HALAL Letter to the COMELEC

20 June 2008

The COMELEC En Banc

Palacio del Gobernador

Intramuros, Manila

Attention: Comm. Rene Sarmiento

Dear Sirs,

I am glad to inform you of the successful conclusion of the research fellowship on automated elections, which I conducted on Apr. 21 – May 30, 2008 at the Oxford Internet Institute of the University of Oxford in the U.K.

I had informed the COMELEC en banc of this research when I was invited on Apri 15, 2008, Tuesday, to present to the En Banc the Halalang Marangal recommendations for improving the integrity of the next elections, and the En Banc had indicated its willingness to review the results of my research.

Enclosed are some of the papers I wrote or updated during my research at Oxford:

  • HALAL Working Paper No. 1: “Double-entry accounting can improve both automated and manual elections”

  • HALAL Working Paper No. 4: “Automating elections: computer have made mistakes too”

  • HALAL Working Paper No. 6: “The Cost of Automating Elections”; and

  • HALAL Working Paper No. 7: “Detecting automated election errors through statistical sampling”

Given our findings about the occurrence of errors in automated elections, some of which could have changes election outcomes, may I reiterate our recommendations for the use of double-entry accounting in the next elections as well as the implementation of a post-election statistical audit as described in our Working Paper No. 7 above.

We would be very glad to meet with the COMELEC en banc or with individual commissioners to answer questions regarding our proposals.

Respectfully Yours,

Roberto Verzola

Secretary-general, HALALANG MARANGAL

Tels. 371-2107, 0929-856-1930, Fax: 372-4995

Advertisements

HALAL papers submitted to the COMELEC

HALAL Secretary-general Roberto Verzola spent six weeks in April-May 2008 on a research fellowship at the University of Oxford Internet Institute in the U.K., where he studied the experiences of other countries in automating their elections.

As a result of his research, he wrote four papers which HALAL later adopted as HALAL working papers and which we later submitted to the Comelec. We also appeared in two Comelec en banc meetings to present the papers.

You may download the HALAL papers here:

Among all our recommendations, the only one which the Comelec seems to have adopted was the recommendation to stay away from direct-recording electronic (DRE) machines like touch-screens and touch-pads.

The most important step in modernizing election tallies: double-entry accounting

The most important step in modernizing election tallies might not be automation, but the use of double-entry accounting methods to make it easier to detect and locate errors and other anomalies. This is my conclusion after doing two election audits: the 2004 presidential election audit, which I did in my personal capacity, and the 2007 mid-term elections, which Halalang Marangal, where I serve as secretary-general, conducted.

The use of single-entry accounting is a fundamental weakness in current election tabulation systems. If we automate the election tabulation at this point, we will be automating a system that is basically flawed, and therefore wasting precious resources on a project that would eventually have to be redone.

Our proposal to modernize our election tabulation through the use of double-entry accounting is contained in a paper (“Modernizing election tabulations through double-entry accounting“) we are submitting at the 2007 Summit on the 2007 Elections, which the Comelec will convene on Sep. 10-12, 2007.

Roberto Verzola, Secretary-general, Halalang Marangal

The “hanging chads” in Florida which worked in Bush’ favor in the 2000 U.S. presidential elections were intentional! –Dan Rather Reports

George W. Bush won the 2000 U.S. presidential elections versus Al Gore because he won in Florida. Florida’s elections were marred by undervotes (abnormally low number of voters casting valid votes for president), miscounted votes (vote for one candidate assigned to a different candidate), and other problems.

The undervote problem was traced to “hanging chads” or pieces of the ballot paper that stuck to the punch card ballots instead of detaching themselves, fouling up the automatic counting process. By reducing the voter turnout in Democratic areas, these problems worked in favor of Bush.

