On Smith’s Cloud Hampshire, Sununu is Flying High

In a UNH poll conducted between 4/13 and 4/22 with about 500 sample size (in other words, likely the same operation done here and here), John E. Sununu is shown to be besting Paul Hodes 46%-41%, with 11% undecided and 2% for someone else.

That’s a fascinating result for the Sununu Dynasty to chew on as they decree whether or no The Younger shall be the nominee – on Cloud Hampshire.

When you add up the 111 registered Democrats, 137 registered Republicans, and 185 registered Undeclareds who were polled on this match-up, you get a remarkable picture of this fantasy state.  Let’s compare it with the most recent data – November 2008 – from a real place that has elections called New Hampshire.

On Cloud Hampshire, Democrats make up only 25.63% of the electorate, Republicans dominate with 31.63% of the share, and the remaining 42.72% are Undeclared.

In New Hampshire, Democrats are actually the party in the driver’s seat, with 29.44% of the people’s choice, followed closely by Republicans with 29.26%. Undeclareds round it out with 41.27%

In Cloud-Cuckoo-Land, the Sprinter has a chance.  Here on earth, with 2% less Republicans and almost 4% more Democrats, it’s harder to say.

Also on Cloud Hampshire, young voters (ages 18-34) are picking Hodes over Sununu by 9 points (47%-38%) – wow!  What a pity that in the clouds, people aged 18-34 only make up 9.75% of the voting electorate.  Now I know that young people are the hardest to get to the polls, but sheesh, Cloud Hampshire, you have a serious civics problem there. Especially since in a real place called the United States of America, exit polls show that 52% of voters aged 18-29 turned out for the election. I’m guessing that translates into something a little higher than 9.75%. And in fact, a little bird tells me, voters age 18-29 made up 17% of the electorate in NH-Sen for 2008.  I wonder what 18-34 would be?

Adding: And what elwood said about CDs 1 and 2.

, , , ,

  • elwood

    Smith always overpolls Republicans. In this example he also overpolls CD-1, where Hodes has never run: 236 to 216.

    • Dean Barker

      An obvious point I should have looked at.

      But, as you say, I got tired.

      • elwood

        when I keep pointing out these flaws – tiresome for the poor reader.

  • susanthe

    does Andy Smith have a job, and why in the world does anyone give any credence to UNH polls, given what a lousy, dishonest pollster he is?

    • DavidNYC

      And we’re just dirty fucking hippies.

  • Andrew Sylvia

    Here it is. It’s 9 months old, but it’s the most recent addition on their website.

    Andy Smith’s poll has 19% more Republicans than Democrats where there are 2% more registered Republicans than Democrats on that list, almost a dead heat.

    A random sample is fine as long as you remember that it is random and not necessarily indicative of reality.

    The writeup at the beginning of the UNH poll doesn’t do that particularly well.

  • historybuff21

    OK a poll with 18% more republicans is valid to those idiots? I mean I think its time to get out of the cloud and rethink the organization of the party, or better yet lets keep it this way. If it stays this way complete disaster might hit at the right time. Sununu I dont think you are really flying high look at reality… you are in a falling party, based on numbers, and the last I heard you lost an election, not exactly the cloud i would want to be on.

    • hannah

      This observation is intended to resurrect the word “slobs” as a designation of lazy buggers who just make messes.

  • Mike Biundo

    It seems both Republicans and Democrats share the common distaste for pollsters when the numbers don’t go their way. It is almost like bipartisanship, almost. Lol.

    • elwood

      We’ve been pointing out Smith’s lousy methodology for a couple of years now, whether a particular poll seems favorable or unfavorable.

      Feel free to search the site. This is not a brand new disenchantment with his methods.

  • xteeth

    I think I would be careful here. As much as I agree with you about  the Smith bias born out again and again when the results are just  plain wrong, normalizing a sample is such an obvious thing that I  would check again to see how he did it. Maybe he didn’t. I don’t  know, but all samples are flawed in some way and in general,  corrective action is taken to mitigate those flaws. The way that this  is done may influence the result but it is generally the first thing  anyone checks. I don’t think he is dumb or incompetant in fact. He  just gets wrong answers. I desperately don’t want us to be wrong in  those attacks we make.

