ELECTIONATE OCTOBER 5, 2012
The first wave of post-debate polls are still in the field, so we're still waiting to determine whether Romney received a sizable bounce. But a baseline is necessary to judge the size of Romney’s bounce, so let's take the polling doldrums as an opportunity to consider the state of the race heading into the debates.
Why an average of post-DNC polls?
First, the pre-debate polls were consistent with the post-DNC average. If the race tightened over the last few days, it only tightened from Obama's post-"47 percent" peak to something more typical of the rest of September. The president was also aided by a wave of good polls on Tuesday and Wednesday, including an NPR poll showing Obama leading by 7 points.
If the race currently approximates the post DNC environment, then a longer-term average might be better for gauging the battleground state polls, since short-term averages are prone to the House Effects of the most recent poll. For instance, Rasmussen has polled recently in some states, but not others. In an unadjusted short-term average or LOESS trendline, whether Obama is closer in one state or another might just hinge on whether Rasmussen conducted a poll in the last week and a half or three weeks ago. If the race is stable, then a longer-term average can even out the bumps created by irregular pollsters without sacrificing accuracy. Indeed, much of the movement in the RCP average is attributable changing the composition of the average, not actual changes in the state of the race.
Most pollsters began their post-debate sample yesterday or will begin today, so we don't yet have a wave of post-debate polls to report on. But there are initial signs of limited utility, but that might just be worth reporting on to tide over those anxious to know what might lie next:
--The RAND American Life Panel showed Obama's 49.8 to 44.1 lead falling to 49.2 to 44.8, a shift of a net-1.5 points. That's not as decisive as the net-3.3 point movement after the first day of the DNC, but it's important to recall that RAND panelists can elect to submit their answers at any time over one week, so changes can occur at an irregular pace and we would not expect yesterday's results to fully reflect the consequences of the debate.
--A Reuters/Ipsos poll found Obama leading 48-43, down slightly from 47-41 prior to the debates. The survey was conducted entirely after the first debate, but it's an online poll with a small sample, so don't put too much stock in it.
--PPP found that Obama's lead was 1 or 2 points smaller in Virginia and Wisconsin than it was in pre-debate polls, but this conclusion was based on just one night of samples, so the margin of error is presumably quite large. For what it's worth, it surprises me that they report initial numbers from the first day of surveys.