According to Part 3 of an investigative report by Dan Rather aired last week in the U.S. (see the transcript here), the “hanging chads” were due to sub-standard paper which were used for the punch card ballots “for the first time” in 2000. In his report, Dan Rather leaves no doubt that the shift to sub-standard paper was intentional and that the punch card ballot manufacturer Sequoia Voting Systems was warned by its employees well beforehand that the sub-standard ballot paper would cause problems.

For interpretations of Dan Rather’s investigative report on this matter, search the Web for “Dan Rather on bad punch card ballots”. You can find one very good interpretation here.

Why is this relevant to Philippine elections? Because the Comelec is due to automate our elections in 2010. The public has to be warned that automating the elections is no guarantee of clean and honest elections. Depending on the approach, automating the elections can make things worse.

While many factors come in, we feel that the most important one is the issue of transparency: will a particular automation proposal make the counting and canvassing process more — or less — transparent? Most automation proposals we have seen reduce the transparency and will take away from the ordinary citizen the ability to audit the vote-counting process in real-time, which we are able to do today under the manual method of vote-counting at the precinct level.

Any automation project should preserve this ability and further enhance it, instead of taking it away altogether.

Recommendations of the HALAL Citizens Audit Report #4 on the 2007 Senatorial Elections

RECOMMENDATIONS

Based on the results of its 2007 citizens’ audit and its experiences in conducting the audit, Halalang Marangal (HALAL) hereby recommends the following measures to correct the problems of the 2007 elections and to ensure that the next election will be conducted in a cleaner and more credible manner:

