Less than 100 days from the 2012 election and the polls show Obama with anywhere from a slim (Gallup +1 Obama, Rasmussen tied) to large (Reuters +7 Obama, Pew +10 Obama) national lead. Most recently a flurry of new polls from the combined efforts of CBS/NYTimes/Quinnipiac polling have raised eyebrows among GOP circles and others. In the polling of three swing states, Ohio, FL and PA, from July 24th to 30th among likely voters Obama led in all three. In the latest round of polling, released today in the swing states of CO, WI and VA Obama led in WI 51%-45%, VA 49%-45% and trailed in CO 51%-45%.
Leaving aside the OH, FL and PA survey results for the moment the latest round of polling in CO, VA and WI is illuminating. The first stand-outs from the surveys are the obvious. In all three states Romney has leads among men while women favor Obama. Romney is also running strongly with blue-collar men, less so but still strong with non-college educated women, while Obama wins among college educated women. College educated men lean towards Romney.
But what is interesting to many pundits, analysts, and amateur election followers (me) is the samples of Likely Voters these surveys are using. Now to be clear, what “Likely Voter” samples individual pollsters use vary. A lot of data is taken into account in deciding on a voter sample such as past election turnout, the filters used to glean who is likely to vote in 2012, demographic weighting, etc. The samples from the CBS/NYTimes/Quinnipiac however have been widely divergent from past turnout results.
Take the sample of Virginia. In the survey a mere 23% identify as Republican, 30% as Democrats and 40% as Independents. Get a little deeper into the weeds and 46% of those independents identify more with the GOP. Ideologically in VA 20% identify as liberal, 38% as moderate and 36% as conservative. Compare this to the 2008 and 2009 exit polls however and there is a significant difference. In 2008 39% identified as Democratic, 33% as Republican and 27% as Independents. In 2009 when Republicans won all statewide executive offices 37% of voters identified as Republicans, 33% Democrats and the rest Independents. So the VA sample from prior election turnout results far over samples independents. Romney leads among this group narrowly in VA which means if the poll is accurate a small segment of independents that feel closer to the GOP are backing Obama. Hard to believe I know.
Now let’s look at WI shall we. In the WI survey 27% identify as Republican, 34% Democrats and 33% Independents. In 2008 the electorate identified as 33% GOP, 39% Democrats and 29% Independents. In the 2010 race for Governor (I purposefully exclude the 2012 gubernatorial recall exit polls) 37% identified as Democrats, 36% GOP and 27% as Independents. This is important because the survey is predicting that Republican turnout will be lower than where it was in 2008. For those who need a refresher course the 2008 election was the worst election for the GOP since 1992 and nationally their turnout was far below 2000 levels.
In CO Romney retains a lead but even that sample is interesting. It over samples Republicans compared to their 2010 turnout levels by a point and under samples Democrats by a whopping six points compared to 2010. Mid-term electorates tend to be whiter and more rural, thus favoring the GOP by the way.
These samples all point to the inescapable fact that the surveyors are not basing their turnout models, at least partisanly, on past turnout. In CO their turnout model gives Romney the edge and in WI and VA their models/results give Obama an edge (though less so in VA). It is possible that the surveyors are basing the number of independents they sample on Gallup’s national tracking numbers. In 2012 the highest number of people nationally who identified as independents was at 39%. It would thus make sense to see many people in these swing states identifying as Independents.
But as Sean Trende of Realclearpolitics has pointed out, simply identifying as an Independent does not mean you really are or even identify as one at the moment. Most pollsters outside of the CBS/NYTimes/Quinnipiac have found more people identify as Republicans or Democrats than Independents. In the question surveys it may be more telling which political party the pluralities that identify as Independents feel closer to. And if this is the case it is bad news for Obama because in all three states pluralities of Independents feel closer to the GOP.
Still, samples and partisanship aside the survey results do not seem to match up to what one would expect of the samples. If a plurality of independents identify as closer to the GOP and Romney leads among independents in VA why is he not ahead? Ditto in WI. And how does he hold a six point lead in CO yet only 40% of CO’s 37% of Independent voters feel closer to the GOP? One is forced to ponder these questions.
It is only on election day we will get to see the final election results and turnout. But until than we must deal with surveys such as the CBS/NYTimes/Quinnipiac sample that seems to be far off from other samples. Makes for interesting news I suppose.