Likely Voters and Early Voters
I often get questions from friends about the polls for the Presidential election. Back in January I discussed the routine abuse of the margin of error. Another big issue in polling is how you determine who a likely voter is.
To determine if you’re a likely voter, pollsters will typically ask if you’re registered to vote, if you intend to vote, and if you voted in the last election (in this case, the last Presidential election). They typically give the most weight to your answer about voting in the last election, so they consider past behavior a better indicator than stated intentions. Some will also ask if you know where your polling place is located, as an additional screen. In this election, with enthusiasm sky-high among the African-American community and young voters – both in favor of Obama – pollsters aren’t as sure about how to screen for likely voters. Gallup dealt with this dilemma by simply publishing an additional set of results – one using their traditional likely voter model, and one that counts you as a likely voter strictly by whether you say you plan to vote. What’s interesting in the final Gallup poll yesterday was that the results of the two models converged – both give a roughly 11 point advantage to Obama. (If you’re curious, back in August Nate Silver wrote an astute critique of how Gallup screens likely voters, arguing that it could be substantially underestimating the ultimate impact of unlikely voters).
Silver also wrote back in May that “youth turnout in the primaries increased by 52 percent as a share of the Democratic electorate.” This was part of an analysis of what the electoral map would look like if Obama could substantially increase turnout among young voters and the African American community. Silver went on to say:
The ability to bring new voters to the polls remains Barack Obama’s most significant electoral advantage, both relative to Hillary Clinton and John McCain. Indeed, current polling may already be underestimating Obama’s strength against McCain if it does not account for improved turnout among Democratic-leaning groups like young voters and African-Americans, who have participated in record numbers in this year’s primaries. If Obama can parlay that advantage with a strong ground game, he very much could redraw the electoral map.
Running counter to this argument is Gallup’s results published yesterday, showing no indication of an overall surge in youth interest in the election over 2004 levels. But in terms of the final outcome, they give themselves some wiggle room by saying “[voter mobilization efforts by the campaigns] can convince people with little motivation or interest in the campaign to actually vote on Election Day.” Gallup’s assessment of youth interest contradicts other evidence, but if Gallup is right, this is where Obama’s massive get out of the vote effort may prove decisive. In 2004 the Democrats registered new voters in much greater numbers than the Republicans, just as they have for this election, but in 2004 they didn’t turn out their voters nearly as well as the Republicans.
The actual turnout numbers for young voters and African American voters are what will make the difference between a narrow Obama win and big one.
The other thing to keep an eye on is the significance of the huge number of early voters. The national numbers for early voters are about 10% higher than 2004, with Democrats submitting 1 million more early ballots than Republicans. And the LA Times reports: “In three swing states — North Carolina, New Mexico and Colorado — the number of voters who have already cast their ballots has reached more than 70% of the number who voted there in 2004.” The question is whether the campaigns are simply banking votes early, or if the early votes are presaging even bigger – and equally lopsided – turnout numbers today, Election Day.
Most of the major polls have Obama up between 6-8%, which I think is about right. In 2004 I made predictions for the 13 swing states, and got 10 of them right (and since two of my wrong predictions were Ohio and Florida, I got the outcome wrong too). I’m not going to do a state by state breakdown again, as there are now lots of people already doing it, with skills and resources beyond mine. Instead, I’ll point you to political scientist Larry Sabato’s Crystal Ball map, which I think looks just about right. He has the electoral college going to Obama 364-174 (I credit Sabato for coloring every state blue or red, and not wimping out like many of the others making predictions, by leaving the close states as uncounted tossups).
Even if there is a big Obama win, be wary of pundits using the word “landslide.” While there is no hard and fast definition, if you look at the 3 elections since WWII that are generally considered landslides (Johnson v Goldwater, Nixon v McGovern, Reagan v Mondale), the popular vote gap between the winner and the loser was around 20%.
If Obama overperforms his poll numbers in states where he is very closely behind McCain, he could turn some red parts of Sabato’s map blue: North Dakota, Nebraska’s 2nd district (Nebraska divides it’s electoral votes by Congressional districts), Georgia, Indiana, and Montana. If he significantly underperforms, then this is what possible winning electoral maps look like for McCain.