  1. Identify and bring to justice all who were involved in the Maguindanao election fraud. The Maguindanao election fraud should be thoroughly investigated and those found responsible should be brought to justice, including those who falsified or who forced Comelec officials to falsify results and official documents; those who hid, stole or destroyed the original documents containing the true results, if any; those who ordered the murder and/or kidnapping of election officials who exposed the fraud; those who, despite the statistical and circumstantial evidence, officially accepted these results and foisted them on the Filipino people; and those who offered to pay or otherwise reward local officials deliver them the votes, regardless of the true people’s will.
  2. Exclude implausible, impossible results. In the case of highly implausible results, the burden of proof should be on Comelec officials and others who assert that these numbers are truthful. In the meantime, while the investigation is going on, the implausible results should be excluded from the official 2007 results. The statistically impossible results should be excluded forthwith from the official results. On the concern that this will disenfranchise the voters in those provinces and cities, our answer is that they were already disenfranchised by those who manufactured these results. Excluding statistically impossible results simply prevents the further disenfranchisement of those whose votes would be negated by the falsified returns.
  3. Comelec retraining. All Comelec officials should undergo retraining on the importance of the required statistics in Comelec forms, especially the number of voters who actually voted and the number of precincts actually tallied. The latter should be the sum of the same statistic from lower levels, and not simply copied from the total number of precincts in the province. No ER, SOV or COC should be accepted or submitted for further canvassing unless all information required are filled in. Strict disciplinary measures should be taken against Comelec officials who submit or accept incompletely filled-up forms. Comelec officials who do not know how to add should be summarily dismissed from the service.
  4. Redesign Comelec forms. The SOV should be redesigned to include the number of ERs actually canvassed in the same column as the number of registered voters and the number of voters who actually voted. The COC should also be redesigned to include the number of voters who actually voted in the first page of the COC. Because they can be looked up from existing Comelec records, the number of registered voters and the total number of precincts are non-essential information and may be omitted from the forms.
  5. Provide certified true copies of ERs, SOVs and COCs to all who want them. Requiring that a copy of the ER be posted in every precinct was a good step in transparency though it was not observed in many instances. Likewise, the provision in the new automation law for 35 copies of the ER to be made available was a good transparency measure, though it was not implemented. HALAL recommends that ERs be encoded into computers as soon as they are brought to a municipal canvassing center. HALAL recommends, once the ERs are encoded, that original printouts be made available at cost to any interested party, during the municipal canvassing and afterwards, as certified true copies of these ERs. Similarly, original printouts of SOVs and COCs should be made available at cost to any interested party, during the municipal canvassing and afterwards, as certified true copies of these documents. With so many copies of results circulating, tampered or manufactured election documents can be easily exposed. HALAL further recommends, once ERs are encoded, that all subsequent consolidations, additions of row and column totals, computations of indicators like voter turnouts and ballot fill-up rates, and the printing of certified true copies of SOVs/COCs be done with computers. This will minimize clerical errors arising from mistakes in copying numbers or adding them, improve the legibility of the documents, and facilitate the printing of an unlimited number of copies of official election documents for any interested party.
  6. Put all ERs, SOVs and COCs online. All ERs, SOVs and COCs should be put online or at least be made available to any interested party, together with their file checksums, at no additional cost, making the entire tabulation process completely transparent.
  7. Retain parallel counts in subsequent elections. As this audit shows, a parallel count facilitates the monitoring of the official canvass and the investigation of electoral fraud. We recommend that parallel counts be made a continuing part of the canvassing process, even under a regime of automation. Automation will make it easier to provide not just a single accredited group but any political party, poll watching group, media, or anyone at all with copies of ERs, SOVs and COCs, who may then may do their own parallel count. This way, automation will enhance transparency and enable more people to participate meaningfully in poll watching, instead of the typical automation goal to “minimize human intervention”.
  8. Retain manual precinct-level counting in the next elections. HALAL further recommends that the manual counting of votes at the precinct level be retained in the next election. The manual counting of votes at the precinct level is superior to automated counting in terms of openness, transparency, and providing an invaluable lesson in civics to all participants and witnesses. Because it is non-automated and open, it is slow enough that any citizen can actually audit in real-time the counting process. But because of its massively parallel approach of simultaneous tallies in all the two hundred thousand plus precincts in the country, tt is also fast enough that results are usually in within 6-12 hours.
  9. Automate the printing, public distribution and consolidation of results. HALAL suggests that the following phases of the canvassing process be automated in the next elections: the on-site printing of ERs, SOVs and COCs, so that anyone can get official copies of these documents; the consolidation of ERs and COCs into SOVs and the generation of COCs from SOVs; their public distribution through the Internet/Web; and the generation of the final results. Such automation can be done at relatively low cost by providing at least two computers with printers and battery backups per municipality (per district in chartered cities), together with operators familiar with spreadsheet programs. In the next election, official municipal results should be printed out and at the same time saved on USB sticks and put online, where these can again be printed out upon receipt at the provincial level for double-checking before they are consolidated into provincial results. A similar procedure can be followed for national consolidation. By making all intermediate results available to the public at every step, automation can enhance the transparency and therefore the credibility of the whole process, at the same time enabling any citizen to participate more fully in the electoral process.
  10. Municipal-level audit by Comelec. Finally, HALAL recommends that the Comelec conduct a similar audit comparing the Comelec and NAMFREL results at the municipal level. HALAL offers to help the Comelec conduct such an audit. This official audit can provide the basis for determining where fraud was actually committed, gathering the evidence, and pinpointing culpability. Eventually, administrative and/or criminal charges should be filed against those who are found responsible for the fraud.

Halalang Marangal 2007 Citizens’ Audit: Report #4

HALAL released on July 19, 2007 its audit report #4, covering the 2007 senatorial elections. To get the full text of the report, see the following link:
Halalang Marangal 2007 Citizens’ Audit: Report #4

Findings of the Halalang Marangal Citizens’ Audit Report #4 on the 2007 Senatorial Elections

FINDINGS

The findings of the HALAL audit may be summarized as follows:

  1. Essential information such as the number of precincts actually tallied and/or the number of voters who actually voted were missing or plain wrong in the election reports of one-third of all the cities and provinces canvassed, indicating laxity bordering on negligence and/or incompetence on the part of municipal, provincial and national Comelec officials who prepared, submitted or accepted those reports.

