[Note: My apologies for the formatting concerns that I could not address. I can provide the full tables upon request.]
I wrote a piece on Murphy’s Law shadowing every step of the automated election, with particular emphasis on the voter’s interface with the technology. I will revisit these issues, plus the PCOS and queuing, ex-future manual counts as well as provide some assessments coming from other citizen groups on the other aspects of this brand of technology chosen by the COMELEC.
1. Functional Literacy of Voters
At least 15 percent of the population 10-64 years old is not functionally literate. A functionally literate person is one who can read, write and compute or one who can read, write, compute and comprehend. I posited that with lack of adequate voter education, the traditional ballot wherein a voter fills in the blanks with the names of their chosen candidates would be ‘friendlier’ even for those who can only read and write. With the new ballot, problems such as overvoting [voting for more than the required number of candidates for a particular position] and improper shading are bound to arise.
In an interview by Manolo Quezon of Anna Tabunda of Pulse Asia last May 3, http://www.quezon.ph/2010/05/09/lost-in-the-counting/ she related that she went through the sample ballots respondents used during the surveys to find out the incidence of improper shading and overvoting.
- With regards to improper shading, 12% nationally or 4 million votes will have difficulty with the ‘new’ ballot in actual conditions. Particularly high incidents of improper shading was registered among older voters and in Regions 1, 2, 3, 4A, 5, 6, 7, 8 and 12. This finding is potentially very significant for candidates strong in the Visayas.
- More or less 2% of presidential and vice-presidential votes suffered from overvoting, which would void the votes cast for the positions in actual conditions: potentially 400,000 votes based on an 80% turnout rate.
- Most crucial was her finding that 3% of senatorial votes represented overvoting. This could have drastic results for the actual standing of candidates.
2. Printing and Distributing the Ballot
The printing of 50 million ballots with more than 1600 varying candidate sets [combination of national and local candidates] is a momentous challenge even for the National Printing Office. Can you imagine the distribution system that would deliver the right ballot forms to the right legislative districts, with no room for error? 99 percent accuracy in the delivery of ballots is not acceptable since voters in one district or municipality cannot vote for their candidates in ballots intended for another place.
- Samar ballots wrongly delivered to Iloilo. [via @sunstaronline] [14:48]
- Voting halted in Iloilo town due to wrong ballots http://bit.ly/cSis1b [via @inquirerdotnet] [14:20]
3. Filling up the Ballot and Feeding into the PCOS Machine
The size of the ballot, according to COMELEC, is “8 ½ inches by 26 inches [lately trimmed down to 25 inches] and some 50 million. The ballots would be color-coded and printed back-to-back with the names of the candidates. The national candidates will be found on the front, while the local candidates will be on the back.
OMR technology is used to detect shaded ovals or boxes that correspond to answers (or votes for a particular candidate). But shadings should completely cover each distinct oval or box and should not stray to the adjacent ones; otherwise the machine will reject the form or ballot.
|Timestamp||Name||Precinct Number AND Cluster Precinct #||Name of School, Municipality||City / Province||Problem encountered|
|5/10/2010 0:42:47||Bt||Precinct # 02224 cluster # 64||Dasmarinas village, makati||Makati||Jammed machine!|
|5/10/2010 0:58:06||DZMM||[none specified]||Canoan Elementary School, Canoan||Estancia, Iloilo||3 PCOS broken after paper jam but one was not fixed.|
|5/10/2010 14:04||pinoyweekly||[none specified]||AbraDe Ilog. Mindoro Occidental||Bagong balota may naka-shade nang partylist, ulat ng Katribu Partylist pollwatchers|
|5/10/2010 14:17||workerswatch||[none specified]||Cebu||Voter refuses to accept shaded ballot; BEI fed it to the PCOS before giving the voter a clean ballot|
|5/10/2010 15:41||mindanaoan||[none specified]||Malaybalay City Bukidnon||lots of ballots got rejected. unfortunately, the voters have ALREADY PLACED THEIR THUMBMARKS|
|5/10/2010 16:00:26||bulatlat||[none specified]||Taboc, San Juan||La Union||ballot stuck inside machine|
|5/10/2010 16:04:03||gmanews||[none specified]||Tarlac||Tarlac||PCOS machine stopped while reading ballot. machine shut down|
|5/10/2010 16:13:31||workerswatch||[none specified]||Bagong Silangan||Quezon City||PCOS machines in 3 precincts reject ballots, voters told to return later|
