Oct. 5: Day After Debate, Strong Swing State Polls for Romney

By NATE SILVER

On Friday, Mitt Romney had his best day in state-level polling since at least the party conventions, something that very probably reflects improvement in his standing following the presidential debate in Denver on Wednesday.

Two automated polling firms, Rasmussen Reports and We Ask America, released polls in Ohio, Florida and Virginia on Friday. All of these polls were conducted on Thursday, the day after the Denver debate.

In the Rasmussen Reports polls, Mr. Romney trailed Barack Obama by 1 point in Ohio. But he led him by 1 point in Virginia and by 2 points in Florida.

These are very good numbers for Mr. Romney as compared with the ones we were seeing recently, although part of that is because Rasmussen has shown more favorable numbers for him in these states throughout the year. As compared with Rasmussen Reports' previous polls of the same states, the margin in Ohio held steady, but Mr. Romney gained 2 points i n Virginia and 4 in Florida, for an average gain of 2 points among the three states.

The We Ask America polls suggested that Mr. Romney made much larger gains. He led in all three states in its polls - and gained an average of 7 points from We Ask America's prior polling of the same states.

Mr. Romney's bounce has been less apparent in national tracking polls so far. The Rasmussen Reports national tracking poll held steady, showing a 2-point lead for Mr. Obama. Mr. Obama actually gained 1 point in the Gallup national tracking poll, however, and about 1.5 percentage points in the online tracking poll conducted by the RAND Corporation.

Another online tracking poll, from Ipsos, suggested a strong trend for Mr. Romney, however.

The Ipsos polls are confusing because it has released polls covering various time intervals in the past few days, but they tell a potentially interesting story if you work through them carefully.

In a poll of about 500 voters that Ipsos conducted immediately after the debate, late Wednesday night and early Thursday morning, Mr. Obama still led by 5 points. However, Mr. Obama's lead was just 2 points in a poll Ipsos released Friday, which included interviews from Monday night (before the debate) through Friday morning.

The inference I make from these Ipsos polls is that Mr. Romney must have polled very well in the most recent interviews it conducted, late Thursday and early Friday morning, quite possibly leading Mr. Obama, in order to have made up so much ground.

It may have been that Mr. Obama's problems were growing worse throughout the day on Thursday as criticism of his debate performance was amplified. That would also help to explain Mr. Romney's very strong performance in the We Ask America polls on Thursday.

Of course, the Rasmussen Reports polls were also conducted on Thursday, and Mr. Romney made more modest gains there. It is harder to make inferences from the RAND Cor poration and Gallup tracking polls, because they use lengthy seven-day field periods and conducted only one full day of interviewing after the debates. But the fact that Mr. Romney actually lost ground in those polls is not very consistent with the 7-point bounce that the We Ask America polls imply, even considering that most of its data was predebate.

Polling trends can sometimes be odd in reaction to news events. One factor is that supporters of a particular candidate may be more enthusiastic, and more inclined to respond to surveys, after he gets a favorable development in the news cycle. The methodology that a pollster applies, particularly its likely voter model, may amplify or mitigate these effects.

The We Ask America polls, for instance, had a lot of voters who identified as Republican in their samples. I don't think there's necessarily anything wrong with that - I'd rather that pollsters give the most honest snapshot of what they were finding in the fiel d on the day that they conducted their interviews. Part of the reason that critiques about “oversampling” Democrats or Republicans are misguided is because the party identification breakouts themselves provide interesting information. It's logical to conclude, for instance, that Republicans may have been especially likely to respond to pollsters after Mr. Romney's strong debate performance. That would also explain why Mr. Romney's bounce was more modest in the Rasmussen Reports polls, as they weight their samples by party identification (a poor methodological choice, in my view), which may dampen the enthusiasm effect.

There is another type of polling bias, however, which is potentially more relevant when there is polling after a major development in the news cycle. Namely, polls are very probably biased toward high-information voters who take more interest in the news and are more likely to respond to political surveys. This issue may be more profound in automated polls, which have especially low response rates - often only 3 or 4 percent of the people they call respond to them.

So it's hard to distinguish a genuine shift toward Mr. Romney, from a real but potentially temporary shift based on changes in voter enthusiasm, from an artificial change caused by a bias toward heavy news consumers.

But now there's another complication: the government reported a strong jobs report on Friday, which changed the tone of the news cycle. To the extent that the polls reflected people's reaction to the news coverage of the debate as much as the debate itself, the jobs report could blunt some of Mr. Romney's momentum if the tenor of news coverage changes.

The FiveThirtyEight forecast did show a clear shift toward Mr. Romney on Friday, giving him a 15.1 percent chance of winning the Electoral College - up from 12.9 percent on Thursday.

My subjective view is that, despite the somewhat mixed messages that the polls gave about th e magnitude of Mr. Romney's bounce, this is still too conservative. The forecast model is pretty “smart” about distinguishing random movements in the polls from real ones, and so can be fairly conservative in interpreting the data. However, it does not have the advantage of knowing that the shifts may have come for a good reason - in this case, Mr. Romney's strong performance in the debate.

So I would bet on Mr. Romney right now given the odds the model offers - but I'd have done so more confidently before the morning's jobs report.

It's going to take a few more days for the forecast model to catch up to the news, and I don't think there's any alternative but to keep an open mind about the polls for right now.