  2. There were statistical and circumstantial evidence indicating that the results from Maguindanao were fraudulent, yet these were accepted by the National Board of Canvassers as part of the official results.

  3. The voter turnout was abnormally high in at least two provinces: Maguindanao (93.5%) and Sharif Kabunsuan (91.5%) and some municipalities of other ARMM provinces.

  4. The ballot fill up rate was statistically impossible in at least four cities: Quezon City (15.3), Mandaluyong (13.0), Pasig (12.6), and Paranaque (12.6) and in 6 of the 22 municipalities of Maguindanao.

  5. Significant changes in rankings suggest anomalous provincial/city Comelec results in at least twelve provinces and cities: Maguindanao, Sharif Kabunsuan, Tawi-tawi, Basilan, Lanao del Sur, Kalinga, Davao del Sur, Apayao, Sulu, Sultan Kudarat, Pasig, and Occidental Mindoro.

  6. Compared to 2004, vote discrepancies in 2007 were lower in 25 provinces/cities including some ARMM provinces, but higher in 73 provinces/cities.

  7. The ten biggest net gainers from the discrepancies were Recto, Villar, Zubiri, Escudero, Pangilinan, Pichay, Defensor, Sotto, Arroyo, and Singson, in that order.

  8. The biggest loser was Kiram, who lost votes even in his home region ARMM as well as in other provinces where his TU partymates gained votes.

  9. The votes for minor candidates (Kapatiran, KBL and Cantal) were shaved nationwide.

  10. Among the major candidates, the other big losers were Aquino, Montano, Osmena and Coseteng.

Automating the counting of votes can make things worse

Automating the counting of votes can make things worse
by Roberto Verzola, Secretary-general, Halalang Marangal

[A pioneer in the local desktop computing and Internet scene, Roberto Verzola built the “first Filipino computer” in 1982, set up the first online systems at the Senate and House of Representatives in 1991, and was awarded by industry the title “father of Philippine email”. One of the convenors of a new election watchdog called Halalang Marangal, he can be reached at rverzola@gn.apc.org]

Current proposals to automate our elections can make things worse instead of better, because the counting becomes less instead of more transparent. The misfocused objective of “minimizing human intervention” will result in fewer and instead of more witnesses when fraud does occur, making it easier for cheats to cover up their crime once they break the system.

Consider the basketball game scoreboard. The scoreboard is updated manually, with chalk for blackboards, sign pen for whiteboards, and push button for digital displays. Manual, but satisfactory. Suppose someone proposes to “minimize human intervention” by putting on the hoop an electronic detector that automatically updates the scoreboard every time the ball falls into the hoop. There is one catch: the audience will see the scores only at the end of the game, not each time a score is made. Surely, such automation will be unacceptable to basketball fans. Consider another example: the taxi meter. Would you prefer the computerized meter to a mechanical one, if the computerized display showed you the fare not while it was updating itself along the way, but only upon reaching your destination?

Many election automation proposals today are of a similar nature: the public will lose its chance to witness the votes as they are individually counted, and will only be shown the totals when the automated canvassing is over.

If this loss of transparency is unacceptable in a game or a taxi meter, then more so in an election.

Many automation proposals are based on flawed assumptions:

1. Automation will eliminate human intervention. It will not. Automation can only reduce, but never eliminate, human intervention. Automated systems will always have points of human intervention: the programmers updating the software; the technicians maintaining or repairing the machine; the staff feeding the ballots to the machine; the staff handling the final output; etc. Reducing human intervention can actually work in favor of the cheats, who will now need to recruit fewer accomplices and deal with fewer potential witnesses to the fraud.

2. Automation will minimize if not eliminate cheating. It can do no such thing. If they work as intended, automated machines can only: a) speed things up, and b) follow more faithfully the instructions of those who program them. If they are reprogrammed to cheat, the machines will follow the new instructions just as faithfully and quickly. If you want examples, just search the Internet for the keywords “U.S. automated election fraud”.

