Skip to content
 

Voting Technology and the 2008 New Hampshire Primary: Herron, Mebane, and Wand don’t find any problems

There’s been a lot of talk about the recent New Hampshire primaries. Now it’s time to hear from the experts, in particular, Michael Herron, Walter Mebane, and Jonathan Wand, the political scientists who, among other things, did the definitive estimate of the Florida vote from 2000. Their punchline: “with respect to Hillary Clinton’s surprise victory in the Democratic Primary and the notable differences across vote tabulation technologies in Clinton’s and others’ levels of support, our results are consistent with these differences being due entirely to the fact that New Hampshire wards that use Accuvote optical scan machines typically have voters with different political preferences than wards that use hand counted paper ballots.”

Here’s their paper, and here’s the executive summary:

We [Herron, Mebane, and Wand] address concerns that the reported vote counts of candidates running in the 2008 New Hampshire Presidential Primaries were affected by the vote tabulating technologies used across New Hampshire.

• In the Democratic Primary, Hillary Clinton was more successful in New Hampshire wards that used Accuvote optical scan vote tabulating technology than was Barack Obama, receiving 4.3 more percentage points of the vote there (40.2% for Clinton versus 35.9% for Obama). In contrast, Clinton did worse than Obama in wards that counted paper ballots by hand, trailing by 6.1 percentage points (33.7% versus 39.8%).

• In the Republican Primary, Mitt Romney trailed John McCain by 3.6 points in Accuvote wards and by 15 points in wards that counted ballots by hand.

• In New Hampshire the choice of vote tabulation technology is made ward by ward, and electronic technology was used in wards that typically differ demographically and politically from wards that count ballots by hand. Wards that selected electronic tabulation are disproportionately from the southeast part of New Hampshire, and they tend to be more densely populated and more affluent. Accuvote and hand count wards have also typically produced divergent voting patterns in elections prior to the 2008 primary. It is plausible that most or all of the observed differences between vote tabulation technologies in the votes candidates received reflect such background differences and not anything inherent in the tabulation methods.

• Using a subset of New Hampshire wards that have similar demographic features and voting histories but differ in their vote tabulation technologies, we find no significant relationship between a ward’s use of vote tabulating technology and the votes or vote shares received by most of the leading candidates who competed in the 2008 New Hampshire Presidential Primaries. Among Clinton, Edwards, Kucinich, Obama and Richardson in the Democratic primary and Giuliani, Huckabee, Paul, Romney and McCain in the Republican primary, we observe a significant difference only in the votes counted for Edwards, and that difference is small (a deficit of between 0.6 and 3.4 percent in the hand-counted votes).

• With respect to Hillary Clinton’s surprise victory in the Democratic Primary and the differences
across vote tabulation technologies in Clinton’s and others’ votes, our results are consistent with these differences being due entirely to the fact that New Hampshire wards that use Accuvote optical scan machines have voters with different political preferences than wards that use hand counted paper ballots.

10 Comments

  1. CHCH says:

    great news! and very interesting besides.

    The only question I have is whether such techniques have sensitivity to the kind of skew that would be expected of a programming "error" (systematic or stochastic) in accuvote machines, or if the point is more that the demographics are so inextricably tied to vote method that such variability could never be dissociated from the demographics. I imagine it would be easy to rerun the analysis after manually "tweaking" the results in one particular candidate's direction…

  2. Spl0it says:

    Why don't we just wait for the re-count to finish? Considering they've already found issues with the re-count this seams a little premature.

  3. 7oby says:

    In Hillsborough County, Clinton gained 675 votes – a jump of more than 2 percent. Time will tell if more errors are forthcoming.

    Source:
    http://www.gnn.tv/headlines/16476/Diebold_dispari

    I disagree, I hear things like being within 1% is really good for vote counting machines (which I disagree with, how hard is scantron?), but a 2% difference is pretty remarkable.

  4. JD says:

    The 2000 and 2004 elections were rigged, period.

  5. Richard Parker says:

    You can always round up a few underpaid 'scientists' to write up something following your point of view.

    The report you quote sounds remarkably like 'smoke-bombing' – that is confusing everyone with a heap of jargon. They don't reveal their methods at all, or the raw data. All you have to go by is the bullshit.

    Are they really professonals at Columbia, or just over-smart students? Did they really do 'the definitive estimate of the Florida vote from 2000' or was that a load of rubbish as well?

    The Supreme Court dummies quashed the investigation of that fraud, so really, no-one knows.

    Thanks for bringing this up into the internet

    I'll file it under 'Stupid Official Explanations'.

    regards

    Richard

  6. Richard Parker says:

    You can always round up a few underpaid 'scientists' to write up something following your point of view.

    The report you quote sounds remarkably like 'smoke-bombing' – that is confusing everyone with a heap of jargon. They don't reveal their methods at all, or the raw data. All you have to go by is the bullshit.

    Are they really professonals at Columbia, or just over-smart students? Did they really do 'the definitive estimate of the Florida vote from 2000' or was that a load of rubbish as well?

    The Supreme Court dummies quashed the investigation of that fraud, so really, no-one knows.

    Thanks for bringing this up into the internet
    I'll file it under 'Stupid Official Explanations'.

    regards

    Richard

  7. bubbalo says:

    Richard instead of speculating why don't you provide evidence regarding NEW HAMPSHIRE. Suggesting they are blowing smoke because you can't understand the report is officially stupid.

  8. Sean Martin says:

    As Andrew suggests: it is quite an astonishing coincidence that there would be such a strong correlation between choice of candidate and choice of voting method. Hmmm…

  9. Garrett says:

    Richard,

    >>"They don't reveal their methods at all, or the raw data.>"Are they really professonals at Columbia, or just over-smart students?

  10. CHCH says:

    I am reposting a comment from my blog which supports my feelings about this article.

    "… I went over the Herron/Mebane/Wand matching study carefully and found the following:

    1) Their study design doesn't have enough power to detect the difference in question, even at the 90% level.

    2) They started with a large number of variables and narrowed the field to 8 without systematic or stepwise procedures.

    3) Two of their final model variables are linearly related. One variable (wpkerry04pd) is the standard normal transformation of another (pkerry04pd). I'm surprised they were able to get convergence.

    I contacted them asking for clarification and Dr. Mebane responded politely, saying:

    "We'll be issuing a revised paper in a few
    weeks with some new analysis based on augmented data.". He went on to say their variable selections were based on their research of past NH primaries, and couldn't do much about the power given the small number of matches. I didn't press them further.

    "

    Interesting.