NH Primary Recount: Sat AM Update

Just under 80,000 ballots have now been recounted. (Note: Weare has not, and the Bedford vote for Richardson should read 183 original, 184 recount. That transcription error appeared yesterday too, and was then corrected).

The recount now includes 14 hand-count towns casting a total of 5,661 ballots.

The total undervote (votes not counted originally then counted at the recount) is 521. The total overvote is 426. Factoring in scatter reallocation, the total miscount rate is 1.17%.

Analysis below.

Of the miscount:

• 141 miscounts come from Manchester Ward 5, where poll workers mistakenly tallied Vice Presidential write-ins for Clinton, Obama and Edwards as Presidential votes.
• 100 apparently come from a clerical error in Wilton, where 286 Obama votes were reported as 386.
• 50 come from New Ipswich, where 77 Richardson votes were reported as 27.

The Wilton and New Ipswich explanations are admittedly only educated guesses on my part.

Without these clerical errors, the miscount rate would be 0.8%.

Not discounting apparent clerical errors:

• The miscount rate in scanner precincts is 1.1%
• The miscount rate in hand count precincts is 2.6%.

•  27% of the total state now recounted.

  Sorry to post and run. I don't usually do that, but I'm trying to get the recount numbers out ahead of speculation.

  by Elwood Dowd on Sat Jan 19, 2008 at 05:55:27 AM PDT

  •  What a mess

    Even though 1 or 2% doesn't seem like a huge issue in the over-all picture; if it were yours or my vote it is a big damn deal. If we take these percentages nation wide, it could mean as much as several million lost or over counted votes.
    We haven't gotten serious about the voting system since 2000! We have got to have a better system.

    by gabie on Sat Jan 19, 2008 at 06:07:28 AM PDT

    •  I would say it is more complicated

      All these errors, taken together, have had very little effect on the vote totals of the leading candidates (or any other candidate, for that matter). The net impact is far under 1%. That wasn't a foregone conclusion -- and it isn't a conclusion yet. Just how things have gone so far.

      I don't think we need perfect vote totals -- but we need reasonable assurance that the errors haven't altered the outcome, and we need to learn from errors so we minimize them in the future. If we Just Assume that the errors are harmless, we are in a bad way.

      That's why I support rigorous routine post-election audits. That's HVM's Stand (tm).

      by HudsonValleyMark on Sat Jan 19, 2008 at 06:21:51 AM PDT

  •  Thanks for the update. n/t

    by vivycakes on Sat Jan 19, 2008 at 06:58:07 AM PDT

  •  thenk you Elwood

    you deserve a medal.

    by DemFromCT on Sat Jan 19, 2008 at 07:10:27 AM PDT

•  what about Nashua 5?

  any word on that?

  by HudsonValleyMark on Sat Jan 19, 2008 at 05:56:15 AM PDT

  •  Brief article in Nashua Telegraph

    says it was human error. So, either another case of write-in misassigned to Presidential candidates, or an error in transcribing the scanner results.

    by Elwood Dowd on Sat Jan 19, 2008 at 02:09:48 PM PDT

•  Interesting

  that the miscount rate is lower for scanner precincts than hand count precincts. I know it's unwieldy (and formidably expensive in large states like CA) but I wish we could automatically recount all elections.

  After Florida 2000, I don't trust our electoral system.

  by coral on Sat Jan 19, 2008 at 06:05:52 AM PDT

•  Regardless Who Wins or Loses...

  This NH recount is accomplishing something that should be the goal in all elections, to get the most accurate and error free count of all the peoples votes as is humanly possible.  This should be the goal of all the States for every election.  If there is any suspicion of the counted votes not matching the actual votes there ought to be a recount of paper ballots.

  Ask a voter if he would mind if his vote didn't get counted even if it would not effect whether his candidate won or lost?  I know I would mind!

  Recounts are not about confirming conspiracy theories, they are about making sure every person's vote is correctly counted!

  Kudos to Dennis Kucinich and the NH Secretary of State for valuing our votes so highly. I hope this catches on in other States.

  by Doctor Who on Sat Jan 19, 2008 at 06:41:10 AM PDT

  •  Would you also insist...

    that the home team bats the bottom of the ninth, regardless of whether they have already won?

