Skip to main content
ABC News
Sept. 18: Obama’s Bounce Erodes in Two Tracking Polls

President Obama has been on the decline in the FiveThirtyEight forecast. On Tuesday, his chance of winning the Electoral College dropped to 72.9 percent, from 74.8 percent on Monday. Mr. Obama is now well off his peak in the forecast, 80.8 percent, which he reached on Sept. 12.

What’s causing the change? There are two major factors.

First, Mr. Obama has experienced an unmistakable decline in the two longest-running tracking polls, Gallup and Rasmussen Reports.

In the Gallup survey, Mr. Obama’s lead fell to one point over Mitt Romney on Tuesday — the same number that he started with before the Democratic convention, and down from a high of a seven-point lead last week.

Mr. Obama has similarly lost his bounce in the Rasmussen Reports tracking poll; in the version published on Tuesday, he was tied with Mr. Romney when voters who leaned toward a candidate were included in the tabulation, and trailed him by two points otherwise.

There are other tracking polls that show more ambiguous results. Mr. Obama’s bounce has held steady (or perhaps even improved slightly) in the online tracking poll published by the RAND Corporation, showing that he maintained a lead of about three points over the past week. The Ipsos online tracking poll had Mr. Obama with a four-point lead on Tuesday, down from a peak of seven points, although that poll has been volatile.

But the Gallup and Rasmussen Reports polls can have a lot of influence on the forecast at times when there is a potential turning point in the race. The trendline adjustment that the model calculates compares changes in the results produced by the same polling firms over time. Since the Gallup and Rasmussen Reports national tracking polls have been published almost every day since the spring, they represent highly important data series in this process.

The second major reason for the shift is that the model is designed to take a skeptical view of the polls conducted for a candidate who just held his convention. Typically, the polling bounce that a candidate receives from his convention does not evaporate immediately, but can persist for a couple of weeks. We are still close enough to the Democratic convention that this adjustment applies to Mr. Obama’s numbers in the polls.

The downward adjustment to Mr. Obama’s numbers will gradually fade out over the next week or so. Thus, if he holds his current position in the polls, he will begin to regain ground in the FiveThirtyEight forecast.

However, it is probably not a bad idea to take a slightly skeptical view toward Mr. Obama’s polls in the meantime. This is a tricky year for estimating convention bounces, with the two parties having held their conventions just one week apart, but if the model is reading the data wrongly, it will correct itself soon enough.

Mr. Obama’s state polls have also been fairly mixed over the last week. He has certainly gotten a few very strong ones, like a Washington Post poll on Tuesday that put him eight points ahead among likely voters in Virginia. But there have been other surveys to show rather poor results for him, like a Rasmussen Reports poll that put him two points behind in Colorado. In general, the impression conveyed by the state polls is a fairly modest bounce in the swing states.

Still, it might be acknowledged that this is among the more confusing periods that we’ve seen in the polls. How do you reconcile a poll showing Mr. Obama with an eight-point lead in Virginia with another putting him at a two-point deficit in Colorado?

Why, if his standing eroded in the Gallup and Rasmussen Reports tracking polls, did he also receive a rather strong poll from NBC News and The Wall Street Journal on Tuesday, which put him five points ahead nationally among likely voters?

These polls almost seem to inhabit different universes. As I’ll describe in a separate article, the methodological choices made by pollsters may have something to do with it; Mr. Obama’s bounce has been much more noticeable in polls that use live interviewers and that call cellphones.

But it does not do any good to pretend there is a consensus in the polls when there isn’t. Sometimes there is simply no alternative to remaining patient until one emerges. The downward trend for Mr. Obama in the Gallup and Rasmussen trackers is closest thing we have to a theme in the polls for the time being.

The Impact of ’47 Percent’

There has also been the introduction of a new event in the news cycle: the release of a video, taped secretly, showing impolitic comments that Mr. Romney made at a fund-raiser.

I begin from the premise that there is reason to be skeptical that Mr. Romney’s “47 percent” comments will have all that much effect on the polls. The news media often jumps the gun in declaring events to be “game changers” when they later prove to little effect on the numbers. Mr. Romney’s comments about Libya last week, for instance, were supposed to be very damaging to him, but if anything the numbers have moved toward him since then.

I do not mean to suggest that campaign controversies like this one never matter to voters. But I do think that reporters in Washington or New York, myself included, are not always the best judges of which are the exceptional cases. Furthermore, these judgments are likely to be influenced by the recent polls, meaning that analyses anticipating future reaction among voters may really be lagging indicators.

I have my own instincts about Mr. Romney’s remarks, which are roughly as follows: even if his outlook is a bit less negative than it seemed a week ago, he is nevertheless the underdog in the race, and not in a position where he can afford to alienate any voters who might allow him to climb to 50 percent of the vote. His coalition may already be drawn too narrowly, and this won’t help him with that.

But I’d place rather little value on my instincts, and rather more on the polls. We should know more about the state of the campaign a week from now than we do today.

Nate Silver founded and was the editor in chief of FiveThirtyEight.


Filed under