The transparency issue

Editor’s note: The Manila Times today publishes the third part of the serialization of Mr. Monsod’s assessment of the May 2010 automated polls. The assessment was delivered on November 9, 2010, at the Ateneo de Manila Rockwell campus in Rockwell, Makati City, as a lecture, under the Jaime V. Ongpin Foundation Lecture Series.

Third of six parts

I am convinced that the Comelec (Commission on Elections) did not start off to be non-transparent in implementing (poll) automation. Its good faith is not in question here. But the Comelec under Chairman Jose Melo was single-minded about automation, which was its greatest strength in making automation happen but it was also its greatest weakness. It soon found itself on a slippery slope as it gave up more and more decision-making powers to Smartmatic for lack of organic technical expertise. (The Comelec Advisory Council cites at least four violations of Smartmatic that should be penalized i.e. CF [Compact flash] card mistake, UV safeguard, time stamps, transmission underperformance.)

Moreover, speed and not accuracy became the overriding consideration, even if it meant ignoring legal provisions, such as Section 197 of the Election Code on the right of voter to a second and third ballot, disabling the feature of a voter verified paper audit trail and dispensing with certain grounds for pre-proclamation proceedings. When the automation schedule seriously faltered, it then waived, ignored or rendered inutile certain legislated safeguards and contractual requirements and de-prioritized other electoral reforms necessary to improve the system itself. When the Comelec ran out of explanations and the questions and criticisms persisted, it simply shut down communications and thereby heightened the mutual distrust. Frustrated IT professionals and experienced election watchdogs (except the Parish Pastoral Council for Responsible Voting, or PPCRV) interpreted the Comelec approach as, at the least, high-risk that put the elections in jeopardy. Or at the worst a conspiracy to make them fail or to allow cheating to take place.

As already mentioned, I believed then, and hindsight tells us now, that the scenario of vast and elaborate conspiracy did not make sense. But the Comelec did commit many mistakes and lapses of judgment, as did Smartmatic, to the point of a potential disaster, as exemplified by the CF cards fiasco during the FTS (field testing and sealing) 4 days before the elections.

We now know that if the reprogramming of the CF cards was just one day delayed, there would have been many precinct failures on election day.

In short, the Comelec was incredibly incompetent but it was also incredibly lucky. But the cheaters will not be in shock and awe next time and they will be more prepared. And the margins may be small enough to change.

The auditability issue

The Comelec has not issued its report on the 2010 elections. The congressional investigation in aid of legislation has not even started and the clock is running. When I asked the career people in the Committee on Suffrage if there were any instructions or framework for assessing the automated elections, I got this response: “Kayo naman sir, napakataas ng hinihingi ninyo. Parang pinapakain ninyo ng lobster ang bagong baby [Sir, you are asking too much, it’s like feeding a new born baby with lobster].”
Whenever the possibility of fraud was raised before and after the elections, Smartmatic’s ultimate answer was that any fraud or even attempt at it can be detected or found out sooner or later, because of the auditability of the system. Let us validate that assurance on four points—the “compensating factors,” the random manual audit, the logs and data from the central server and CF cards and the resolution procedures in protest cases.

1.The compensating factors – I believe that the source code is clean based on the certification of the AES (Automated Election System) by the Technical Evaluation Committee (TEC), one of the few sources of forthrightness on the side of the government. I also inquired from my sources about SysTest (system test) and got a favorable assessment despite its checkered history. But the TEC’s endorsement was conditioned on the adoption and implementation of 30-compensating factors to assure that the AES can securely, accurately and properly be used in the elections, and the additional safeguards mentioned earlier. Unfortunately, the TEC says that there is no feedback or documentation if, in fact, those were all adopted or implemented, except an en banc resolution (from Comelec) directing the Project Management Office dated March 11, 2010 to do so. Thus, a documentary trail to determine what compensating factors and safeguards worked or did not work is not yet available.

2. The random manual audit – At the outset, I must say that, despite my earlier support for a proposal for a parallel manual count, I did not go along with the final appeal because I think that the Comelec had valid reasons for its denial, i.e. the debatable assumption that a manual count is always more accurate than an automated count, and the confusion that might result from it and the lateness of the hour to implement it. But the main argument of the Comelec against the parallel manual count was the random manual audit (RMA). And that, I am sorry to say, was a failure.

