« MT-Sen: Deeper and Deeper | Main | MO-Sen: No Shame »

Tuesday, August 01, 2006

NH-02: New Poll Shows Wider Bass Lead

Posted by DavidNYC

The University of New Hampshire just released a new poll (PDF) on the race in NH-02 (likely voters, May in parens):

Hodes (D): 25 (35)
Bass (R-inc.): 53 (42)
Undecided: 22 (22)
(MoE: ±6.2%)

A few things to note about this poll: Like the last one, it has a huge MoE. Unlike the last survey, though, this poll isn't backed up by any others (yet). A Hodes staffer told me that back in May, the campaign's own polling showed the exact same margin that UNH's did. Hopefully, there will be another poll in the field soon - but this time, I'm betting it won't confirm these newest UNH results. Why do I say that?

First, what's happened in NH-02 over the last three months that could possibly have caused such a huge swing? Do you have any clue? I certainly don't - and that's because the answer is "nothing." This race has barely heated up. Neither candidate is on the air in any meaningful way. The campaign hasn't even experienced much if any controversy yet (Bass's immoral voting record nonwithstanding). And this is born out by the fact that Hodes' name rec numbers are essentially unchanged from the last time out.

Bass, however, saw a big jump (nine points) in his favorability score. Though the PDF linked above doesn't give an exact figure, it does say that Bush's favorability also increased from his May low of 30%. If Bush experienced a jump anything like Bass's (and I wouldn't be surprised if they were similar), you've gotta ask: Why? Is Bush (and by extension Bass) all of a sudden more popular in New Hampshire now than he was three months ago?

I'd have a hard time believing that. But I think I have the answer right here. Check out the partisan breakdown of the survey sample, with the current poll listed first and the May poll in parens:

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.

You may be tempted to dismiss this as so much spin from a Hodes partisan. That I am a big supporter of his I would never deny - it's been plain for everyone to see for months. But again I say, this latest survey from UNH isn't backed up by any other polls; it has a huge MoE; and there is a serious issue with the dramatic, unexplained change in sample demographics. If Charlie Bass wants to sleep soundly tonight because of this poll, that's his mistake to make. But Hodes supporters shouldn't be discouraged and should view this poll with the skepticism it merits.

Posted at 05:14 PM in 2006 Elections - House, New Hampshire | Technorati

Trackback Pings

TrackBack URL for this entry:


Also bizarre: All of Bass's new support in this poll came from former Hodes supporters. The number of undecideds stayed the same.

So what happened? Maybe Hodes killed a kitten with a hammer at a campaign event (so only his supporters knew about it)?

Posted by: bosdcla14 [TypeKey Profile Page] at August 2, 2006 01:11 AM | Permalink | Edit Comment | Delete Comment

Admittedly I, like David, really want this to be wrong. But I, like David, also think, on looking at the data here, that it is genuinely flawed. What on earth would have Bass's favorability at a 2 or 3 year high right now? And the first district's Jeb Bradley similarly exceeding his recent performance?

Posted by: MissLaura [TypeKey Profile Page] at August 2, 2006 01:27 AM | Permalink | Edit Comment | Delete Comment

There are credible technical explanations for this result that don't require agonizing over the result. First, note the high margin of error (6.2%), which occurred because the poll only sampled 261 respondants. Assuming the other poll also had an MOE of 6.2%, that is a potential 12.4 point swing, which could be in line with what happened. The "true" parameter value could be Hodes ~30% and Bass ~47-8% and lie within the MOE of both polls.

But, even discounting the small sample size, the possibility exists that one (or both) of the two polls is flat out wrong. What nobody ever mentiones about the margin of error is that it is (usually for these polls) the MOE for the 95% confidence interval. What that means is that the true parameter value will lie within the MOE 95% of the time, which also means that one time in twenty it doesn't. Assuming independent sampling, this means that there is a 9% chance that at least one of these polls is incorrect.

This is the case for all polls and not just this one, by the way. Five percent of them are wrong, in the sense that the true value lies outside the margin of error.

(This concludes our math lesson for today...)

Posted by: Jay [TypeKey Profile Page] at August 2, 2006 06:53 AM | Permalink | Edit Comment | Delete Comment

Thanks, Jay, for saying things I know but don't want to only be dragging out on the polls I don't like...

Posted by: MissLaura [TypeKey Profile Page] at August 2, 2006 11:44 AM | Permalink | Edit Comment | Delete Comment

Thing is, there's no way to know if this poll is purely flawed without a third poll (or an actual election) to compare it with. I also don't think it's just "plain wrong" - I do think the shift in the sample really explains most if not all of the weirdness.

As Bos points out, the undecideds didn't budge an inch. Are we really to believe that a third of Hodes' supporters just jumped ship for almost no discernable reason? It's possible, sure, but not likely.

Posted by: DavidNYC [TypeKey Profile Page] at August 2, 2006 11:54 AM | Permalink | Edit Comment | Delete Comment

We can all agree that the difference in these two polls cannot be explained by anything happening in the race. Something’s flawed. The question is which poll should we believe? The answer: neither of them. The UNH Survey Center makes no attempt to screen respondents by party affiliation, geography, gender, age, etc. In this age, such sloppy methodology is inexcusable. Increased cell phone use, caller ID, and increased annoyance with telemarketers among other factors have spelled the end of reliable random digit dial surveys. Pollsters must create a survey profile that generally fits the demographics of the population they’re sampling. Like a broken watch that tells the correct time twice a day, a UNH poll may occasionally be accurate but we have no way of knowing when. Best just to ignore them altogether.

