Wednesday, October 10, 2012
Even with a few polls showing Mitt Romney leading President Obama, some Republicans allege that pollsters consistently underrepresent the number of Republicans in their samples. The charge certainly deserves serious attention. Just about every major news story about the election mentions polling. Potential biases in survey results, amplified by their frequency of coverage, threaten to distort people’s assessment of the presidential race, and perhaps even alter the outcome. A lead in the polls helps build momentum, and momentum matters, especially in the last few weeks of the campaign. Even small biases could have a big impact.
Some of the present confusion comes from speculation about what the partisan makeup of these polls “should” look like. Most pollsters adjust their initial results to bring their samples in line with the country’s demographic profile. They weight for variables like sex, age, and education based on U.S. Census data. But partisan identification is an attitude, not a demographic variable. Party identification is fluid for many Americans, and pollsters who have conducted call-back surveys (where respondents are interviewed on two separate occasions) find that many respondents change their partisan identification between surveys. The partisan balance of the country is always changing. For those with intense partisan allegiances, this concept is particularly hard to understand.
Since partisanship is an attitude, then “correcting” polls to fit a certain partisan profile is akin to starting with the conclusion you want and working backwards. It’s manipulative and negates a poll’s ability to inform based on its findings.
Since partisanship is an attitude, then ‘correcting’ polls to fit a certain partisan profile is akin to starting with the conclusion you want and working backwards.
Many critics contrast today’s surveys with the exit polls from the 2004 election, when an equal number of Republicans and Democrats voted. They assume that the 2012 electorate should look more like it did in 2004 than in 2008. But since 2004, the number of self-identified Republicans has dropped and the GOP’s popularity has declined. In late September 2004, 39 percent of Americans identified themselves as Republican, according to Gallup. In Gallup’s latest poll, only 28 percent do so. But where did these voters go? Some may have converted to the Democratic cause in 2008, while others may be sitting this election out. But in six new polls from National Journal, Quinnipiac, CNN/ORC, ABC/Washington Post, Pew, and Politico/GWU, Romney leads among political Independents. With the exception of the Pew poll, these surveys have more Democrats than Republicans in their samples. In addition, most major polls show a higher portion of Independents than voted in the 2008 and 2004 elections. Considering the bitter political battles over the budget, contraception, and the Republican primaries over the past two years, it’s far more likely that the polls reflect movement of Republicans into the Independent camp and not a systemic poll bias.
Figure 1: Independents in the Polls
A portion of these Independents are probably disaffected Republicans who have not changed their presidential voting habits, just their party allegiance. As Election Day approaches, some will come to identify themselves as Republicans. The latest Pew poll, the first following the presidential debate, shows that some realignment might already be happening. The percent of likely voters identifying as Republican jumped 7 points from the previous mid-September poll and Independents tilted more toward Romney. This suggests that some Independents are filtering back into the Republican Party. However, Independents’ more tenuous allegiance to the Republican Party should be taken as a warning sign for Republicans—not something to dismiss as polling bias.
The perception of bias is compounded by the media’s unfortunate tendency to toss all polls into the same blender to produce “comprehensive” averages. But not all polls are created equal. The 2008 presidential exit polls showed Democrats outnumbering Republicans by a 7-point margin. Few expect that the 2012 electorate will look like that of 2008. Some polls, especially polls of registered voters, show a Democratic advantage as large as in 2008. But likely voter samples show a narrower Democratic advantage or a close race despite a Democratic advantage. This suggests that likely voter samples are more trustworthy than registered voter samples.
Figure 2: Partisan Composition of Recent Polls
The partisan composition of the country has changed and will continue to change right up to November 6. But what also deserves attention is the way the polls are being reported. They are increasingly used as scorecards to judge every minor action in the presidential race. In reality, they are too imprecise to warrant such weight and analysis. They are tools that should be used to approximate. Voters should keep this in mind as they are confronted with the inevitable avalanche of polls between now and Election Day.
Andrew Rugg is a research assistant at the American Enterprise Institute.
FURTHER READING: Rugg also writes “Maybe Americans Really Are Ready for Spending Cuts” and, with Karlyn Bowman, discusses “Obama’s Weakness in Historical Context” and “Americans and the Terrorism Threat 10 Years After 9/11.” Michael Barone says “States Aren’t Red or Blue Forever” and “When It Comes to Polls, Readers Beware.” Marc A. Thiessen explains “Like Reagan, Romney Can Still Win.”
Image by Dianna Ingram / Bergman Group