October 25, 2012
Oct. 24: In Polls, Romney’s Momentum Seems to Have StoppedBy NATE SILVER
The term "momentum" is used very often in political coverage - but reporters and analysts seldom pause to consider what it means.
Let me tell you what I think it ought to mean: that a body in motion tends to stay in motion. That is, it ought to imply that a candidate is gaining ground in the race - and, furthermore, that he is likely to continue to gain ground.
As a thesis or prediction about how polls behave, this notion is a bit dubious, especially in general elections. In races for the United States Senate, for instance, my research suggests that a candidate who gains ground in the polls in one month (say, from August to September) is no more likely to do so during the next one (from September to October). If anything, the candidate who gains ground in the polls in one month may be more likely to lose ground the next time around.
(Where might there be clearer evidence for momentum, as I've defined it? In primaries, especially when there are multiple candidates in the race and voters are behaving tactically in choosing among them. But there is little evidence of it in general elections.)
The way the term "momentum" is applied in practice by the news media, however, it usually refers only to the first part of the clause - meaning simply that a candidate has been gaining ground in the polls, whether or not he might continue to do so. (I've used this phrasing plenty of times myself, so I have no real basis to complain about it.)
But there are other times when the notion of momentum is behind the curve - as it probably now is if applied to Mitt Romney's polling.
Mr. Romney clearly gained ground in the polls in the week or two after the Denver debate, putting himself in a much stronger overall position in the race. However, it seems that he is no longer doing so.
Take Wednesday's national tracking polls, for instance. (There are now eight of them published each day.) Mr. Romney gained ground in just one of the polls, an online poll conducted for Reuters by the polling organization Ipsos. He lost ground in five others, with President Obama improving his standing instead in those surveys. On average, Mr. Obama gained about one point between the eight polls.
This is the closest that we've come in a week or so to one candidate clearly having "won" the day in the tracking polls - and it was Mr. Obama.
The trend could also be spurious. If the race is steady, it's not that hard for one candidate to gain ground in five of six polls (excluding the two that showed no movement on Wednesday) just based on chance alone.
What isn't very likely, however, is for one candidate to lose ground in five of six polls if the race is still moving toward him. In other words, we can debate whether Mr. Obama has a pinch of momentum or whether the race is instead flat, but it's improbable that Mr. Romney would have a day like this if he still had momentum.
The FiveThirtyEight model looks at a broader array of polls - including state polls - in order to gauge the overall trend in the race.
Our "now-cast" also finds a slightly favorable trend for Mr. Obama over the course of the past 10 days or so. Mr. Romney's position peaked in the "now-cast" on Friday, Oct. 12, at which point it estimated a virtual tie in the popular vote (Mr. Obama was the projected "winner" by 0.3 percentage points). As of Wednesday, however, Mr. Obama was 1.4 percentage points ahead in the "now-cast," meaning that he may have regained about 1 percentage point of the 4 points or so that he lost after Denver. Mr. Obama's chances of winning the Electoral College were up in the FiveThirtyEight forecast to 71 percent on Wednesday from 68.1 percent on Tuesday.
It's not yet clear how much of this, if any, has to do with the final presidential debate in Florida this Monday, which instant polls regarded Mr. Obama as having won. Instead, it's been more of a slow and unsteady trajectory for him, with Mr. Obama often taking two steps forward but then one step back. It's also not out of the question that the apparent trend just represents statistical noise.
At the same time, there is more reason to take a potential change in the polls seriously if it is precipitated by a news event like the debate. The tracking polls that were released on Wednesday contained only one full day of interviews that postdated the Florida debate. If the debate moved the needle toward Mr. Obama, it should become more apparent in the coming days.
The battleground state polls that came in on Wednesday were generally very close to our model's current projections. For instance, there were three Ohio polls published on Wednesday; one showed a tied race there, while the other two showed Mr. Obama ahead by margins of two and five points.That's pretty much what you'd expect to see out of a trio of Ohio polls if Mr. Obama's lead there were about two points, which is where our model now has it.
Some of the polls, especially the Time magazine poll which had Mr. Obama five points ahead in Ohio, seemed to set off a lot of discussion on Twitter, as though people were surprised that Mr. Obama still held the lead there.
But these polls are really nothing new. Since the Denver debate, Mr. Obama has held the lead in 16 Ohio polls against 6 for Mr. Romney. In Nevada, Mr. Obama has had the lead in 11 polls, to Mr. Romney's 1. Mr. Obama has led in all polls of Wisconsin since the Denver debate, and he has had five poll leads in Iowa to one for Mr. Romney.
Part of the confusion (and part of the reason behind the perception that Mr. Romney is still gaining ground in the race) may be because of the headlines that accompany polls.
We're still getting some polls trickling in where the most recent comparison is to a poll conducted before the Denver debate. We should expect Mr. Romney to gain ground relative to a poll conducted before Denver. (Mr. Romney may have lost a point or so off his bounce, but he has clearly not lost all of it). But it isn't news when he does; Mr. Romney's Denver gains had long ago become apparent, and priced into the various polling averages and forecast models.
The question, rather, is whether Mr. Romney is gaining ground relative to the post-Denver polls - or if, as Wednesday's polls seemed to imply, the race instead may have ticked back slightly toward Mr. Obama.