In the first place, the chairman of the TWG-RMA [Technical Working Group of the Random Manual Audit] is a statistical illiterate like me and should have realized her limitations. Second, she compromised on the use of the RMA for proclamation purposes against her earlier commitment, in the interest of a harmonious relationship with the Comelec. Maybe it is just as well that happened given the poor quality of the RMA and its delayed completion. Third, there was no real statistician on the project except NSO [National Statistics Office] Administrator Carmelita Ericta. But she was invited to join the committee very late, after the sampling design and audit procedures had been decided. The NSO vouches for the correctness of the “edited” results (the raw data had to be edited because of the many mistakes by the teams mainly because of lack of training), but the NSO does not vouch for the adequacy of the sampling design nor the reliability of how the audit was carried out in the field (many were not done at the precinct, nor on Election Day, nor observed by the parties for lack of notice, nor were all the precincts covered by an already small sample). The RMA, released only end-July (2 + months after the elections), showed that the 99.9995-percent accuracy rate commitment of Smartmatic was not met, but it is really useless for its intended purposes except to demonstrate how not to do it the next time around.

3. The logs and data stored in the PCOS (Precinct Count Optical Scan) CF Cards and in the Central Server are supposed to provide a complete documentation of what happened in the elections. These are not yet available to date, which may account for the denial of Cenpeg’s (Center for People Empowerment in Governance) request for information. Moreover, as of last week, 11,021 CF cards were not yet in the custody of Comelec Central. The Comelec did not start to run the logs until last week with the help of the DOST [Department of Science and Technology] because, I was told, certain contractual provisions about a one-time use of the Smartmatic servers and software needed to be resolved. Maybe the Comelec can enlighten us more on the issue.

In any case, the data in the Central Server is incomplete because out of 76,347 precincts, only 68,670 ERs [electoral returns] were transmitted to the Central Server, 450 ERs are “unaccounted” for (998 until last week), and there are 251 “failures of elections” which have no ERs. The rest, 6,428 ERs, were transmitted to the municipal board of canvassers. Incidentally, it still remains to be verified if the 251 “failures,” the 262 precincts with FTS data, the 450 “unaccounted for” precincts and the vote balances of the lowered thresholds for transmission were included in the presidential and vice presidential canvassing.

The 90.35-percent unofficial electronic count and the delayed reconciliation of data sent to the three servers would have been a significant underperformance of transmission, and would have been useless had the presidential elections been closer. The PPCRV parallel count was also useless. It only encoded 53,000 ERs equivalent to a 69-percent accomplishment rate, the lowest in the history of unofficial parallel counts in the country.

It may well be that the audit logs of the Central Server as validated by the data in the CF cards will ensure the auditability of the automated elections. But to do that, the logs, data and CF cards have to be secured in the first place.

4. Recount in election protests – The elections are not complete until the rightful winners are proclaimed. Thus, protests represent an opportunity to test the auditability of the results. But todate, the rules on the recount have still to be agreed upon. I agree with the latest proposed guidelines of the Comelec that the recount should be done on the same basis as the elections, which would include rerunning the ballots through the PCOS machines and the use of the ballot image as primary evidence. A reversion to the old manual “revision” of ballot, as proposed by election lawyers, would mean that the criteria for deciding protests would be different from the criteria to determine other winners in the elections. That would be legally assailable. It is the jurisprudence that should adapt to the new system and not the other way around. In fact, automation is supposed to be so convincingly accurate that the Comelec thought it would result in a 99-percent reduction in protests, at least that was what a former commissioner said in a recent Cebu focused-group discussion among judges and lawyers. How explain then that the number of protests at the Comelec and the House Electoral Tribunal (HET) is higher in 2010 than in 2007—96 vs.73 in the Comelec and 62 vs.49 in the HET. There is also the protest of former Sen. Manuel Roxas 2nd (who lost in the race for vice president) that asks for a forensic examination of the whole system before any recount is started.

In view of all the foregoing, our only verdict on the Smartmatic assurance of a complete auditability of the system is a failing mark at this time. –Christian S. Monsod, Former Chairman Commission on Elections, Manila Times

To be continued