    Ask a voter if he would mind if his vote didn't get counted even if it would not effect whether his candidate won or lost?  I know I would mind!

    What this recount demonstrates is that counting votes is hard, whether scanned or by hand, and expensive. Is it really worth the effort of so many people to ensure the figure is correct to the nearest 1? Is that even possible? The WA-Gov recount of 2004 (?) suggested that if the result was close enough the count was basically a crap shoot.

    I would also prefer that I heard the result within a day or two, rather than having to wait several weeks.

    by leberquesgue on Sat Jan 19, 2008 at 06:56:57 AM PDT

    •  Impatient

      Let me get this straight:

      I would also prefer that I heard the result within a day or two, rather than having to wait several weeks.

      Sounds like you would sacrafice getting the right vote count for getting a fast vote count.

      Sorry, I don't share your impatience.

      by Doctor Who on Sat Jan 19, 2008 at 07:16:29 AM PDT

      •  do we know what we're arguing about?

        Are we arguing about whether it's worth, say, $10 million to distinguish between a margin of 1,234,567 vote and a margin of 1,245,568 votes?

        That's just a hypothetical. I'm just looking for clarification about the scope of the disagreement.

        by HudsonValleyMark on Sat Jan 19, 2008 at 07:24:59 AM PDT

        •  $10 Miilion is a small price to pay

          for confidence in our voting system, the foundation of our democracy. Everyone should go to the polls with the sure knowledge that their vote will be counted.

          by lightfoot on Sat Jan 19, 2008 at 07:32:43 AM PDT

          •  and be certain to give the bill to you ;-)

            I love how everyone volunteers someone else's money. ;-P

            by DemFromCT on Sat Jan 19, 2008 at 10:57:32 AM PDT

          •  for clarity, could you answer the q I asked?

            You would spend $10 million to distinguish between a margin of 1,234,567 votes and 1,234,568 votes?

            Because I don't think that has anything to do with the foundation of our democracy.

            by HudsonValleyMark on Sat Jan 19, 2008 at 11:50:55 AM PDT

        •  That was the gist of my contention, thanks n/t

          by leberquesgue on Sat Jan 19, 2008 at 09:12:55 AM PDT

      •  No

        I would like to know the correct winner. Beyond that approximate vote tallies are great, to indicate the scope of victory and influence of minor candidates, but we don't need precise vote tallies.

        The broader point of my above comment, which might not have gotten through, is that "the right vote count" is close to an impossibility, as well as being irrelevant to almost everyone.

        by leberquesgue on Sat Jan 19, 2008 at 09:16:38 AM PDT

        •  For me, voting is personal.

          I want to know that my vote counts.

          There are other issues besides the correct winner, such as the viability threshold, the amount of MSM attention a candidate receives, admission to debates, "Bandwagon Effect", etc.

          by lightfoot on Sat Jan 19, 2008 at 10:33:06 AM PDT

          •  still trying to clarify the scope of disagreement

            Sure, who won is not the only thing that matters. And, sure, we all want to know that our vote counts.

            How much are we willing to pay for the "n"th degree of accuracy in the count? This is not a rhetorical question. I'm trying to get away from the rhetorical questions.

            by HudsonValleyMark on Sat Jan 19, 2008 at 11:54:01 AM PDT

            •  People are pretty fired up

              about this issue. I am willing to believe a lot of people would volunteer, if the states would open the door to allow it.

              A 1%-2% miscount is pretty high by any reasonable standards that I can think of. That type of error would never be tolerated in the business world where very similar types of counting is done, nor in academia or science.

              Considering that Kucinich is paying for a full recount, I think that the cost of mandatory audits would be quite reasonable.

              Also, improvements in the process, as we discussed above, could help.

              Start with small inexpensive audits. If those show discrepancies then move to progressively larger audits or a full recount as needed. This type of reform need not be expensive.

              by lightfoot on Sat Jan 19, 2008 at 01:36:40 PM PDT

              •  Ahem.

