We got a new national poll today, and it no doubt damned near broke the Drudge Siren (I didn't bother to check, to be honest with you).
And, on a light polling day, it actually offers a teachable lesson about how to read polls. But, first, all the other nums on a relatively light Thursday:
GOP (PRESIDENTIAL) PRIMARY NUMBERS:
NATIONAL (Fox News): Romney 46, Paul 16, Santorum 15, Gingrich 15
NEW JERSEY (Quinnipiac): Romney 51, Santorum 14, Gingrich 9, Paul 7
PRESIDENTIAL GENERAL ELECTION TRIAL HEATS:
NATIONAL (Fox News): Romney d. Obama (46-44)
NATIONAL (Rasmussen Tracking): Obama tied with Romney (45-45)
NEW JERSEY (Quinnipiac): Obama d. Romney (49-40); Obama d. Santorum (51-36)
DOWNBALLOT POLLING:
NJ-SEN (Quinnipiac): Sen. Robert Menendez (D) 44, Joseph Kyrillos (R) 35
PA-SEN--R (McLaughlin for Tom Smith): Tom Smith 29, Sam Rohrer 14, Steven Welch 9, David Christian 7, Marc Scaringi 2
One quick thought about that Fox News poll after the jump.
So, Mitt Romney, having now secured the Republican nod, has moved ahead of Barack Obama in a look ahead to November, according to the Democratic-Republican polling consortium that is contracted to do surveys for Fox News.
Or ... is he? Herein lies a lesson in why sample composition matters.
One of the most reliable maxims in public opinion polling is: if the number is too good (or too bad) to be true, it probably is. Frequently, the reason a poll is either far too optimistic or too pessimistic when compared to its contemporaries is because the pool of respondents has shifted in a way that might be hard to replicate on Election Day.
This current poll by Fox News is an excellent case in point. The sample for this poll broke down thusly: D 40.6, R 39.9, I 19.5. While that is not far off what the sample composition of the 2010 midterm election exit polls looked like, finding a presidential point of comparison is pretty darned rare.
Looking back at the exit polling for the last four presidential elections, we get the following breakdowns:
2008: D 39, R 32, I 29
2004: D 37, R 37, I 26
2000: D 39, R 35, I 27
1996: D 39, R 35, I 26
So, on average, Democrats have had a net edge of roughly four points in elections. Fox News is assuming that the voting pool is going to contain far fewer independents, and far more Republicans than we have seen in a generation.
Is it possible? Possible? Sure. Republican election analysts spent the entirety of 2008 convinced that the polls were skewing Democratic, because the samples looked way too Democratic. Of course, then the D-R gap in the November exit polls was wider than it had been in over a generation.
And (pointing a finger at myself here), a number of Democrats looked at a number of polls in 2010, convinced that Democrats were being undersampled. As it happened, they were undersampled for a reason.
However, a legitimate reason to be skeptical of this Fox News poll is the fact that not only does it look different than previous election years, it also looked markedly different than the critical mass of polls conducted this year. It is pretty tough to find a poll in the last eight months or so where the sample had equivalent numbers of Democrats and Republicans.
Now, if going forward, we see a tightening in the partisan spread in samples from other pollsters, then this Fox News poll becomes prophetic rather than errant. But, until then, it is fair to be a bit skeptical about a poll that presumes an electorate that (a) doesn't look like many we've seen in recent presidential elections and (b) doesn't look like many of its polling contemporaries, either.