Why did President Obama get a bounce in the polls following the Democratic convention?
Part of it may simply be one of the functions that conventions have long served: to motivate voters who are generally loyal to the party, but who had been paying only marginal attention to the race.
At FiveThirtyEight, we measure the “enthusiasm gap” between the parties in a particular way. Specifically, we watch for cases in which pollsters report results among both the broader universe of registered voters, and the narrower one of likely voters. If one candidate’s supporters are more likely to vote, that candidate will do better in the likely-voter polls than the registered-voter ones.
Typically, the candidate who benefits from the comparison is the Republican. Republican voters tend to be older, whiter and wealthier, all characteristics that correlate with a greater propensity to vote.
In midterm election years, when overall participation is lower, the enthusiasm gap can be correspondingly greater. In 2010, for instance, we calculated it to be about six percentage points in favor of Republicans. For the most part, that advantage proved to be predictive of the actual results on Election Day. Democrats outperformed their likely-voter polls by a point or two on average in Senate and governors races, and underperformed them by a point or so in races for the House of Representatives. But in both cases the likely-voter surveys did a much more accurate job of capturing the Republican wave than polls of registered voters.
There have also been midterm election years, like 2006, when there was almost no likely-voter gap, or when it even favored Democrats by a point or so. With lower turnout, there is the possibility of a wider range of outcomes based on the particulars of the political cycle.
In presidential election years, however, when the turnout is higher, the gap has been more consistent: always favoring Republicans, but always narrowly.
Specifically, based on our research, the likely-voter gap has favored Republicans by 1.5 percentage points on average in presidential elections years since 1988, with little variance from cycle to cycle.
This year, however, there was evidence that the gap was larger, averaging close to three percentage points in favor of Mitt Romney.
Our forecasts build in an adjustment for this factor, adding points to Mr. Romney’s column in polls of registered voters, and subtracting them from Mr. Obama’s, in order to make them comparable to likely-voter polls.
Polls conducted since the Democratic convention in Charlotte, N.C., however, suggest that the gap may have narrowed. Although the difference varies from polling firm to polling firm, on the average it has reverted back to the historical average of between one and two percentage points.
In the table below, I’ve presented a comparison of the results of the polling firms that have released both sets of numbers since Charlotte. (I count the telephone and online polls by Ipsos as separate surveys, since they use distinct methodologies.)
There have been outlying results on either side. An ABC News/Washington Post poll released this week showed Mr. Obama leading Mr. Romney by six points among registered voters, but by just one point among likely voters — a five-point enthusiasm gap favoring Mr. Romney.
Conversely, a Fox News poll, released on Wednesday, showed a reverse enthusiasm gap, with Mr. Obama actually doing one point better in their likely-voter poll than in their survey of registered voters.
On average among the five polling results, however, the gap has been 1.4 percentage points favoring Mr. Romney — almost exactly in line with the historical average.
Will Democrats maintain this narrower gap? Or is it a temporary effect from the conventions?
It’s hard to know for sure, of course. Until we see more evidence, the forecast model is still assuming that the gap will be slightly wider than average, perhaps around two percentage points in Mr. Romney’s favor.
But in general, there’s more reason to believe that a shift is permanent rather than temporary when it brings the numbers closer into line with historical norms, as it has in this case.
In addition, there is a question over whether likely-voter models applied before the party conventions are all that informative, since voter enthusiasm can wax and wane over the course of the cycle.
Our research suggests that likely-voter polls become more accurate than registered-voter polls beginning around Labor Day. But it is more debatable which set of polls do a better job of predicting the result before that date.
If there were no likely-voter gap at all, and all registered voters turned out to vote, Mr. Obama would have a very clear advantage in the race, perhaps being favored to win by five or six percentage points.
That case is probably wishful thinking for Democrats. Even in years like 1996 when Republicans had a middling candidate, their voters were nevertheless quite loyal on Election Day.
But if Democrats can keep the enthusiasm gap to its traditional value of about one and a half percentage points, rather than three, that could make quite a lot of difference. Before the conventions our forecast model had generally shown Mr. Obama as a favorite by about two and a half percentage points; now it projects him to win by about four. That’s consistent with what you’d get if the enthusiasm gap had been narrowed by a point or two.
We’ll need to wait for another week or so to see how much of Mr. Obama’s convention bounce carries forward, but we’re getting to the point where polls do provide for a pretty reliable gauge of the Election Day outcome. A four-point deficit in the polls can certainly be overcome, but it isn’t that easy, especially in a year like this one with few undecided voters.
Incidentally, the fact that Republicans already were “fired up” before the Republican convention may explain why Mr. Romney got a small convention bounce. Their convention got middling television ratings, with the exception of the Fox News coverage, where there was little decline from 2008.
Not every viewer of Fox News is a partisan Republican, and the network generally does a compelling job with its political news coverage. But its viewers lean toward being both conservative and highly politically en
gaged, just the sort of voters who were already showing high levels of enthusiasm for voting Republican this election. If few people watched Mr. Romney’s convention beyond that group, then perhaps there wasn’t much potential for his numbers to grow, almost without regard to how the convention was staged.
Wednesday’s polling data was a little choppy. Mr. Obama lost two points in the Rasmussen Reports tracking poll, which now shows him ahead of Mr. Romney by just one point. That’s a decent result for Mr. Obama relative to their usual take on the race, since the poll is somewhat Republican-leaning, but also suggests a clear decline from his peak in the poll a couple of days ago.
On the other hand, Mr. Obama extended his lead to seven points in the Gallup national tracking poll, up from six on Tuesday.
I’m not sure that it’s worth reading too much into either trend. It is likely that we’re going to see a proliferation of polls in the forthcoming days; we’ve already seen this to some extent at the state level, although many of the polls have been of noncompetitive states. We will be trying to determine what the consensus of the evidence holds, rather than making guesses based on the fluctuations in tracking polls — a much preferable circumstance, on balance.
The expectation embedded in the forecast model is that Mr. Obama’s convention bounce is about fully priced into the polls by now, and that his numbers probably will be facing some downward pressure over the next week or two. However, it also seems unlikely that his numbers will decline all at once, barring intervening news events.
If Mr. Obama’s polls do not show any signs of decline at all over the next week or two, that would qualify as good news for him — and, obviously, it would be bad news for him if the race reverts back to its preconvention norms almost immediately. But for now, our forecast has stabilized a bit, with Mr. Obama holding in the range of about a four-point lead in the popular vote and an 80 percent chance of winning the Electoral College.
Of course, there has been a significant news event, with the killing of the United States ambassador in Libya and three members of his staff. I’ll have a brief comment about that in a separate post.