Let’s Talk About Polls
September 20, 2012 2 Comments
There have been any number of Republicans who have complained about the media polls which show Obama up in this election. Pew Research shows Obama up 8, NBC shows him up 5. Any number of polls have shown Obama up 5-10 over the summer. Now there’s a web site up and running called UnSkewedPolls.com which purports to unskew the polls. Not surprisingly these polls show Romney up 7.8%. This website takes the raw data from all of the media polls, which is generally available, and weights it in a way more favorable to Romney.
Keep in mind very few polls actually show the raw data in the results. Each poll weights data based on how they believe turnout will be on election day. Thus if they sample 100 Democrats which represent 20% of the people polled but the pollster believes 30% of the electorate will be Democrat, they weight the result of those 100 Democrats up so that each of them represent more than 1 vote in the final poll. Polling companies also weight their sample down if they poll to many Democrats, for example. This is common practice and generally isn’t something to be concerned about. Unless of course you consider some of the outlandish predictions the pollsters have concerning the electorate on election day.
It’s important to look at the history of elections over the Obama years. In 2008, a landslide victory for the Democrats, the electorate was 36% Democrat, 29% Republican and 35% independent. Obama won by 7 points in an election that was D+7. In 2010 the D/R/I turnout was 35/35/30. We all know that 2010 was a landslide victory for the Republicans. It’s hard to imagine either party winning in a landslide in this election. With that in mind the D+7 of 2008 is unlikely but it would represent the ceiling for the Democrats. Likewise it’s unlikely that we’ll see a repeat of 2010 when party turnout was even. So even represents the ceiling for the GOP. Republicans tend to be more enthusiastic about voting this year, so we’re more likely to see a turnout of D+2 or D+3.
That brings us to the media polls and why they’re so outrageous. The average media poll weights the results at D+9. Some have gone as high as D+13. Others show the party split closer but under weight independents, who tend to favor Romney in nearly all the polls. The recent NBC poll for example grossly under samples independents, who only comprise 16% of the poll. They weight Dems at 42% even though in Obama’s landslide election in 2008 Democrats turned out 6 percentage points less than the poll. The results aren’t even believable because they don’t mirror what we’ll actually see in turnout on election day. It isn’t even close.
Which brings us to the UnSkewedPolls.com analysis. I’ve had a half a dozen emails telling me how amazing this all is and how exciting it is for Romney. Unfortunately they’re doing the same thing the media sites are doing. They’re taking all of the media sites raw data, which is available online and weighting it at R+4. They’re weighting it at 33/37/29. No wonder they’re coming up with Romney up 8 points. Unfortunately our 2010 landslide victory was D+0. Only in Republican fantasy land will Republicans out poll Democrats on election day by 4 points. Just like it’s Democrat fantasy land to believe the media polls which suggest Democrats showing up in droves over and above the 2008 Obama landslide.
This election is going to be somewhere in the middle of 2008 and 2010. I say it will be D+2 or D+3, meaning it’s slightly more favorable to Republicans. If we take the unskewed polls results of Romney up 8 and we consider that they’re off by 6-7 points it suggests that Romney is up by 1 or 2 points. Until today (which I suspect will be an outlier) Rasmussen showed Romney up 1-2 points. That’s probably where we’re at right now nationally. Of course we all know it’s the state results that matter and in that Romney is down big in Ohio and Virginia, including in Rasmussen polling. We shouldn’t get worked up over the absurd national polls, the state polls should cause us a lot of alarm.