. Only 12 of their polls overestimated the popularity of the Democratic candidate by the same 3%. While to most people evidence of such an imbalance would suggest there was a serious malfunction in the Rasmussen measurement tool, their leader Greg Rasmussen when asked for an explanation, replied that he “can’t imagine any need to respond”.
A bias of +4% is a big deal, particularly in elections where the popularity of the leading candidates is evenly split among potential voters. A first-time candidate favored by this bias is given added legitimacy in the perception of the voting public and increases his or her chance of winning the election. That almost happened in the Nevada senatorial election between Harry Reid and Sharron Angle. With the Rasmussen poll showing Ms. Angle ahead of Mr. Reid by 3% just three days before the election, the media was predicting a major upset for the Tea Party candidate. Unfortunately for her, Mr. Reid won by 5%. As to what may have happened, please see my previous post.
Was Mr. Reid almost a victim of biased Rasmussen polls? Quite possibly. As suggested in a story by Las Vegas Sun reporter J. Patrick Coolican “A bad poll can hurt a campaign, creating a narrative for defeat in the media, deflating donors and volunteers.” The story quotes Mr. Reid’s internal pollster Mark Mellman as saying “The problem is that these things can have an impact. …So the media has a special responsibility to get it right. If you’re not going to get it right then don’t bother.” But how are journalists to decide who is right? Internal poll results are revealed at the discretion of the client. While Mr. Mellman claims his internal polls were much closer to election day results, he did not reveal his data publicly at the time of the Rasmussen polls which unfortunately undermines their credibility.
We all know from Research Methods 101 that polls are supposed to measure public opinion, not create it. Well, so much for that theory. In the rude world we live in, polls are as often used for the purpose of influencing us as much as for informing us. This brings us back to our question: Are Rasmussen polls simply a GOP marketing initiative to make their candidates (often unknown, first-time Tea Baggers) more legitimate and attractive to the voting public? The evidence provided by Nate Silver’s analysis is compelling. But as we learn again and again, the real world has more twists and turns than the human imagination can conjure.
Recently, CNN reported that its exit polls from the midterm election showed that the turnout of conservative voters was 42%, significantly higher than the usual 31-34%. At the same time the Independent turnout was only 37% (versus 53% in 2008) and the Democrat turnout was 35% (the lowest since 1986). If we make the reasonable assumption that the increase in conservative support was influenced by the Tea Baggers, this could explain why many polls produced inflated statistics for Tea Bag candidates, including Sharron Angle in Nevada. As suggested in my earlier post, the increased size of this group as well as their heightened enthusiasm may have resulted in their overrepresentation in poll samples and thereby overstated the popularity of GOP candidates in some midterm elections.
While this demographic explanation doesn’t excuse Greg Rasmussen’s imperious attitude towards questions about his methodology, it does offer a reasonable explanation for the GOP bias noted in Nate Silver’s analysis.
That said, the Rasmussen results look very, very suspicious. It’s not just the numbers. The pollster Greg Rasmussen is by his own admission a conservative. Historically, Rasmussen polls have a right-wing slant to their results. The GOP needs to be very careful about how they use public opinion polling and the perception it creates in the public mind. They already have been labeled as The Party of No. If further analysis finds that partisan politics has played a role in influencing poll results, the GOP will have a new and more damaging moniker– The Party of Lies.
Don’t do it boys. There’s much more to lose than to gain in this bargain, not just for the political fortunes of the GOP but for all democratic discourse in American society.
