I tend to avoid the national polls since I started this blog to discuss the Battleground states. And not national polls, in my opinion, can distract from where this race will actually be decided — which is in a handful of Battleground states. While very smart people make compelling arguments that national polls are normally a better “tell” than battleground state polls, there is very little normal about the 2012 election. We’ve posted a couple of times how a) various allegedly credible pollsters can have both consistent as well as egregious biases in their results., and b) polling voter screens often have built in Democrat biases even without nefarious intent. That is why while polls are fun to watch, I don’t get too excited one way or the other until we get closer to election day and many of these inaccuracies are sifted out.
With that in mind, many pro-Romney people have been gnashing their teeth over the spate of national polls over the last couple of days showing Obama with a sizable if not commanding lead. Has Obama’s unprecedented early ad spend of relentless negativity actually impacted the race or are all THREE polls outliers versus the Gallup and Rasmussen polls that have shown the race consistently knotted up? Quick tangent on Rasmussen Reports — his polling has called the last few elections more accurate than any pollster pretty much nailing the last two Presidential election outcomes within 1 percentage point of the actual results. Any time a Lefty dismissed Rasmussen as a “Republican,” they do so at their own peril. Either he is accurate or he is not. And the overwhelming evidence shows he is the most accurate Presidential pollster anywhere.
But today pro-Romney people are concerned that Fox News (Obama 49, Romney 40), CNN/ORC International (52 to 45), and Reuters/Ipsos (49 to 42) all have Obama with sizable leads in national polling. Three independent pollsters, including “fair and balanced” Fox news can’t all be wrong by roughly the same amount could they? The answer is they absolutely could be independently and consistently wrong if they use the same sampling bias.
Let’s break them down 1-by-1:
First, the Fox News poll data reveals they surveyed registered voters. That alone is with approximately 2-points in favor of the Democrats. In the 2010 mid-terms the registered voter surveys revealed a pro-Democrat bias versus likely voters by four-points but in a Presidential election year, this disparity can be expected to shrink. Note among “extremely interested” voters (i.e. people nearly guaranteed to vote), the race was a dead heat 48 – 48. Second the party ID was Democrat 44, Republican 35, Independent 20 or D +9. In 2008 during a huge Democrat wave year (Bush fatigue, war fatigue, financial crises, recession, unappealing Republican alternative, Republican brand damage, etc.) the party ID of voters gave Democrats an advantage between D +7 or 8. This followed 2004 when the party ID on election was perfectly split between the major parties. Now following 4 years of Obama, an anemic economy, relentlessly high unemployment, the Tea Party revolution, increased party registrations, better alternative candidate, Republican brand revitalized and a host of other pro-GOP factors there is a near zero percent chance that Obama meets or exceeds his 2008 voter turnout advantage. Yet, this poll and many others reflect a November turnout far in favor of Democrats than they can reasonably expect. My personal guess is that on election day party ID will favor the Democrats by 2 percentage points of D +2. Suddenly an Obama headline lead of 9 points in this poll quickly shrinks to somewhere between 2-3 percentage points — well within the margin of error and consistent with Gallup and/or Rasmussen
The CNN/ORC International poll also uses Registered Voters (as a subset of the even less accurate survey of “adults”). That shaves approximately 2-points from the Democrats to begin with (as discussed above). Second, any survey of “adults” has averaged a 7-point advantage for Democrats versus likely voters screen (per Nate Silver, FiveThirtyEight blog, New York Times: “surveys conducted among likely voters are about 7 points more favorable to the Republican than those conducted among all adults, whether registered to vote or not…there is evidence that likely voter polls are more accurate, especially in Congressional elections.”). This renders the rest of the “adults” survey largely meaningless. But for the Registered voter screen, a breakdown of party ID would have been extremely helpful. Unfortunately CNN provides no such data, but you can back into a fair guess of the split based on a question orf self-identified “Republicans” and “Independents leaning Republican” in question 23 of the survey. This reveals a Republican ID of 45%, leaving 55% Democrats or Independent leaning Democrats. CNN provides no further information so you have no idea the percentage of leaners this includes (i.e. a way to separate out out actual Independents) but in an already curious survey fraught with issues, this is only the latest. Anyway, the 45 – 55 split results in a D +10 advantage, again far above the D +7-8 in 2008 and exceedingly higher than my estimate of a D +2 advantage in November 2012. With so much wrong in this survey, its meaningless to even guess what an adjusted outcome would be in this poll.
Finally, the Reuters/Ipsos poll uses the same methodology as CNN/ORC and surveys over 1000 “adults” and then provides race-horse results of registered voters showing Obama with a 7-point lead. This is consistent with Ipsos methodology as explained here. Beginning with acknowledged inaccurate survey samples weakens confidence in the poll out of the gate but at least Reuters/Ipsos provides a voter ID screen (final page: Demographics). This came out Democrat 32, republican 28 and Independent 40 or D +4. When pressed, the advantage shrunk to D +2. These are fair representations if slightly high for Democrats on the first screen. But if only 1 of 3 polls has Obama with a 3-5 point lead while the other two have Obama with a slim or non-existent edge, I’m not going to get too worried.
Once likely voter screen become the staple of all polling services (next month) huge outlier polls like these three will be a long distant memory. For now, I’m going to readily ignore them and so should you.