Would Romney get a 3-point bounce or a 5-point bounce? That was the question heading into Romney’s acceptance speech last Thursday. While sparse polling over Labor Day added to the uncertainty, the question today has changed. Did Romney even get a bounce at all?
The national tracking polls show the race exactly where they’ve shown it since the spring and the few state polls are largely consistent with pre-convention polling. If you squint hard enough, there’s a case that Romney improved slightly; but his small gains are not clear enough to be distinguished from static.
For months, Gallup and Rasmussen have shown a tight race and a narrow Romney lead, respectively; today, Gallup and Rasmussen show a tight race and a narrow Romney lead, respectively. While Rasmussen saw a shift in Romney’s direction, it was only in comparison to the anomalous finding that Obama led by as much as 3 points prior to the RNC. In contrast, Gallup hasn’t shown any movement in Romney’s favor and, if anything, actually shows Obama gaining 2 points, with the president settling into a 1-point lead.
In my view, this afternoon’s CNN poll showing a tied race among likely voters is largely consistent with the view that Romney didn’t receive much of a bounce, if any. While a tied race among likely voters and a 7-point edge among registered voters represents a 2-point Romney gain from CNN’s prior survey, Obama’s 7 point edge among registered voters is exactly the same as it was a month ago and better than all but one of CNN’s polls since Romney secured the Republican nomination.
Some have cited Reuters/Ipsos as a sign of a Romney bounce, but a closer look shows that nearly all of Romney’s gains came prior to the heart of the RNC. Since Romney’s speech, the Reuters/Ipsos poll hasn’t shown any clear movement. And whatever the merits of the Reuters/Ipsos poll, it’s a new poll introduced immediately prior to the conventions, which makes it harder to judge if the shifts in the survey are typical static, or real divergence from a longer-term trend. That’s especially important with Romney’s gains coming prior to his own speech.
The state polling is convoluted as well, but here there is some evidence for a slight Romney bounce. It’s largely on this basis that Nate Silver concludes that Romney gained between two-and-a-half and three points. But there are strong reasons to question the use of these state polls in judging Romney’s bounce. First, the state polls conducted before the RNC are a little too old. All but one of the previous polls were conducted prior to Paul Ryan’s selection as Romney’s running mate, which appeared to reduce Obama’s margin by 1 or 2 points among registered voters. So while it’s defensible to rely on these PPP polls to argue that Romney has gained 2 or 3 points since late July or early August, it’s not clear that Romney gained 2 or 3 points as a result of the convention.
Second, even if the polls had been conducted immediately prior to the convention, the baseline for comparison isn’t as well established as the national trackers. After months of daily polls from Gallup and Rasmussen, we have an extremely good idea of the “normal” results for these surveys. But it’s harder to argue that PPP’s tied race in North Carolina, for instance, represents a “gain” for Romney. In Florida, North Carolina, and Colorado, PPP had only released one previous likely voter survey, while PPP hadn’t even conducted a single likely voter survey in Michigan. Attempts to include more data in the baseline just wind up including previous registered voter surveys, which might well have the effect of inflating Romney’s bounce.The only state poll with a decent baseline comes from SurveyUSA, which showed Romney jumping out to a 3-point lead in North Carolina, compared to a tied race a few weeks ago. But that SurveyUSA poll reflected Obama’s best showing in months, and their prior poll showed Romney up 5 points.
But perhaps this is the simplest test: in the absence of the Republican National Convention, would anyone interpret these polls as signs pointing toward the possibility of recent movement in Romney’s direction? I wouldn't, and Nate Silver’s model seems to agree: The recent movement in the “Now-cast” is indistinguishable from similar bumps and valleys over the last few months. It's telling that the reason we're arguing about a bounce isn't because of some new poll, but instead because there was a convention last week, so analysts presume that the slightest trends in the data point toward genuine movement that would otherwise be dismissed as static. In the absence of better data, the responsible position is that it's unclear whether Romney received a bounce. But the fact that Romney's bounce was difficult to discern, at best, suggests there might not have even been one.
As noted last week, there is only a weak relationship between the size of a post-convention bounce and Election Day performance. And historically the incumbent’s standing post-convention is more important than the challenger’s, so Obama’s position next week will be of greater importance than Romney’s position today. Romney’s weak bounce could either be a sign of deep weakness that Obama can exploit this week, or a sign that Obama won’t receive a bounce either in this highly polarized political climate. But the fact that Romney wasn’t able to take a lead following his convention should still be troubling for Boston. No modern candidate has won the presidency without seizing a lead following their own convention, and while that should not be interpreted as an ironclad rule, it’s still a sign that he could be in some trouble.