ELECTIONATE OCTOBER 16, 2012
After the first presidential debate, YouGov recontacted likely voters in 25 states, including the eight battlegrounds and found Romney making small gains. These state findings come on the heels of last week’s YouGov/Economist national survey, which found Obama maintaining a 3-point lead among likely voters. Although later polls have since shown Obama holding a modest edge, the YouGov/Economist poll was the only survey showing Obama maintaining a national lead when it was released least Wednesday.
But the surveys weren't just your typical polls, and not just in the sense that they were conducted online. YouGov recontacted respondents from last month's polls, so Obama's resilience means that few of the voters surveyed by YouGov switched allegiances after the first debate. This raises the possibility that shifts in the post-debate polls are due to changes in response rates, not genuine shifts in voter opinion. Changes in response rates are quite common after party conventions and they are partially responsible for the post-convention “bounce” phenomenon. The resoundingly positive media coverage for Romney after the first debate could have plausibly triggered a similar movement in the polls. According to YouGov’s Doug Rivers, Democratic panelists were 4 percent less likely than Republican panelists to respond to their most recent request, and that’s despite higher response rates than telephone pollsters. And YouGov is not alone at hinting at a response-rate related bounce for Romney; many of the surveys with methods that tend to control for changes in response rates, like RAND and Rasmussen, showed Romney making fewer gains than more traditional telephone surveys weighted to demographic targets.
However, interviewing past respondents has risks of its own. YouGov managed to recontact more than 80 percent of respondents in some states, but the drop-off voters might well be the ones who shifted toward the undecided column or have become less likely to turnout in November. The experience of being interviewed the first time might also cause voters to respond differently than they would if they hadn’t previously been interviewed. None of these risks necessarily invalidate YouGov's findings, but they're all reasons why the polls method produce different numbers than a more traditional survey.
Of course, telephone surveys have real limitations in era of cell phones and declining response rates. In particular, it’s hard to argue that automated surveys with vanishingly low response rates and no interviews with a large share of the electorate represent a better method for gauging public opinion than an online panel, even if one isn’t yet convinced the online panels have come of age. The ability for YouGov to provide hard data demonstrating that Democrats aren't responding as often as Republicans is potentially very informative and not easily achieved with telephone surveys.
But although YouGov raises the possibility that the movement in the polls might not actually reflect real change, YouGov’s state polls align well with the CBS/NYT/Quinnipiac and NBC/WSJ/Marist polls conducted last week—two of the only live interview firms to conduct post-debate battleground state surveys. Those surveys didn't show much movement in Obama's direction either, and the battleground state polls haven't shown as much movement in Romney's direction as the national polls. But at the moment, YouGov’s national survey isn’t a complete outlier either, although it is tied with the ABC/Washington Post poll as the least favorable post-debate national survey for Romney.
The YouGov poll raises the possibility that at least some of Romney’s post-debate gains are due to changes in response rates rather than voter preferences, even if it’s not clear exactly what share of Romney’s movement can be attributed to either possibility. If Obama has a decent debate performance and suddenly the polls move back toward YouGov (without similar movement in YouGov, RAND, or Rasmussen in Obama’s direction), it might be fair to reinterpret Romney’s bounce as driven primarily by changes in partisan response rates. But it will be even more interesting if YouGov and the other pollsters diverge through November.