Four political science professors are out with a new study that makes a rather novel and thought-provoking case: Swing voters are largely a "myth."
You can find the link to this report co-authored by Andrew Gelman over on Monkey Cage, but here are the operative points:
We find that reported swings in public opinion polls are generally not due to actual shifts in vote intention, but rather are the result of temporary periods of relatively low response rates by supporters of the reportedly slumping candidate. After correcting for this bias, we show there were nearly constant levels of support for the candidates during what appeared, based on traditional polling, to be the most volatile stretches of the campaign. Our results raise the possibility that decades of large, reported swings in public opinion -- including the perennial convention bounce" -- are largely artifacts of sampling bias.
The study focuses on the last 45 days of the 2012 presidential campaign and tries to measure the actual swings over that period by adjusting not just for demographics, but also for partisanship. The theory is that the swings are as much of lack of poll response from supporters of the slumping candidate as they are about voters changing their minds.
So while national polls showed a significant swing toward Mitt Romney following the first presidential debate -- his high moment of the campaign -- this study shows a smaller, less-statistically significant shift.
Whereas adjusting only for demographics yields a six-point drop in support for Obama in the first four days following the first debate, adjusting for both demographics and partisanship indicates that in actuality there was likely only a two to three point movement. More generally, throughout our polling period, adjusting for partisanship reduces swings by more than a factor of two relative to adjusting for demographics alone. Notably, while controlling only for demographics suggests Romney took the lead after the first debate, additionally adjusting for partisanship shows that Obama in fact led the race throughout.
Here's how that looks, with the demographic-only line being the fainter one and the demographics+partisanship line being darker:
As you can see, a much smaller shift occurs after the first debate on Oct. 3, 2012, when you control for partisanship -- though the shift does notably wind up being about four points, by the week after the debate.
The study is an interesting one, and it's quite feasible that interpreted swings in polling do have something to do with supporters of a slumping candidate simply not responding to pollsters. That's a fair point. To say the study suggests swing voters are a myth, though, goes too far. For a few reasons:
1) A presidential campaign isn't a great gauge of swing-iness. After all, the vast majority of Americans knew a lot about Romney and President Obama by late September. And polling at the time suggested this familiarity did indeed leave room for only small swings.
A Pew poll as early as April 2012 showed just 7 percent of Americans said they weren't leaning toward either candidate at that point, and by July, a Washington Post-ABC News poll showed just 6 percent said there was a chance they would change their mind and vote for the other guy over the final three-plus months of the campaign. In 2008, that number was 10 percent, and in 2004 it was 12 percent. Clearly, the swing voter pool is shrinking as Americans become more polarized.
But that's in a presidential campaign in which people almost surely decide much earlier than in a normal race. And given the small universe of swing voters, the fact that there was a three- or four-point shift after the first debate suggests swing voters did exist ... and swung (albeit less than we might have thought).
2) Here's how the study shows the actual shifts, with most voters swinging not from Obama to Romney, but from someone else to Romney or from Obama to someone else.
But not voting or switching between third- and major-party candidates are swings all their own. And in regular polls, if the supporter of a slumping candidate loses the will to respond to a pollster, it's quite possible he or she has lost some motivation to vote, as well.
This study suggests a steady, defined electorate 45 days out, and that's just not the case. Motivation matters. So the faint line above might vary more when partisanship isn't controlled. But in the electorate, partisanship isn't really controlled.
3) The poll was conducted using an opt-in survey for XBox users. As the study's authors concedes, this leads to a sample that is overwhelmingly younger than the general electorate and overwhelmingly more male. The trade-off, they note, is that they are able to have a massive sample size -- more than 80,000 respondents.
While the unrepresentative sample is one thing, the opt-in nature of the survey is perhaps more likely to skew the results. After all, those motivated to respond to polls of their own volition, we would argue, are probably more politically engaged. And more politically engaged people tend to be firmer in their convictions and candidate choices.
This study does suggest that the number of swing voters is far smaller than a lot of people might think. That's true. But other polls (including those mentioned above) already made pretty clear that was the case when it came to the 2012 presidential race.
Whether there are that few swing voters in the actual electorate and in non-presidential elections, I'm not so sure.