                The state of New Hampshire DOES "open the door" to volunteers during the recount, and most towns are eager to recruit poll workers for election day.

                by Elwood Dowd on Sat Jan 19, 2008 at 02:11:51 PM PDT

              •  you still haven't answered my question

                If I have no idea what degree of accuracy you think is essential, then I have no idea whether we agree.

                Please note that the net change in vote totals so far is tiny.

                I say, audit at least as much as is necessary to confirm the winner (and keep people on their toes).

                by HudsonValleyMark on Sat Jan 19, 2008 at 02:58:35 PM PDT

                •  Sorry

                  I was into about the seventh paragraph of a long detailed answer then I accidentally clicked something in Firefox and Google came up, and I went back and my comment-in progress was gone...arghh.

                  I really appreciate that you are interested in this topic. Let me start again but more briefly.

                  I was aware that NH was using volunteers in the recount from reading most of the posts from this rjones2818 diary, however election laws and processes across the country are a crazy quiltwork, so that is why I specified "if the states would open the door", especially regarding recounts and audits.

                  You want a simple number for accuracy, but the answer is not that simple, sorry. Some states have an automatic trigger for recounts when an election is close, which I believe is a good start, however in my view that is necessary but not sufficient to reform the election system.

                  In my view one critical reform that is needed is an audit process that uses some type of stratified sampling to test each component of the voting system along with the voting system as a whole, per precinct, and ultimately in the general per state. That way we can catch "random" gross errors that still deprive voters of their individual votes, and we can also detect systematic errors that are indicative of electoral fraud.

                  In designing such an audit process we would need to look at each precinct and examine the system in place. What are the key components? What degree of confidence do we want to have in each component? What degree of confidence do we want in the entire system, per precinct? The degree of confidence determines the number of sample ballots we would need to ue in the audit. This is where the number come in, not in the actual vote percentage. This is fairly elementary statistical theory. We would want to postulate a hypothesis test for each critical component in the system that would give us some N% degree of confidence in the proper functioning of that component. I think N would have to be at least 95% and perhaps 99%, but again this test would probably need to be designed per precinct.

                  From the diary, above:

                  Of the miscount:

                     - 141 miscounts come from Manchester Ward 5, where poll workers mistakenly tallied Vice Presidential write-ins for Clinton, Obama and Edwards as Presidential votes.
                     - 100 apparently come from a clerical error in Wilton, where 286 Obama votes were reported as 386.
                     - 50 come from New Ipswich, where 77 Richardson votes were reported as 27.

                  So in New Ipswich, 50 Richardson voters would have been deprived of their vote if not for the recount. I wonder how they feel now? Some people would have us believe it is better that they did not know, but I would not dare tread down that path. We must audit in a way to identify the source of the errors to continuously refine the system. How did this error occur in New Ipswich? Who or what is responsible? What is wrong with building-in some accountability in the voting system?

                  Let me take your original post as an oversimplified example. So we have teams of two people in a loud crowded room, one on each team reading the votes and  the other tabulating the votes. Let's say we have Ole Bill, long-time volunteer, who has become hard of hearing and error-prone in this environment. Is an auditing system in place to identify the problem so somebody can tell Bill to quit being so damn stubborn and get that hearing aid already?

                  Continuing with my semi-humorous but hopefully illustrative example, say in the case of fraud, ole curmudgeon Goofus Mordoch just returned from a midnight meeting of Ron Paul supporters in the ruins on Mt. Moosilauke, where it was decided that Paul needed a couple of percent more to keep his candidacy alive so he could run independent in the general and siphon off of the Dems. Could we catch Mr. Mordoch?

                  If optical scanners are so great, why do we have errors? Can we identify the source of these errors? Can we improve the process? Without audits, I am afraid we a flying blind.

                  (Heh, not so brief after all)

                  by lightfoot on Sun Jan 20, 2008 at 07:11:03 AM PDT

                  •  OK...

                    Not that it is 'on you' to solve the problems of the world, but in order to operationalize your idea of statistical confidence levels, I think you need to pin down the referent. What counts as "proper functioning"?   That will depend on your tolerance for error.

                    "If optical scanners are so great, why do we have errors?"

