Right now, the Real Clear Politics (RCP) average of polls shows a dead heat between President Obama (47.5%) and Mitt Romney (47.3%) nationally. The moving average, taking into account the 11 most recent national polls, has included only “likely voters” since Gallup switched to using its likely voter model in early October.
Based on two years of academic research in polling methodology, when it comes to assessing polls and their validity, I have one central rule: be constantly vigilant. I am skeptical of every poll I see because their results are so sensitive to sampling, question wording, question order, acquiescence, and social desirability biases that any number of factors can invalidate their results.
But I adamantly believe that taken together, polls can reveal trends. In particular, the same poll conducted with the same methods repeatedly over time helps us better identify the electoral system’s moving parts. Considered together, different polls with wildly different methods and different definitions make it far more difficult to declare trends definitive, but using them, we can still sketch a general idea of where things are going.
For instance, consider the polls currently included in RCP’s average. Five show a tie, three show a 1-point Obama advantage, and one shows a 1-point Romney advantage. With statistical margins of errors ranging from 2.6% to 4.0%, they all show statistical dead heats.
But two of the polls show 5-point advantages for one candidate or the other. Those stand out, and they should. Gallup’s shows Mitt Romney leading among likely voters by 5%, and the National Journal’s shows President Obama leading among likely voters by 5%. How can two polls, both purporting to sample “likely voters” together account for a full 10-point swing?
One issue: sampling bias. When the respondents polled and deemed “likely voters” don’t look like the people who are actually likely to vote, the poll has a biased sample. Gallup is fairly transparent about how it determines whether someone is likely to vote. It assesses respondents’ answers to seven questions and whether they meet certain criteria.
Based on its polling during the first three weeks of October, Gallup found that the demographic breakdown is similar to 2008’s among those it deems likely voters. The only major difference? Party affiliation.
In Gallup’s 2008 final pre-election poll, Democrats comprised 39% of estimated likely voters, independents comprised 31%, and Republicans 29%. When you take into account independents’ lean, likely voter numbers broke for Democrats, giving them a 12-point advantage (54% to 42%).
This cycle, the advantage is reversed, but the margin is slimmer. Gallup's late October 2012 numbers showed that Democrats comprised 35% of estimated likely voters, independents comprised 29%, and Republicans 36%. When you take into account independents’ lean, likely voter numbers broke for Republicans, giving them a narrow 3-point advantage (49% to 46%).
These numbers align with the larger trend among polls and 2010 voter turnout: higher Republican participation and lower Democratic participation. That is why the National Journal’s poll confounds me. It estimates a 5% Obama lead over Romney, but assumes that Democrats outnumber Republicans by 8% among likely voters – a point higher than Democrats’ 2008 advantage. Though there will likely be significant variation among individual states (especially swing states), assuming an 8-point likely voter advantage for Democrats nationally stretches the bounds of credulity.
Turnout among millennials (who have tended to vote for Democrats in recent years) could be at its lowest point in decades, a massive reversal from 2008. Democrats are not mailing in requested early ballots as frequently as Republicans. And enthusiasm among Democratic voters has been on the decline for a while now (see: here, here, and here.) After all, the “Obama That I Used to Know” parody went viral among liberals for a reason.
No one can say what turnout numbers will actually be in the end, but assuming an 8-point advantage for Democrats – larger than 2008’s – ignores both waning Democratic enthusiasm and growing Republican enthusiasm since then. And even if Republicans’ enthusiasm peaked in 2010, it is fairly safe to say that Republicans are more fired up today than they were four years ago. Contrary to what the National Journal projects, that is a serious disadvantage for Democrats.