The pundit battle over Nate Silver and Intrade odds-making over an Obama victory on November 6 has covered "skewed" polls, discussions of oversampling of one political party over another, and finally, estimated turnout of voters by race. An odd pattern has emerged, one in which Romney appears to be set back in the closely-watched Real Clear Politics poll averages each time his polling average over Obama gets to the 1% mark.
The pattern is apparent in battleground state polls as well. Take Iowa for example.
Last week, a Rasmussen poll showing a tie and University of Iowa and PPP polls showing Romney up one point in Iowa was quickly overshadowed by three successive polls showing Obama ahead by 2 to 6 points.
If this were all people looked at, they would be convinced that the president was safely ahead in the battleground state of Iowa. Much of the skewed poll/"oversampling" argument rests on disagreements about who will turn out on Election Day. Will the electorate look like 2008, with a high Democrat turnout, or 2010, with a high Republican turnout? Many arguments and polls ignore actual voter registration, which is available via most state Registrar of Voter websites, and is frequently updated.
The most recent NBC/WSJ/Marist poll that shows Obama up a whopping, dominant 6 points in Iowa was weighted for the following percentages among likely voters: 34 Democrat 31 Republican 34 Independent. Obama won Iowa in 2008 by 9.5 percentage points and his final Real Clear Politics average in the state was 15.3 points over John McCain. Few pollsters covered the same states and time periods in 2008 as in 2012, but a Survey USA poll of Iowa voters conducted just prior to the 2008 election that showed Obama up by 15 points over McCain used the following partisan breakdown: 45 Democrat 29 Republican and 25 Independent.
A quick look at the Iowa Registrar of Voters website shows that today, actual Iowa registration is 32 Democrat 33 Republican 35 Independent. Republicans overtook Democrats in party registration in 2009 in Iowa and have continued to hold that advantage. Looking within the poll itself, 48% of both likely and registered voters "approved" of the job Barack Obama is doing as president, yet the public is asked to believe that a mysterious 2% of those who disapprove will still vote for the president.
Let's look at the most recent NBC/WSJ/Marist poll of Wisconsin, which contributes to the president's RCP advantage of 5 points in the Badger state. This poll samples 34 Democrat 29 Republican and 35 Independent voters. Of these voters, 49% approved of the job Barack Obama was doing, and 48% stated they would vote for Tammy Baldwin, the Democrat candidate for Senate. Baldwin was up only 1 point on Tommy Thompson, former Wisconsin governor and Republican candidate for Senate. Somehow, another 1% said they'd vote for Barack Obama for president, and 1% fewer for Mitt Romney and Paul Ryan. They did not actually say this, of course. The poll was simply adjusted to reflect pollster estimations of who would show up on Election Day, and pollster priorities were on the presidential, not the Senate race.
Wisconsin does not record voter registrations by party, but an LA Times article from May, 2012 indicated that Governor Scott Walker had moved into the lead in the hotly-contested recall election, and that Republicans had taken the advantage in the state. The Milwaukee Journal-Sentinel has merely reported that, although Wisconsin has voted for the Democratic candidates in the past five presidential elections, the state's more recent voting history has been characterized by extreme swings from one part of the state to another.
Finally, much has been made of the CBS/Quinnipiac polls that are single-handedly making it look like Barack Obama will cruise to swing state victories in Ohio, Florida and even Virginia.
The most recent Quinnipiac party ID breakdown was Florida: 37 Democrat, 30 Republican and 29 Independent, Ohio 37 Democrat, 29 Republican, 30 Independent, and Virginia, 35 Democrat, 27 Republican, 35 Independent. The excuse for these breakdowns has little to do with 2008 turnout and a lot to do with who the pollsters believe will show up to vote on election day.
So, let's look at Florida, with CBS/Quinnipiac saying Obama is ahead by 1 point, 48% to 47%. Actual Florida Registrar of Voters closing figures for the 2012 primary elections were 36 Republican 40 Democrat 21 Independent. Slightly different from the weighted poll. Although partisans might be excited by a 4-point Democrat advantage in party registration, the poll just removed 4 points from the Republican registration advantage and conveniently ignored the presence of Florida governor Rick Scott and Senator Marco Rubio, and other trends, such as first-term Congressman Allen West leading his race despite a snowstorm of attacks from the left, and the strong showing of Adam Hasner, a strong Republican challenger for an open Congressional seat in a new, Democrat-leaning Congressional district that covers parts of Palm Beach and Broward counties. Hasner is polling evenly in local polls against his Democrat opponent, Lois Frankel, but he has received the same poll-depressing treatment as more high profile Republican candidates —like Mitt Romney — from larger polling organizations.
Way back in early 2008, polls were still favoring Barack Obama, but he had a different opponent: Hillary Clinton. Clinton won the New Hampshire primary, but one would never have expected that from the polls conducted before the election. The Atlantic and the Huffington Post featured her striking pickup of eight points over her poll support of only 30%. At the time, writers promoted the "Bradley effect," although polling didn't support that thesis. The Bradley effect refers to the 1982 California gubernatorial race in which highly qualified African-American candidate Tom Bradley, a popular former mayor of Los Angeles was polling well ahead of his Republican opponent, George Deukmejian. Bradley lost the election by less than 4,000 votes with nearly 8 million votes cast. But as the New Hampshire chart shows, Obama's vote total was similar to poll numbers, and it was Clinton who surged far ahead from her poll level of support. The "Bradley effect" only applies if far fewer people vote for the minority candidate than tell pollsters they will vote for them.
No one really knows what will happen on Election Day. Judging by other manipulations we have seen throughout this grueling campaign season, such as the $1 billion spent on attack ads by both campaigns, which appear to have had little effect, and the "War on Women" that exists primarily in media minds, it's a good bet that these poll games will also have little effect.