                    Well, several reasons, some of which have nothing to do with the scanners. But to me the question almost seems facetious: I'm having a hard time thinking of a process that doesn't have errors. Of course I agree that we need audits in order to understand the nature and extent of those errors, as well as to deter fraud.

                    But as long as we are stuck at the level of abstraction of 'making every vote count,' it seems hard to design an audit to match our goals.

                    by HudsonValleyMark on Sun Jan 20, 2008 at 10:49:21 AM PDT

                    •  Make every vote count.

                      Clearly it is a worthy goal, not an absolute. Really I would turn one of your questions around and ask you, what "N" percent of vote spoilage do you see as acceptable?

                      Why do optical scanners have errors? With this question I would like some specifics on how the systems fail, so we can address the problems and design a process of continuing improvement in the electoral system. But it seems that no such process is in place, which is a shame. Yes, all systems have errors, but just the fact that we are communicating via the internet using personal computers tells us that some systems can be built virtually error-free. If your computer memory flipped 1% of its bits every session, we would not be having this discussion .

                      "What counts as "proper functioning"? In most cases when a paper ballot is available for comparison, it is a simple binary value: "Was the vote counted properly, yes or no?" In  simple case with two candidates running, or say a yes/no vote on a proposition, the statistics are binomial. We could audit by formulating a hypothesis test on the component: Does the component count the vote properly? The degree of confidence we choose determines the number of samples needed for the audit.

                      The details above are just an example. Perhaps it would be better to compare a voting system component's statistical distribution with hand-counted ballots using a non-parametric method such as a Kolmogorov-Smirnov test. Further research would be required. But with the stubborn resistance that I see to any electoral reform, I often wonder why we are not applying ourselves to this problem and going down the progressive path towards the goal of "making every vote count", and instead choose to live in darkness regarding our election system.

                      It seems there is an elitist attitude toward the common voter, "It is better they do not know".

                      Does that about nail it for you?

                      by lightfoot on Mon Jan 21, 2008 at 06:48:31 AM PDT

                      •  BTW, Thank you

                        Thank you for asking the important and intelligent questions. Having the opportunity to engage in any kind of intellectual conversation on dKos is rare event in my experience here.

                        Also, if my choice of words at times is overly emotional, please forgive me, as this is an issue that I feel passionate about.

                        by lightfoot on Mon Jan 21, 2008 at 07:00:15 AM PDT

                      •  hey there

                        My last comment was a bit terse because I was running out the door, but I'm grateful that you took it in the spirit I intended.

                        I in fact work with lots of other good people on election audit design, so I take this very seriously too.

                        Let me first say more or less what I think, then return to some of your comments.

                        WRT "spoilage," I would distinguish between accurately recording voter intent and accurately tallying voter intent. We want to get misrecording as close to zero as possible. We probably can't get to zero, because voters don't always follow directions, and if we bang too hard on trying to reduce their error rate, we may actually discourage more people from voting than we benefit. (Also, it's damn hard to measure that error rate; even the idea of "intent" can get fuzzy, especially down-ticket where voters may not have known that a particular contest would be on the ballot.) But clearly some equipment is more reliable than other equipment, and clearly some interfaces and ballot designs are more human-friendly than others. Some things are known about this, but routine audits could help us to learn much faster from experience.

                        From a technical standpoint, I'm not sure we have a better idea right now about minimizing misrecording than to use hand-marked ballots, ideally in conjunction with scanners that are programmed to reject apparent overvotes. (Maybe also undervotes.) In principle, DREs could further reduce misrecording; as to Actually Existing DREs, there doesn't seem to be much consistent difference, but when DREs misrecord, they can really misrecord.

                        We also want to minimize mistallying, but the stakes are somewhat different, because (as I use the words here) mistallying error can be corrected after the initial count, while misrecording error cannot. From the audit/recount results I've looked at, most mistallying is due to voters' not quite following instructions, and mistallying rarely alters vote shares by more than 0.1%. But there is no logical limit to how large mistallying could be. I think that the prevailing rate is acceptable provided that appropriate measures are taken to prevent when possible, detect and correct when necessary, error rates that are unusually high and/or could alter the election outcome.

                        --

                        Yes, it seems that a lot more can be done to promote process improvements in election administration.

