"The only poll that matters is the one on Election Day."
This well-worn campaign cliché is taking on new meaning in this midterm election. Polls this year are differing sharply based on various assumptions of who will actually turn out.
This is nothing new -- many pollsters are screening for so-called "likely voters," and their methods can vary widely. And it's also no surprise that many "likely voter" polls are finding different results than polls of all registered voters.
Whatever the methodology or numbers, these polls collectively demonstrate one thing: the makeup of the electorate will determine this midterm election. Victory or defeat in many races will hinge less on changing voters' minds, and more on changing the voters who turn out.
Pollsters' assumptions about the makeup of the electorate in November, and how they weigh their participants' responses accordingly, can have the same impact on poll numbers as turnout has on elections. Take Florida, for example: one poll showed Republican Gov. Rick Scott leading his opponent, former Gov. Charlie Crist, by five points. Another poll taken soon after found Crist leading by three points. The first poll weighed responses for partisanship assuming that the Florida electorate would be 5% more Republican than Democratic (35% to 30%), and only 20% African American or Hispanic. The other poll showing Crist in the lead was comprised of 40% Democrats and 38% Republicans.
Similarly in Colorado, a recent poll by Quinnipiac University showed a 10 point lead for Republican Bob Beauprez over Gov. John Hickenlooper using a sample comprised of 34% Republicans and 27% Democrats, and only 8% Hispanics. Another poll released the same day found a two point lead for Hickenlooper with a sample of 33% Republicans and 31% Democrats, and 12% Hispanics. (Note: The Quinnipiac University poll has been heavily criticized for a GOP-skewed sample; for example, their sample of Hispanic voters was 33% lower than the proportion of 2010 turnout.)
Generally speaking, the "likely voter" polls assume a high percentage of progressive voters will stay home during the midterm election, and are therefore less favorable for Democrats than polls of all registered voters. We can see this clearly in a recent CNN poll, which showed New Hampshire Democratic Sen. Jeanne Shaheen leading by seven points among registered voters (which is in line with America Votes' internal polling and most public polls), but tied among "likely voters." And in Arkansas, CNN found Democratic Sen. Mark Pryor trailing by 2 points among "likely voters," but ahead by 9 points among all registered voters.
The key variable that is difficult to capture in many "likely voter" polls is the impact progressives' Get Out The Vote (GOTV) efforts will have on the electorate. Just like poll numbers can be altered by the makeup of the sample, Election Day results -- "the only poll that matters" -- can be altered by the makeup of the actual electorate. Right or wrong, these polls reiterate exactly how important it is for progressives to turn our registered voters into actual voters by November 4.
America Votes and our partners are working to turn out voters who stayed home in 2010 by running tightly coordinated ground games across our 20-state network. For example, America Votes has targeted nearly 1.1 million progressive voters in Florida, and 500,000 progressive voters in Colorado who are likely to stay home, or "drop off" in midterm elections. In Wisconsin, we are targeting nearly 500,000 "drop off" progressive voters and pin-pointing those most likely to be impacted by the newly implemented voter ID law. Our coordinated programs in these states will connect with targeted households multiple times between now and Election Day.
Although some polls are screening "likely voters" too tightly, progressives should not dismiss them. The fact is that midterm electorates are undoubtedly much different than presidential year electorates. But from Florida, to Colorado, to New Hampshire -- polls show that the majority of voters back progressives in key races, and we have the infrastructure and resources to make an impact.
Polls are based on statistical methodology and electoral assumptions - part science, part art. But at the end of the election cycle, our collective work on GOTV is the key to getting the results we want.