3. Safeguards can prevent cheats from manipulating an automated system. This is an illusion. Cheats can master automation technologies as well as anybody else. Sooner or later, they will be able to identify the system’s weak points and break it. I have worked with automated machines at the level of machine language and individual chips; at this level, many things are possible.

4. The main cause of cheating is the slow manual count. This confuses the symptom for the disease. Obviously, cures based on wrong diagnoses will probably be wrong too.

In fact, the precinct-level manual counting of votes is not slow. In most precincts, it is over within several hours. More than that, the precinct count is the most transparent part of the whole process. Here, like the audience in a basketball game, the public can see each vote counted and the candidates’ score updated, vote by vote. Beyond that, the precinct count is an invaluable lesson in civics for our youth. Here, parents can bring their children and watch democracy – or tyranny – in action, live. Cheating that occurs at this level often involves brazen, in-your-face kind of acts that no machine can stop and no cheat can hide.

In truth, it is not the slow count that leads to cheating but the other way around: it is cheating that leads to a slow count. especially at the municipal and provincial levels. The slow count is a symptom, an effect, of the disease. It is cheating, the disease, which causes the slow count. The real cure for cheating is to punish the cheats. Our laws say that an election cheat shall be barred from holding public office for life. That alone, plus the jail terms, will make our elections clean and honest. Sadly, the 2004 election cheats went scot-free, kept their positions, or even got promoted.

Should we junk election automation then?

No, but automation should be used mindfully, not to minimize human intervention but to maximize transparency, or the ability of interested third-parties and the public in general to double-check and audit the system.

A very good example of enhancing transparency is Senator Serge Osmeña’s proposal to use digital imaging technologies to make more copies of the election return (ER) at the precinct-level. The more copies of the ERs circulate, the more difficult for cheats to cover up their crime.

Systems expert Manuel Alcuaz’ proposal for LCD projectors at the municipal level is also a good example. Projecting ERs on a big screen enables more people in the audience to audit the ongoing canvass.

Halalang Marangal’s proposed citizen’s tally is still another example. We will ask non-partisan volunteer precinct watchers to use cellphones to text the results to our databases, where the 250,000 or so precinct results will be stored. The individual precinct results can then be requested by the public by text/SMS or via the Internet. Anyone can also download the entire database or get it on a CD.

This low-cost, highly transparent approach will empower ordinary citizens to audit the results of the elections: they can now go to a voting center, watch the precinct count, jot down the results, and compare the results posted on the database with what they saw with their own eyes. If they have a computer, they can even get the CD version, or download the database from the Internet and do their own citizen’s tally.

Once they adopt “maximizing transparency” rather than “minimizing human intervention” as objective, technical experts can no doubt come up with even better schemes.

Unfortunately, the attention of Congress has been focused on hardware-intensive proposals that will not only waste our scarce resources automating the most transparent portion of the whole electoral process but may even make it easier for future cheats to cover up their manipulation of the results.

PCIJ Report on Teletech, voting machine assembler for ES&S

Teletech is the company operating in the Philippines assembling voting machines for ES&S. According to Dan Rather, these flawed machines might have altered the results of elections in the U.S.

  1. The Philippine Center for Investigative Journalism (PCIJ) has written an investigative report on Teletech. (See PCIJ report here.)

Brief Description of ES&S proposal to Comelec

According to the Comelec, the ES&S proposal to automate the next Philippine elections can be briefly described as follows:

  • One of the largest companies in the industry
  • OMR based voting/counting machine
  • Includes transmission facilities via modem
  • Machines are made in the Philippines
  • Flexible acquisition options

The ES&S proposal was submitted November 2006. In his recent HDnet show, Dan Rather exposed the flaws of ES&S voting machines made in Philippine “sweatshops”.