                        Checking tally accuracy is really not a simple binary, because (1) some voters mark their ballots in ways that human observers would consider unambiguous, yet scanners probably should not be programmed to recognize, and (2) some voters mark their ballots ambiguously. And then of course there are all the ways that scanners can produce Just Plain Wrong results. (If you don't know it already, I think Doug Jones has a treasure trove of information on some of these subjects.)

                        "We could audit by formulating a hypothesis test on the component: Does the component count the vote properly?" --Are we testing against a 0 error rate, or something else? You seem to be fudging on the crucial aspects of your test. (Or maybe part of the confusion is that you may or may not have switched from a binomial CORRECT/INCORRECT to a different binomial CHOICE A/CHOICE B.)

                        "...I often wonder why we are not applying ourselves to this problem..." Speak for yourself (grin). No, it's a fair question why 'election science' isn't more advanced -- although it might be further along than you realize.

                        To some extent, election officials sometimes resist post-election audits because these are perceived as one more onerous item on the to-do list, rather than opportunities to learn and to improve. (And, probably, some fraction of election officials actually is crooked as hell.) But in truth, I think the rest of us were slow to catch on about applying auditing principles to elections, so there wasn't much for election officials to resist! We're all figuring it out together.

                        by HudsonValleyMark on Mon Jan 21, 2008 at 08:10:30 AM PDT

                        •  Thank you very much

                          fo your reply, HudsonValleyMark, you have given me some good insights and leads to further research and educate myself on this topic, as well as some juicy food fo thought.

                          For example, I understand your point about the binomial problem; my mistake was thinking to simplify CHOICE A/CHOICE B into CHOICE A/not CHOICE A (alternately CHOICE B/not CHOICE B) when for a full analysis it would seem to be important that we also identify the case CHOICE A and CHOICE B.

                          I'll leave this discussion with the inspiration to go off and do some further "homework" on this topic that I care so much about. In our feelings towards electoral reform I think we are not so far apart in spirit. Cheers!

                          by lightfoot on Mon Jan 21, 2008 at 10:27:31 AM PDT

    •  Um, not to mention the obvious or anything.....

      but baseball is a game and our right to vote and be counted is not.
      Dennis Kucinich has paid $27,000 out of his campaign funds to pay for this recount.  For the sole purpose of voter-confidence and integrity, I rather know if we can trust our system or not ahead of the actual election...

      by ikrisarus on Sat Jan 19, 2008 at 07:31:48 AM PDT

•  I used to do the handcount

  in the town next to Wilton.  I started when I was 18 and I always thought we should count it more than once.  We would get a group of ballots, by 50s or 100s, and one person would read the name while the other marked down the tally.  When we were done we'd total it up, circle the count, put a rubber band around it and hand it in.  Then we'd start on the next batch.  

  We sat at long tables and called the names across the table to our partner.  But there were 10-12 others doing the same thing on each side of the table.  It would have been really easy to hear the person next to your partner calling out Carter and end up marking it down on your tally.  I don't know if that ever happened but I don't think the conditions were ideal.

  by sobermom on Sat Jan 19, 2008 at 06:57:26 AM PDT

  •  This an insighful comment.

    What is needed is some follow-up on how to make the system work a little better. A recount or audit can find these problems too.  In this case, maybe the answer is a simple as having a little more separation between people, different seating arrangements or bigger tables?

    by lightfoot on Sat Jan 19, 2008 at 07:37:46 AM PDT

    •  Absolutely

      There are problems that can be fixed quite easily.  As you said, smaller tables with more separation or separate tables.  It does take a long time to do handcounting.  But I also think that a recount to double check in the same night is not unreasonable if the conditions aren't great.  The more people that are there, the quicker it will go.

      by sobermom on Sat Jan 19, 2008 at 07:44:45 AM PDT

      •  Sounds reasonable to me

        just like double-checking your math problem. I would still like to see mandatory audits. I used to live in New Hampshire, in West Lebanon, and I know it is a great place where you can trust your neighbors and you do not have to lock your doors. That is how I felt, anyways.

