|
Sometime in the near future, we can expect to see new polling on the state's congressional races from the UNH survey center. Before those numbers come out, I want to do a little refresher on why no one should take them very seriously.
Short version: UNH polls have serious sampling problems. Specifically (at a minimum), partisan identification fluctuates so wildly that looking at trend lines is misleading. The UNH poll in January is different from the UNH poll in April which may well turn out to be different from the UNH poll in July. Not garden-variety, all-polling-contains-some-error different. Apples and oranges different.
Long version: Back in May, I wrote a diary about the recent record of this poll, and specifically about these sampling problems. Rather than excerpt the entire diary, I'm bumping it to directly below this post.
But here I want to get back to the core problem, so I will repeat my excerpt from DavidNYC at noted national horserace blog Swing State Project.
Democrat: 23% (25%)
Republican: 32% (22%)
Independent: 38% (43%)
Unregistered: 7% (10%)
I think that just about explains things: The current sample has 10% more Republicans than the prior poll. While I'm aware that party ID among independents tends to shift with the political winds, there's been nothing to suggest that even so much as a gentle zephyr has been blowing in the GOP's direction in New Hampshire over the last twelve weeks. What's more, if Bush's favorability shows an increase comparable to Bass's, that would make him more popular in blue New Hampshire than in the nation as a whole - and more popular than he's been in the state since January.
So, what's been going on with the sample in recent UNH polls? Let's look a take at the first district, where Carol Shea-Porter's popularity apparently took a pretty striking nosedive between January and April. At least, in Andy Smith's sample, which it must be understood is distinct from reality.
With first-district samples of 267 in January and 251 in April, here's how partisan identification broke down:
1/08 4/08
Democrat 37% 26%
Republican 27% 26%
Undeclared 28% 42%
Not registered 8% 6%
Basically, we're to believe that Democrats took an 11-point tumble while Republicans held just about steady and undeclareds skyrocketed. I can believe that in the wake of the primary, a few people who had voted for one party reverted to identifying as undeclared. But this sample strains credulity to a truly insulting degree, relying on reporters not to read closely enough to see what's going on (whether it be grounded in sloppiness or intentional manipulation).
I'm on the edge of my seat waiting to find out the partisan distribution of the upcoming poll. Anyone got any bets?
Update: A quick gander at the numbers of likely voters in the presidential poll released today showing Obama up by just 3 points, in contrast to other recent polling, shows this partisan breakdown:
Democrats 25%
Republicans 28%
Undeclared 40%
Presumably this is just a subset of the sample being used for the downballot races, which won't be screened for likely voters, but if it's a sign of things to come, color me excited.
And a reminder: 50,000 more people voted Democratic in the NH presidential primary than voted Republican this year.
|
Blue Hampshire RSS