There's Going To Be a Recount (Of Sorts)

This morning Xiomara Castro and the LIBRE Party filed a formal set of complaints about the vote counting process and its lack of transparency, documenting errors and discrepancies in the formal counting of the tally sheets of the over 16,000 Mesas Electorales Receptoras (MER). 

LIBRE representative Ricki Moncada then read the document to the assembled press.

The Tribunal Supremo Electoral (TSE) head David Matamoros agreed to a public recount of the Actas: not the votes themselves, just a recount of the votes as recorded on the tally sheets.

This, of course, is a compromise.  Ballot boxes will not be re-opened; individual votes will not be recounted. 

It's a vexing compromise because some of the problems that LIBRE alleges include alteration of the Actas themselves, which they say they can document. 

Remember that the election procedures gave the political parties a number of ways to get copies of the scanned/reported Acta at each step.  The party representative at every MER was supposed to be given a copy of the scanned Acta to take back to the party. Once the scanned Acta hit the TSE's computers, a copy was supposed to be sent to each party, and to the foreign vote auditors.

LIBRE says it has Actas sent to them by the TSE with scan dates of the early morning hours of the election day, bearing data that looks like the test data used to validate the system in earlier runs.  LIBRE also says they have copies of Actas that don't match the Acta image in the TSE central computing database, with different signatures and vote tallys.

Since the election itself on November 24, there has been a public recount of the scanned actas going on at this site:  http://conteo.votosocial.org.  In addition, there has been a Facebook-coordinated effort to identify Actas that contain problematic information or were mis-recorded in the TSE's vote counting system. 

Through these independent projects, more than 1600 problematic Actas had been identified by Saturday, November 30.

There is a further 2000+ Actas which the TSE has "sequestered" because of unspecified problems, for which vote counts, and images, have not been released.  Then there are a series of MERs for which vote totals are recorded, but no Acta image is posted.

I have been involved in the public recount of the Actas, entering the values from the scanned images of the Actas into a system that then recounts the votes once 3 separate reviews of the transcription agree that the data are correct. I also have reviewed Actas flagged on the Facebook page as problematic.  I can say first-hand that I found inconsistencies in more than 500 Actas I've reviewed over the last week. 

Some of the inconsistencies were transcription errors: the TSE had an enormous problem going from the hand-written numbers to recording those numbers in their MS-SQL database.  Over time, the TSE seemed to be correcting these transcription errors, though in a non-transparent fashion since they never acknowledged a single one of them.  Many still remained as of this past Saturday.

More troubling, though, is that the vote totals on far too many Actas added up to more than the number of people who were reported to have voted in that particular MER. 

Each Acta contains a field "Ciudadanos que votaron", which the TSE training manual documented as being calculated by taking the total number of ballots at the start of voting, and subtracting the number of blank ballots remaining at the end of voting.  The starting number of ballots and the calculated "Ciudadanos que votaron" are recorded on the official tally sheet.  The total number of votes being reported on the tally sheet should add up to the number of "Cuidadanos que votaron" but very frequently it does not. Based on my experience of trying to review the results, minor errors of 1, 2, and 3 over-votes are common, while over-votes of 50 or more happen less often. 

Reviewing and recounting the Actas alone will not correct these over-votes.  They merely become  enshrined in the result.

The public vote count shows results that differ from the TSE count, though not enough to change the outcome of the election.  But there are still 4.4% of the Actas which cannot be validated because the TSE released no image of them. This is enough to affect the margin between the two leading candidates, which might reflect the tighter race that most observers expected.

Then there's the issue of database security.  Anonymous Honduras has twice penetrated the vote counting center, and currently (late afternoon on December 2) has replaced the TSE's main web page with their own.  Their penetration made it clear they could have easily, and invisibly, changed the results in favor of any candidate they wanted to.  From details like their ability to show administrative tables, it seems that they had complete control over the database, and the TSE was apparently none the wiser.

So, there will be some sort of a recount of the Actas, with representatives of the political parties present to agree that the data entered into the system is what is on the Acta. The TSE has no idea what the procedures will be, or how they will do this, but something will happen. 

It's a step in the direction of transparency. But not the kind of recount that would put to rest, ultimately, the kinds of doubts that have been raised.

TSE Vote Counting Errors

The OAS audit of the TSE's software was accurate in its concerns: the vote counting part is turning out to be less than perfect.

Observers have noted several anomalies, some of which have been corrected, and others of which have not.  Here's an example that was still visible in the system as of this morning.

Consider the following screenshot (from the TSE's website here) taken at 1:19 pm today.  Click on the image for a larger version.

This shows the scanned Acta for MER 3311 of the Escuela Leonardo Aguilar Oseguera school voting site in San Pedro Sula. On the left is the scanned Acta; on the right are the vote totals that the TSE credited to each candidate for this MER.

Precisely how those vote totals enter the system is unclear from TSE-provided information.  Are the handwritten Actas submitted to an automatic optical character recognition system (ReadIRIS was mentioned in one TSE story about the vote counting system) or are the tallies entered by someone reading and entering the values for each candidate?

In either case, the vote total listed for Nasralla on the TSE website is only 3 votes, not the 103 votes listed on the tally sheet.

We are not alleging any mischief here (although that's possible). Rather, this is precisely the kind of sloppiness the OAS audit of the software used led us to expect.

This Acta should have failed the software confidence checks (because the number of votes in the tallies does not match the number of votes allegedly cast in this MER). But it either wasn't flagged by the system supposed to catch such inconsistencies (it should have detected as an under vote), or else a human intervened and OK'd the result: human error.

In either case the erroneous numbers were submitted to the vote tallying system with these inconsistencies and were tallied, giving Salvador Nasralla of the PAC party 100 fewer votes than he actually received from this mesa.

These are all officially approved tallies, officially approved errors, not the Actas sequestered for inconsistencies. An Acta does not appear on the website if it hasn't been included in the vote totals, so the TSE thinks there's no problem with this one.

We beg to differ. 

This is just a single example of the kinds of errors we are seeing in the vote totals publicly provided by the TSE.  The sad part is that it is not unique. There are many similar kinds of errors visible in the Actas posted on the TSE website.

The software audit by the OAS said the vote totaling system did not meet international standards for such systems and needed a lot more work (and a reimplementation) to do so.  It looks like the TSE went ahead and used this poor quality software, which would mean it should independently audit all of the Actas and vote totals for consistency and sanity (meaning, are the numbers reported consistent with the number of voters in a specific mesa) before declaring anyone a winner.

Other errors that were apparent earlier are now systematically being corrected, so it will not surprise me if by the time this blog entry is published, the errors I know about (including the one above) have been made to disappear.

But it is these kinds of failings, and silent corrections without acknowledging the failings, that reduce public confidence in the TSE and its official results, leading people like Salvador Nasralla and Manuel Zelaya Rosales, ask for a public, open, recount of the official Actas before the TSE declares a winner.

A public recount of the Actas would go a long way to providing confidence that everyone is adding up the same numbers to get the same vote totals, which frankly would improve public perception of the TSE.

Without that, no matter how many international observers say the process on Sunday was legitimate, there will remain large segments of the Honduran people who believe that the process after Sunday was not.