        But corruption exists, and I do not understand the Pollyanna attitude of some on this site that seem to deny it (not referring to you at all, but the attitude of Kos and other leaders here).

        I still favor mandatory audits for elections. It would not be that expensive, it would catch gross errors that could turn close elections, and it would increase confidence in our voting system, which would in turn probably increase turnout. If people think their vote might not count, they might not bother to vote.

        by lightfoot on Sat Jan 19, 2008 at 08:01:04 AM PDT

        •  heh

          I do not understand the Pollyanna attitude of some on this site that seem to deny it (not referring to you at all, but the attitude of Kos and other leaders here).

          I have had a lot of things said about me, and heard a lot of things said about Kos.

          Polyanna isn't an especially frequent complaint. ;-)

          I share your attitude towards FL, which has a bit of history (so does OH). There can't be enough scrutiny there. That has little to do with NH, where, as you point out, things are done differently.

          What I would hope this result would do is force people to reevaluate both OS devices, and those who overstate problems. The devices worked very well and the problems have been grossly overstated.

          Since those, up until now, appear to be hard facts, Kos and the site leaders happen to have been correct in their judgment about this particular recount. And what you need to understand is the damage done to the effort to improve elections by those who unjustifiably cry wolf. We don't want NH to be used as an excuse to do nothing elsewhere, just because this went well in NH.

          by DemFromCT on Sat Jan 19, 2008 at 01:31:39 PM PDT

          •  we've had 2 days

            don't we have 3 weeks total?

            Ya gotta admit, this has captured peoples attention. I'm hoping it sets into peoples minds as a kind of standard since New Hampshire so far seems to be running a tight ship. Rather than being damaging it might turn out to override the damage done by irrational claims and give people a hunger to see this happen more often in cases where there are questions. I hope so at least. And maybe it will get the half cocked activists to stop. But I'm happy none the less because even with all the yelling that went on all over the place, we got to see some democracy in action. I hope people get addicted to it.

            The other thing I wish would happen would be for a state to do a random sample hand recount and publish the results to give people an idea of what the norm should be.

            It seems a lot of people are treating this like a sporting event, pulling for either big discrepancies or flawless counts, not really caring about just knowing what happened. Not so much as a few days ago. And like a sporting event, carrying all the ill will and name calling that can go on between superfans.

            by YetiMonk on Sat Jan 19, 2008 at 04:17:14 PM PDT

            •  actually there aren't that many people

              paying attention at all, outside of us. '-P

              See link.

              That's what I mean about this being a political process and not simply a weighing of this expert vs that expert. This is not on the voting public's radar at all.

              But us? We're political junkies. ;-)

              by DemFromCT on Sat Jan 19, 2008 at 04:32:51 PM PDT

    •  The answer is optical scanners.

      That's what the recount says, pretty clearly.

      by Elwood Dowd on Sat Jan 19, 2008 at 02:12:56 PM PDT

•  When purchasing a computer server

  the goal is "5 9's", that is, the probabilty that the server will be working at any given time is 99.999%.

  The margin in the Florida 2000 presidential election was 0.09% before the recount was stopped by the U.S. Supreme Court.

  Although a recount did not change the result in this election, in general a miscount rate of 1.17% is not acceptable. We can do better. I am glad Kucinich took this early opportunity to point out the flaws in our voting system.

  Mandatory auditing would be a step in the right direction.

  by lightfoot on Sat Jan 19, 2008 at 07:28:15 AM PDT

  •  1% is perfectly acceptable.

    If it can be made better, great. But realism is in short supply these days.

    if nothing else, the inexact science of elections should be fully understood.

    Because it is an inexact science, I am in full agreement about mandatory audits and the paper ballots that allow them (and recounts) to happen. if elections were as precise as you suggest, it wouldn't be necessary to do either.

    by DemFromCT on Sat Jan 19, 2008 at 11:00:58 AM PDT

  •  What mythical process do you propose?

    Comparing server availability - and no, most servers are NOT spec'd out for five 9's - to a process that involves heavy human interaction, interpretation of hand-marked ballots, and mechanical equipment - is nonsensical.

    by Elwood Dowd on Sat Jan 19, 2008 at 02:19:24 PM PDT