    • susanthe

      He’s a pal of Jeb Bradley’s. Every single poll he ever did about CSP and Bradley was way off. I guess I don’t have your kind, trusting, nature. I don’t think it was accidental.  

      • xteeth

        I just remember going off into analysis of variance routines where you could shuffle this here and that there and suddenly everything was clear if that was what you started out believing. The big joke was, in this very mathematical, statistical area of experimental psychology, that what you actually accomplished was a change in the subjective probability of the experimenter. Thus you could have whole programs that would veer off into some arcane area really until someone would show up and point out how funny it was. The mathematics were assumed to be done properly – that is not addition errors etc. One kind of funny thing from a short stint I spent at Soc Rel at Harvard analysing data from some African study of elememtary children. They used a program called “Data Text” which took a set of data and ran every known statistical test that could be applied to it. According to probability, if your set level is say p = .05 that means that if you run 100 tests on completely random data the results will be significant in five of those tests. So you take those five results and publish. Fun Huh?

    • Mormo

      Normalize, that is, or so I have heard. (It’s also a poll of “likely” voters…)

      Did you know that in Cloud Hampshire, Obama/Biden defeated McCain/Palin among 2010 likely voters by a narrow plurality, 49.4-47?

      With the great caveat that this is intended as a likely voter poll, re-weighted to the August ’08 registration numbers posted above, the Sununu-Hodes numbers would be:

      Hodes 42.4
      Sununu 45.1

      Which isn’t much closer…but Hodes shows a very small lead among registered Undeclareds, 42-40.

      Also, this frankly sounds pretty good to me…Sununu has a much more famous name and statewide history. Hodes has been the CD2 rep for two and a half years. It’s easy to see growth opportunities for Hodes, notably in CD1, and harder to see any voters persuadable for Sununu. I think it really comes down to national climate and the campaigns..I’d look for a lot of nastiness from Sununu to try to make the election like 2004-Pres, to turn off 2008 Obama voters.

      Also because he is not a nice person.

      • Mormo

        That is, the August 08 numbers Andrew posted above. Using Ray Buckley’s post-Election Day numbers from this diary, the numbers (EVEN BIGGER CAVEAT, THESE INCLUDE E-DAY REGISTRATIONS PRESUMABLY LARGELY TO VOTE OBAMA!) would be:

        Hodes 42.7
        Sununu 44.6
        Other 3.2

        Looky I play with numbers too!

        PS— The voter rolls increased by over 10%, 89,385 voters, from August to post-E-day. You have to figure some of those new voters will turn out for 2010 if we do our work!

        • Ray Buckley

          that assumably this is the same poll that has Lynch over 70% and Obama over 60% (and all those issues as well). If true, and if it is true it is skewed, then using Mormo’s corrective formula it would show that the president and governor even more popular. Which, to me, would be even worse news for NH Republicans.

      • elwood

        If it were simply a matter of UNH having a sample that was representative except with regard to party IS this would work.

        But the poll overcounts older voters and overcounts the First District too.

        So by re-weighting based on the sample, you are extrapolating from an older Democratic sample slanted to the East.

        • Mormo

          Indeed, the corrections hardly amount to a hill of beans, what with the tiny sample size, the smaller subgroup sizes, and the likely voter screen, whatever it is. I would argue that the best numbers that we have are party registration, and that by reweighting to come a least a little closer to the “true” (whatever that is for this “likely 2010 voter sample”), we also correct some of the CD1 oversampling and youth undersampling—that these are at least partly collinear with party registration. I’m looking for the best we can do out of some bad numbers to begin with (a small uncorrected sample).

          It should be noted that some reputable pollsters don’t weight by party I, and have good reasons for doing so. I think that given a small sample weighting becomes more important; as the pollster you would want to squeeze as much bias as possible out of it, even at the risk of also squeezing out a little information.

  • Smuttynose11

    I’ve always had an issue with the relatively small UNH polling samples. In this one, it looks like they polled just over 400 people. I went to school at Quinnipiac in Connecticut and worked at their polling institute for a couple semesters. The polling samples were virtually always over 1,000,  and sometimes as much as 2,000 at Q-pac. I think that poll has established such a good reputation, largely because of those large polling samples it draws from.

Site maintenance and hosting by Hoeferweb