http://www.dailykos.com/story/2009/8/30/774450/-Are-All-Likely-Voters-Republicansby Steve Singiser
It's 2003 all over again.
In recent months, a number of polls released on the state of the 2009 and 2010 campaigns have made banner headlines for substantial Republican leads. In the fine print around paragraph five, however, it was revealed that among registered voters, the margin was considerably closer.
For polling junkies of a progressive persuasion, this was a neverending source of aggravation back in the day. That said, it was based on some fairly sound logic. Simple math teaches us that a proportion of registered voters rarely, if ever, vote. To the end, we also know that a couple of the most loyal voting blocs for the Democrats (young voters and African-Americans) are two of the voting blocs whose voter behavior are the most spotty.
That said, however, the likely voter screen has not always tilted reflexively Republican. In 2006, for example, the five CNN polls on the generic ballot test for Congress showed a slightly wider Democratic margin among likely voters (17.0%) than among registered voters (16.4%). The same was true with the four polls by Newsweek, where Democrats fared slightly better among likely voters (15.0%) than they did among registered voters (13.5%). In the interests of full disclosure, other pollsters maintained the traditional GOP advantage, even in a great Democratic year like 2006. Gallup, whose likely voter model has taken some criticism over the years, had the GOP faring about four points better with likely voters than they were among registered voters. In the most notable example of this effect, a mid-September Gallup poll had the Democrats leading by nine points among registered voters, but dead-even among likely voters.
This year, however, there have been some likely voter screens that have raised eyebrows. Take, for example, SurveyUSA's late July poll on the Virginia Governor's race. Cited by many as a sign of the ascendancy of the Republican Party (all three GOP statewide candidates had solid leads, ranging from 11-15 points), SurveyUSA gave away a critical hint about their 2009 likely voter screen. Amid the mountain of data (SurveyUSA has always been great about laying out all the stats), we see that the likely voters selected by SurveyUSA supported John McCain by a nine-point margin over Barack Obama in 2008. Of course, that's not close to what actually occurred in 2008--Barack Obama carried Virginia by a 53-46 margin.