On election night, the exit poll result showing voters under age 30 had increased as a proportion of the electorate - 19%, up from 18% in 2008 - was perhaps one of the most shocking conclusions of the night. Poll after poll showed serious disappointment, disillusionment, and detachment from voters under age 30. Republicans in particular expected that the Obama electorate could not be replicated. According to the exit polls, they were wrong.
Yet the Census Bureau released their report on voting rates by demographic group from the 2012 election. Based on a supplement to November's Current Population Survey, the report tells a slightly different story than the exit polls about what happened with the youth vote in this last election.
In 2008, the CPS supplement estimated that voters aged 18-24 cast about 12.5 million ballots, comprising roughly 9.5% of ballots cast, an estimate not so terribly different from the exit poll finding that 18-24 year olds were 10% of voters in that election. Broadening out to the full 18-29 group, the CPS estimates this age group cast 22.4 million ballots out of the 131.4 million overall - about 17%. Again, not so far off from the exits.
However, in 2012, the CPS shows about 11.4 million voters were 18-24 years old, or 8.5% of the electorate, while the exit polls actually showed the proportion of voters under age 25 going up to 11%. Looking at the broader 18-29 group, they show about 20.5 million votes cast, comprising 15.4% of the electorate - a far cry from the 19% reported in the exit polls.
So what happened? How could two different, reputable studies come to such different conclusions about the changing proportion of the electorate under age 30?
My theory is that issues with self-reporting of voting behavior may have had something to do with it. In 2008, the CPS tables show an estimated 131,144,000 votes were cast, which is pretty much right on the money. Yet in 2012, the total they give for "reported voting" is estimated at 132,948,000. The FEC begs to differ, with an estimated four million fewer votes having actually been cast. While their data matched up perfectly with reality in 2008, there's a pretty big gap this time around.
The Census Bureau folks acknowledge such discrepancies are possible, and this isn't a huge knock against them. They note that their report may overreport votes in part due to people who truly voted but whose ballots were invalidated, as well as people who misremember their own voting history or who feel social pressure to say "I voted" even when they know they actually didn't vote. (UPDATE: Michael McDonald, also writing for Huffington Post Pollster, eloquently dives into the mechanics behind the CPS and the potential over-reporting issues here.) There's an excellent paper by Masa Aida and Todd Rogers that matches up self-reported voter intent with who actually shows up, and their very first sentence notes:
People tend to over-estimate the likelihood that they performed a socially desirable behavior in the past (eg, whether they voted), and to over-estimate the likelihood that they will perform a socially desirable behavior in the future (Silver, Anderson, and Abramson 1986; Snyder 1974; Epley and Dunning 2000).
However, just because there's evidence that people over-reported their own voting behavior doesn't mean this automatically leads to underestimating the youth vote. The question we really need to answer is whether or not older voters were, frankly, more apt to feel obligated to lie about their own voting behavior. It's not inconceivable to think that a young person would feel less social pressure this time around to vote and therefore would feel less pressure to misstate whether they voted to conform to expected norms. On the other hand, if young people felt less social pressure to vote, then maybe youth turnout really did drop.
There's also the fact that exit polls are far from perfect. (This is nothing new of course; I found a great little interview with Warren Mitofsky from twenty years ago where he's being asked about the challenges of exit polling.) You're not legally required to take an exit poll in the same way you're legally required to reply to the CPS. If you're a busy person, stopping to fill out a lengthy questionnaire may be the last thing you want to do just after voting. Not to mention, shifts away from election day in-person voting have presented a major challenge to the exit polls. However, on the bright side, the exit polls are ultimately weighted back to known precinct returns.
All of which is to say that both studies have their imperfections, but are the best data we have. Both CPS and the exit polls have sample sizes that are huge compared to the traditional polls we consume day in and day out. While both CPS and the exit polls showed similar results for the proportion of voters under age 30 in the 2008 election, they find very different conclusions about what happened with youth turnout in 2012. There's still some more research to be done to get to the bottom of where that gap comes from.
What this debate over the proportion of young voters really overlooks is the far, far more important issue of how those young people voted. Over at The Fix blog, Scott Clement and Sean Sullivan write about this same report today, concluding:
Here is a group that by and large votes Democratic. The more the party can grow the pool of younger votes, the better it is likely to fare overall. Conversely, the smaller the pool becomes, the better it is for Republicans, who don't fare so well with the youth vote.
I think this focuses too much on the turnout question. It assumes that Obama's massive margins are the new normal and that all Republicans can do is to hope young people stop showing up at the polls. Like I wrote last year on this blog, worrying about whether young voters will be 16% or 17% or 19% of the electorate is an interesting game, but the bigger question for Republicans is how they plan to stop losing young voters by such huge margins.
The good news for Republicans is that as recently as the 2000 election, the VNS exit polls showed there was effectively no correlation between age and voting. Republicans don't have to give up on the youth vote and just hope they stay home. They should instead focus on winning them.
In the mean time, the debate over whether or not young voters increased or decreased as a proportion of the electorate remains unsolved. I'd argue it's also not nearly as important as what Republicans plan to do to win over more of those young people who do show up.