UH, WAIT A SECOND – AT 9:01 A.M. ET: There's a Pew poll out that has the Obamans beside themselves with enthusiasm, but a second look may temper their glee. From John Podhoretz at Commentary:
A huge Pew Forum survey just released shows what President Obama desperately needs to see in the polling: He’s at 50 percent among registered voters, with Mitt Romney at 43 percent. This is very important for Obama and his campaign, because they know full well an incumbent president at 45 or 46 percent in the polls is far more likely to lose than win. In the history of public-opinion surveys, no first-term president has won a second term without polling at 50 percent or higher near election day.
But wait. What’s this? Scroll to the final page and you learn that of the 2373 registered voters, 837 identify as Democrats and only 636 as Republicans. That translates to a sample that’s 35% Democrat vs. 28% Republican (36% are said to be independents). That 7-point gap between Democrats and Republicans is the same 7-point gap that showed up in the 2008 exit polls, in which Democrats made up 39 percent of the electorate and Republicans only 32. Does anyone really believe that will be the case in 2012, which is certain to be a much closer election however it goes with a far more revved-up Republican electorate than in 2008?
In the midterm election of 2010, Democrats and Republicans each made up 35 percent of the electorate. The latest data, reported here by Bloomberg News, suggest voters who formerly identified as Democrats are leaving the party in substantial numbers in swing states to affiliate as independents—while Republicans are gaining registrants.
Not to mention the Pew poll is of registered voters only, not likely voters—and at this point in the race, it’s really time to start looking for likely voters only.
COMMENT: Well said. That's why it's so important to look at the internals of polls. We're getting a lot of surveys, and some of them are poorly done.
You might ask why anyone would publish a poll with questionable internal numbers. The answer is money. These polls cost a great deal to commission. There is a temptation for news organizations to publish them, as they generate buzz and readership, even if they're proved wrong in the long run.
July 13, 2012