UNH can afford to be sloppy because no one who makes decisions based on polling data actually pays them. Candidates and PACs and 527’s and campaign committees that use polls to target resources would never rely on numbers that swing so wildly with little or no explanation. And trust me, this race has been polled to death. There are polls out there that have a high degree of confidence that could tell us exactly what is going on with this race. We may not see them, but they’re there, and just like the invisible black hole, we can infer some things about their properties from their visible effects.

So far, the effects imply that they’re not good for Paul Hodes. Watch what the pros who have access to reliable poll numbers are doing. Example: right after Hodes posted the most impressive fundraising quarter in NH history, National Journal dropped this race from their top-50 watch list. Why? Because of all the little tips and tidbits and crumbs that the Journal reporters pick up from hanging out in Capitol Hill bars with party operatives, including what their latest internal polling is showing them. The DCCC Red to Blue program features their 34 most promising challengers…but not Paul Hodes. Why? Because the DCCC thinks that at least 34 of their challengers have a better shot at scoring a win than Hodes based on their internal polling. So when considering the meaning of polls, listen for the dog which in this case is not barking.

I’m very confident that we’re going to take control of the House in the fall, and possibly even the Senate. But for all my friends in New Hampshire, I’m sorry but I’m afraid that it just isn’t terribly likely that Hodes will be part of the new majority.

Posted by: pollphreak [TypeKey Profile Page] at August 3, 2006 03:12 PM | Permalink | Edit Comment | Delete Comment

Good analysis of how sampling inconsistencies can impact polls. UNH polls are notorious for this, and I haven’t put much faith in them since they put Katrina Swett within 6 points of Bass and closing fast with a week to go in ’02 (UNH also had Shaheen beating Sununu). I had done a lot of volunteer work on Dick Swett’s ’94 campaign and was looking for Katrina to exact revenge on behalf of all of us who had grown tired of Bass’s cloying toadiness, but she (and Shaheen) lost in a result that had zero correlation to UNH’s polling.
Taken together, these two polls make no sense. Clearly, one or both don’t reflect reality. I’d like to think that the first does and the second doesn’t, and you make a fairly reassuring argument in support of that. However, something nags…
In the April poll, UNH samples 101 self-identified D’s, 30 I’s, and 65 R’s. In this poll, that ratio is 92/30/83. Fewer D’s, same amount of I’s, and many more R’s. Clearly, a great deal of Bass’s improvement over the summer can be attributed to nothing more than the fact that more R’s answered the phone, as you’ve explained very well. So which sample is closer to reality? The one thing that jumps out at me is that the April sample has more self-identified D’s than R’s and I’s combined. Yes, NH-02 is Kerry Country, but that sort of sampling would be more typical of Massachusetts, not NH. The August sampling, with somewhat more D’s than R’s, corresponds more closely with the leaning-Dem nature of this district.
Beyond that, there are 3 things that worry me about this poll. First, even ignoring the sampling boost that UNH gave Bass, it’s clear that he’s somehow found a slight breeze to ride. In April, he was getting 23% of D’s, 39% of I’s, and 73% of R’s, Now, he’s getting 27% of D’s (a 4-point gain), 48% of I’s (a 9-point gain), and 84% of R’s (an 11-point gain). He’s improved across all subgroups, and a little math will reveal that even with the April sample, Bass improves to 46-28, an 18-point lead. Something’s out there, even if we can’t see it, and it’s helping Bass and hurting Hodes. Maybe it’s something as simple as Bass sticking it to Bush on stem cell research, which might explain why Bass's approval rating went up even as Gregg's and Sununu's went down. Who knows?
Second, and more disturbingly, there is no way to escape the simple fact that after 3 years of running virtually non-stop, Hodes is known by fewer people than Joe Schlobotnick. When the April poll was taken, there were 6 months until the election. Now we’re down to 3, and Hodes hasn’t moved an inch. Jim Craig in NH-01, who’s never run for anything more than State Representative, has 35% recognition. How can a guy that got 125,000 people to check off his name just a few short years ago have virtually no name recognition? If Bass were unpopular, that might not be a problem, but Bass is now as popular in his district as Judd Gregg (apples to apples here, since it's the same sample). Just imagine if Hodes had the opportunity to take his case individually to each voter, which would be the ideal situation. Well, it won’t do any good if the man is a complete stranger, especially when 58% are already favorably disposed to Bass. Who’s going to listen to him?
Finally, I was disappointed when I read Hodes’ campaign manager’s quote in yesterday’s Boston Globe – “Frankly, I don’t know that much has changed in the last three months.” The biggest daily in New England calls for spin, and the best you can do is the verbal equivalent of deer in the headlights?
I hope you’re right, but my faith is beginning to quiver…

Posted by: Aquagirl [TypeKey Profile Page] at August 3, 2006 03:43 PM | Permalink | Edit Comment | Delete Comment

That Globe story was pretty short. We don't know how much the spokesman actually told the reporter, what clever spin s/he might have ignored. Coming up with good spin is one thing - but making sure the tradmed prints it... well, I don't really think there's any way you can ensure that.

Posted by: DavidNYC [TypeKey Profile Page] at August 3, 2006 04:07 PM | Permalink | Edit Comment | Delete Comment