Tuesday’s polls obeyed by largely the same themes as Monday’s. There were signs of tightening toward Mitt Romney in the national polls, but they were harder to discern in state surveys, which remained strong for President Obama.
In this case, however, the data from national polls won out in the forecast, with Mr. Obama’s odds of winning the Electoral College declining to 84.7 percent from 85.7 percent on Monday.
A Wall Street Journal/NBC News poll published on Tuesday showed Mr. Obama with a 3-point lead among likely voters — down from a 5-point advantage in a poll they took just after the Democratic convention. There are a few caveats here: Mr. Obama’s lead was 5 points with third-party candidates included on the ballot. And his standing did not decline in the version of poll among registered voters. (Although that’s quite a mixed blessing for Mr. Obama because it implies that the gap between likely voters and registered voters has widened.) But The Wall Street Journal and NBC News conduct top-notch surveys, so this is a decent data point for Mr. Romney.
And an unambiguously strong poll for Mr. Romney came late on Tuesday, when National Journal published a survey showing a 47-47 tie among likely voters. The trendline is favorable for Mr. Romney in this poll as well, because its previous survey had shown Mr. Obama with a 7-point lead.
Mr. Romney also gained 2 points in the Rasmussen Reports national tracking poll and 1 point in the weekly national poll published by Public Policy Polling. A contradictory data point was from the Gallup national tracking poll, which had Mr. Obama gaining 2 points.
Mr. Romney got several poor state surveys on Tuesday, however. Most notable were a Virginia poll, from Roanoke College, which gave Mr. Obama an 8-point lead among likely voters, and a Nevada poll, from the firm We Ask America, which gave him a 10.5-point lead in that state.
It’s hard to know whether this reflects an increasing gap between the swing states and the rest of the country, or, rather, is more of a statistical quirk. Our research suggests that both state polls and national polls are useful in calibrating an overall estimate of the national popular vote. In some years, like 1996 and 2000, state polls were generally highly accurate, while the national polls had a bias on one or another direction.
There is a slightly clearer trend in our “now-cast,” our estimate of what would happen in an election held today, where Mr. Obama’s projected margin of victory in the popular vote has fallen by 0.7 percentage points over the past two days.
If the race has tightened by more than that amount, we’re going to have to start to see more evidence of it in the state-by-state polls. But perhaps Mr. Romney has a bit of a tailwind heading into the first debate on Wednesday.
Pennsylvania Dropping From List of Swing States
On Tuesday, a Pennsylvania judge delayed implementation of provisions of a new voter identification law, essentially neutering its impact for this November’s elections.
Based on academic studies of the effect of changes to voter ID laws, we had estimated that the law would have reduced turnout in Pennsylvania by 2 percent and reduced President Obama’s margin relative to Mitt Romney by about 1 percentage point. So while that isn’t a huge effect, it is now one less thing for Mr. Obama to worry about.
It is also one less reason for Mr. Romney to compete in Pennsylvania, where his standing was already very tenuous. As of Tuesday, the forecast model — which incorporated a modest adjustment for the voter ID law that has now been reversed — gives Mr. Obama a 97.7 percent of winning the state’s electoral votes on Nov. 6.
Part of the reason for the model’s certainly about the outcome in Pennsylvania is the stability of the data there. Mr. Obama has led in 34 consecutive polls of Pennsylvania, a streak that dates to early February.
This may not be a coincidence. Our research on individual-level voter demographics from the 2008 exit polls suggests that Pennsylvania is a relatively “inelastic” state, meaning that it is hard to move the numbers there. Mr. Obama can expect to rack up huge margins in the urban centers of Philadelphia and Pittsburgh, while Mr. Romney can expect to win almost all of the counties in the central part of the state (save for the occasional college town).
There are certainly some swing voters in the state, as in the Philadelphia suburbs. But elections in Pennsylvania generally come down to turning out the respective party bases — and it looks, simply, as if Mr. Obama’s base is a little larger than Mr. Romney’s there. Now he’ll have a bit less trouble getting his voters to cast their ballots.
Pennsylvania has steadily been falling on our list of tipping point states — those that are most likely to provide the decisive electoral vote in the event of a close election — and now ranks in ninth position.
Its rank is even lower, 13th, on a related list, our return on investment index, which also considers how expensive it is to compete in a state based on its population.
The return on investment index can also be interpreted as the likelihood that an individual voter in a state would determine the Electoral College winner with her vote. In Pennsylvania, that index is 0.3 — meaning that a vote there is almost 30 times less important than in neighboring Ohio, where the index is 8.4 instead.
According to National Journal, Mr. Romney’s campaign has not spent a single dollar on advertising in Pennsylvania since May 1; outside groups that support Mr. Romney ceased their advertising there in August.
It’s hard to know whether Mr. Romney’s decision to avoid spending money in Pennsylvania should now be thought of as being prescient.
On the one hand, it may be better to pull out of a state early than to do so late, after having spent a bunch of money there. On the other hand, Mr. Obama’s campaign ceased its advertising expenditures in Pennsylvania in July, no longer having the need to call Mr. Romney’s bluff — and Mr. Romney’s lackluster effort in Pennsylvania has put him in a
very defensive Electoral College posture throughout the campaign.