In the days following Hillary Clinton's surprise victory in the New Hampshire primary, some voting activists online began to speculate that the results may have been compromised by computerized voting systems used in the state.
About 80 percent of voters in New Hampshire used "optical-scan" voting systems -- they cast their votes on a paper ballot which was then scanned into an electronic counting machine. Those systems are made by a company with a foul record of election integrity, Diebold.
To me, the initial evidence for fraud looked thin, but I was glad when presidential candidate Rep. Dennis Kucinich asked for a hand recount of the paper ballots -- here was a chance to clear up the whole mess for certain.
He paid the state nearly $30,000 to conduct the recount. That was enough money to count all the Democratic ballots in the state's most populous county -- Hillsborough, which includes the cities of Manchester and Nashua -- and some, but not all, of the ballots in another big county, Rockingham (which includes Portsmouth).
After a couple weeks of counting, the state concluded Kucinich's requested recount on Wednesday, when funds ran out. At the time, officials had counted nearly 40 percent of the state's votes by hand.
Here's the upshot: The recount resulted in very few changes from the original results.
For instance, these are the results for Hillsborough County, where the majority of votes were tabulated on Diebold machines, and where Clinton won a plurality.
As you can see, the recount revealed only minor changes. Among the major candidates, Bill Richardson's total was most significantly altered by the recount -- and among the top four candidates, he was the only one to gain votes in the recount.
Clinton, Obama and Edwards all did better in the original count than the hand count, but not by much. See full results from the Hillsborough recount here.
Because not all of Rockingham County's votes were counted before Kucinich's money ran out, I can't present a similar table for results there. The Secretary of State's Office has posted results of the partial recount on the Web -- as you can see, in places where the the ballots were manually recounted, the results in Rockingham also changed very little from the numbers we saw a couple weeks ago.
In the town of Derry, for instance, Clinton got 2,387 votes in the original count, and 2,390 in the recount; Obama's went from 1,632 to 1,636. Similar small changes were seen in Londonderry and Salem, two other big towns in the county.
So what does this recount tell us about voting in New Hampshire?
Was there fraud? Did the machines malfunction, or in any other way help Hillary Clinton win?
Voting activists are not satisfied, and Kucinich, even though he's no longer running for president, is asking for a more comprehensive recount. And, in fact, some relatively important discrepancies were turned up; I'll go over those below.
But one point bears explanation. Before the recount, many who were suspicious of the results fixed on the fact that Clinton seemed to have done better in voting jurisdictions that used optical-scan machines.
One explanation for the discrepancy that activists floated was fraud: Clinton could have done better in the machine-counted areas because the machines there were rigged.
Does this recount prove otherwise? To many activists, it won't; they say that unless you count all the machine-counted ballots by hand, you can't be sure that the machines somewhere weren't rigged.
Fair enough -- as I've said before, people are right to be skeptical of machine counts. Reformers have long called for voting officials to institute an automatic audit of all election results; officials ought to move immediately to do so. When an election concludes, a percentage of the ballots should be selected at random and counted to make sure that the machines' results hold water.
Because we know that the machines can be rigged, until we conduct election audits automatically, every election will be plagued by controversy. (Note that such an audit only works for optical-scan voting systems, which produce a paper trail. Paperless touch-screen machines have no place in our elections.)
Still, one can take solace in the recount results. In the most populous county in the state, there were only marginal differences between the original count and recount. Hillary Clinton really did win there. The machines in Hillsborough and in Rockingham counted the votes correctly the first time. This is good news.
You might still wonder about Clinton's voting-machine advantage -- if it wasn't fraud, why did Clinton seem to do better in places that used electronic voting machines?
An academic study by political scientists Michael Herron, Walter Mebane and Jonathan Wand released this week adds evidence for the most plausible answer, that New Hamphire locations that use similar voting technology tend to be of similar demographics, and, thus, tend to vote similarly.
The paper -- here's the PDF -- concludes:
In other words, when controlling for demographic differences, differences between machine-count regions and hand-count regions in New Hampshire disappear.
So why is Kucinich asking for a full statewide recount?
Kucinich and other activists have pointed to two relatively major discrepancies found in the recount of votes in Hillsborough County, specifically in ballots cast at Manchester's Ward 5 and Nashua's Ward 5, where optical-scan Diebold systems were used.
In Manchester 5, as shown below, the recount resulted in a significant loss of votes for major candidates:
And in Nashua 5, the recount left Clinton and Edwards with fewer votes:
The percentage change in these places was greater than in many other locations. Why? What happened here? Why did the original count -- the machine count -- get it so wrong?
State officials have explained the problems as "human error." I left messages for the city clerk in Manchester for a better explanation; I'm waiting to hear back from her.
But this afternoon, I spoke to Paul Bergeron, the city clerk of Nashua, and he explained the discrepancy in his city (the indefatigable Kim Zetter of Wired News got in touch with him earlier this week, scooping me).
Bergeron explained that when an optical-scan machine recognizes that a voter has voted for "write-in" candidates on a ballot, the machine sets that ballot aside for a hand-counting of the written-in names. In Nashua Ward 5, 172 people had written in names in the vice-presidential slot, and so on election night, officials examined these ballots to count the V.P. choices.
But when counting the ballots, officials also recounted choices for the presidential races on those ballots -- but the machines had already tallied those selections. The presidential vote for those ballots was thus counted twice. "They simply goofed," Bergeron told me. The problem was corrected in the recount.
A similar problem seems to have occurred in Manchester Ward 5. There, some voters wrote in leading candidates' names as selections for vice president. After the machines had counted the presidential votes on those ballots, a clerk who examined the ballots added the write-in V.P. votes to the presidential category: If you wrote in Obama for V.P., the clerk added that to Obama's presidential total.
That was a mistake -- a write-in vote for Obama for V.P. should not count as a vote for him for president. And so, in the recount, the candidates lost votes.
Voting activists have also pointed to other problems they say occurred in the recount. Of particular concern is the "chain of custody" surrounding the ballots that were recounted; videos and photographs, they say, show the ballots were vulnerable to tampering as they were transported across the state.
Based on these problems, Kucinich sent New Hampshire Secretary of State Bill Gardner a letter requesting a full recount of all ballots in the state.
Kucinich is not offering to pay for such a recount. An aide at the Secretary of State's Office told me that the state will not be paying for statewide recount itself. So it's pretty clear that a full recount of the Democratic ballots in New Hampshire will not occur.