Obama Would Be Big Favorite With 'Fired Up' BaseBy NATE SILVER
There's one advantage that President Obama has that Mitt Romney probably doesn't. If he can get a good turnout from his base, he'll be the heavy favorite to win in November — even if Mr. Romney gets a strong turnout as well.
On average over the last five public surveys, 35 percent of registered voters identify themselves as Democrats and 30 percent as Republicans. That advantage is down somewhat for Democrats since 2008, but it is an advantage nevertheless.
In essentially every recent presidential election, however, the Democratic candidate has performed worse among actual Election Day voters than among the broader pool of registered voters. There is no reason to think that this year will be an exception. Recent surveys that compare likely-voter with registered-voter results suggest that there could be a turnout gap of around three percentage points favoring Mr. Romney. That's larger than the historical average, when it's been in the range of one or two points.
Our election forecasts build in a likely voter adjustment for this reason. If a pollster publishes both registered-voter and likely-voter results, we use the likely-voter version of their numbers. And if only a registered-voter version is available, we shift the numbers by two or three points toward Mr. Romney in order to make it equivalent to a likely-voter poll.
But what would happen if all those registered voters really did turn out?
I decided to run a version of our "now-cast" — our estimate of what would happen if the election were held today — on a registered-voter rather than likely-voter basis. (The "now-cast" is a little simpler than our Nov. 6 forecast, which also incorporates a convention bounce adjustment and measures of economic performance, and so the now-cast is a little simpler to interpret for purposes of measuring the effects of the likely-voter adjustment.)
This special, registered-voter version of the "now-cast" applies just the opposite of our usual process. If both registered-voter and likely-voter versions of a poll are available, I instruct the model to use the registered-voter numbers. And if there's only a likely-voter survey, the model shifts the numbers toward Mr. Obama by a couple of points as a proxy for a registered voter poll.
In the regular, likely-voter version of our "now-cast," Mr. Obama is estimated to have a 68 percent chance of winning the Electoral College in an election held today. But on the basis of registered voters, he would be a 91 percent favorite. Instead of being ahead in the popular vote by a hair over one point, he'd be expected to win by around four. And he'd be projected to win 322 electoral votes, rather than 291.
It's very unlikely, of course, that Democrats will turn out as high a percentage of their voters as Republicans do. The demographic groups that favor Democrats are just harder to get to the polls.
But if Mr. Obama can narrow Republicans' advantage in this area — reducing the gap between registered-voter and likely-voter polls to one or two points rather than three, he can get halfway there, and make Mr. Romney's task much harder.