This is a followup on three previous diary entries:
Let summarize what follows: as bad as it looked previously, it's worse.
If you've read the previous diary entries, or you've been following the FreedomToTinker blog, you know the background: Sequoia Voting Systems is a voting machine vendor, and their machines can't count. Well, not correctly, at least. This is often considered a problem for voting machines.
Sequoia has done everything it can to prevent independent third parties from scrutinizing their systems, including directly threatening a noted group of Princeton researchers led by Dr. Ed Felten.
Before getting into the latest revelation to come out of this, let me explain something about crosschecks in voting systems. Because fraud is a constant concern, all voting systems and procedures are designed to sanity-check numbers against each other. For example, the total number of votes cast for all candidates on the ballot in race 1 can't exceed the total number of voters. (It might be less due to undervoting in that race.) The number of votes cast for all candidates in race 2 of party Y should equal the number of votes cast for party Y candidates in race 2. (That is: one sum by candidate, one sum by party.) And so on. In some cases, fixed rules can be used, because we know a priori that they'll never be broken. In other cases, reasonable approximations can be used based on voter patterns: for example, in a general election, the Presdential campaign is unlikely to be undervoted more than a state or local campaign. Those approximations can incorporate broad patterns (like that one) or specific ones like "in the history of general elections in state X, undervoting for senate seats has never exceeded 4%", by setting the check threshold at, say, 6% and flagging results for review if it's exceeded.
Whatever those crosschecks are, they're important, because they not only provide some measure of assurance that the voting system is functioning properly, they also defend against some attacks against the integrity of the process. For example, if there are 10 machines used in Precint 123, then the total number of votes cast in race X on all 10 should be exactly equal to the number of votes recorded as cast in race X at the county level.
With that in mind, let's check Dr. Felten's latest finding: NJ Election Discrepancies Worse Than Previously Thought, Contradict Sequoia's Explanation.
He's been looking at some new summary tapes. These are produced by each machine and are signed off by election officials. A previous discrepancy in a number of these was met by an incomplete, marginally plausible explanation from Sequoia that "a voter would be allowed to vote in one party while being recorded in the other party's turnout." If true, this is a bug, and a fairly serious one, but it's at least conceivable that it addresses that situation. But it absolutely doesn't cover this one:
The Republican vote totals are Giuliani 1, Paul 1, Romney 6, McCain
14, for a total of 22. The Democratic totals are Obama 33, Edwards 2,
Clinton 49, for a total of 84. That comes to 106 total votes across the
two parties.
The turnout section (or "Option Switch Totals") shows 22 Republican
voters and 83 Democratic voters, for a total of 105.
This is not only wrong - 106 votes cast by 105 voters - but it's also
inconsistent with Sequoia's explanation. Sequoia says that all of the
voters show up in the turnout section, but a few might show up in the
wrong party's turnout. ("In every instance, however, the total turnout,
or the sum of the turnout allocation, is accurate.") That's not what we
see here, so Sequoia's explanation must be incorrect.
Sequoia's previous feeble explanation was that (say) someone who voted for Romney (R) would have their vote correctly counted for Romney, but that it might be included in the total number of (D) votes. So the candidates would all end up with the right number of votes, but the per-party totals would be wrong. This is of course not only an obvious error, but a major one that should have been caught in the first few hours of quality assurance testing.
But even if we accept this flimsy excuse, it doesn't cover this latest problem -- because it deals with the total of all votes. What the machine is telling us is that 106 VOTES WERE CAST BY 105 VOTERS.
"That is not even good enough to be wrong."
But, as Dr. Felten observes, that's not the end of it:
And that's not all. Each machine has a "public counter" that keeps track
of how many votes were cast on the machine in the current election. The
public counter, which is found on virtually all voting machines,
is one of the important safeguards ensuring that votes are not cast
improperly. Here's the top of the same tape, showing the public counter
as 105.
The public counter, as he goes on to note, is actually important enough that poll workers are required to sign off on it -- to certify that the number recorded by the public counter is correct. This is one the safeguards in the process: for example, if the public counter reads 105 but there are only 50 votes for president, that's an undervote marked enough to warrant attention. And obviously if the public counter reads 105 but there are 106 votes for president, something is seriously wrong. As we see.
This is not a one-system problem, either:
Another of the new tapes, this one from West Deptford in Gloucester
County, shows a similar discrepancy, with 167 total votes, a total
turnout of 166, and public counter showing 166.
Without a review of ALL the tapes from New Jersey, it's impossible to know how widespread this problem is, whether all the errors are off-by-one, whether all the errors are in the same direction, what might be causing them, or anything else. Finding out will require a fully independent investigation -- exactly what Sequoia is trying to block by threatening anyone who might conduct one.
And it also leaves unanswered a larger question: if these machines can't manage to perform their basic functions properly, what else is wrong with them that isn't so obvious?