|5/10/2010 16:35:16||pinoyweekly||[none specified]||Bagong Silangan||Quezon City||ballot too wide, margins were cut to feed it to machine. happened in 4 precincts|
|5/10/2010 16:43:59||spirodon||[none specified]||Holy Spirit High School||Manila||paper jams in PCOS machines|
|5/10/2010 16:45:56||edgyboi||[none specified]||Pagbilao||Quezon||several ballots were pre-shaded, PCOS seal tampered|
|5/10/2010 17:23:42||gmanews||[none specified]||Pembo Elementary School||Makati||PCOS glitch due to paper jam|
|5/10/2010 17:40||mindanaoan||[none specified]||precinct 228 kauswagan||CDO||voter complains that he received pre-shaded ballot|
|5/10/2010 18:04:43||mindanaoan||[none specified]||Naawan||Misamis Oriental||PCOS machines reject ballots. BEIs realized that ballots were too wide and had to cut them|
|5/10/2010 18:16||coderadiophils]||[none specified]||Bgy. Burgos-Padlan, San Carlos City||Pangasinan||may ballots na may shade na bago pa man bumoto ang voter -EO|
4. PCOS Machines and Queuing Issues
Optical mark recognition (OMR) is the scanning of paper to detect the presence or absence of a mark (e.g. shaded box) in a predetermined position. It evolved from the IBM punch cards of old where punched holes were sorted to represent information. OMR is the simplest of commonly available form processing technologies (e.g. Lotto tickets). OMR equipment has been available for many years and has nowadays reached very good levels of reliability. But OMR has relatively stringent requirements for the successful processing of the paper forms. Thus, countries with very dusty or humid climates and poor transport infrastructures are discouraged from using OMR. Special questionnaire/ballot design restrictions include the quality of the paper along with precise specifications regarding the printing and cutting of the sheets.
Precincts have been re-configured into clusters of one thousand (1000) registered voters who will go through one (1) PCOS machine. With voting hours from 7:00 a.m. to 6:00 pm, there would be theoretically 1000 voters to use 1 machine in eleven (11) hours. Note that only the list was expanded but the polling place, usually a public school classroom, – that is, in the past for two hundred (200) voters -. remains the same but this time for 1000 individuals.
Example from former Department of Education Undersecretary Mike Luz:
Finally voted in Bacolod as number 303 of about 1000 in Cluster 286 after a wait of over 4 hours (1:50 p.m.). Earlier was given waiting number 418. If only 303 of 418 (up to his earlier assigned number) voted, then 115 must have gotten tired of waiting and left. That’s a 27.5 percent drop-out rate. If this is a typical take-up rate (72.5%) all over the country, this election could have a lower voting rate than in the past.
|Table of Voters’ Messages to Election Watch on http://bayanihan0nline.wordpress.com|
|Timestamp||Name||Precinct # & Cluster Precinct #||Name of School, Municipality||City / Province||Problem encountered|
|5/10/2010 0:31:02||coderadiophils||[none specified]||Juan Sumulong Elementary School||Antipolo||machine not working|
|5/10/2010 0:32:04||Bian Villanueva||[none specified]||Hacienda Luisita||Tarlac||2nd machine that broke down as of 2:39pm|
|5/10/2010 0:33:32||workerswatch||[none specified]||Gulod, Novaliches||Quezon City||another PCOS machine fails|
|5/10/2010 0:55:26||ayliya||Precinct 0384 A||Cainta||Rizal||defective and not yet replaced as of 10:51am|
|5/10/2010 0:56:24||popisunga||[none specified]||Las Piñas||Metro Manila||PCOS malfunction. delayed voting for 2 or so hours|
|5/10/2010 16:01:15||briankingong||[none specified]||Lawa, Meycauayan City||Bulacan||3 out of 16 PCOS machines malfunction|
|5/10/2010 16:04:54||kontradaya||[none specified]||Brgy. Guitnang Bayan 1, San Mateo||Rizal||PCOS machine broken. Only 9 people were able to vote as of 9am|
|5/10/2010 16:05:21||kontradaya||[none specified]||Muntinlupa||Metro Manila||faulty machines|
|5/10/2010 16:08||NBN_Ch4||[none specified]||[none specified]||Masbate||2 PCOS machines caught fire during elections|
|5/10/2010 16:14:22||pmcalara||precinct 36||San Francisco, Tigaon||Camarines Sur||PCOS machine not working|
|5/10/2010 16:15:10||briankingong||2106A-494||District 2||Quezon City||PCOS machine busted|
|5/10/2010 16:15:53||briankingong||2106A||Quezon City||Metro Manila||PCOS machine not working|
|5/10/2010 16:20:41||reklamotion||Clustered precinct 200||Molino elementary school||Bacoor, Cavite||broken PCOS machine|
|5/10/2010 16:21:46||jaimehernandez||Cluster 96||Brgy. San Francisco||Biñan, Laguna||Previously tested PCOS 3403096 now hangs|
|5/10/2010 16:34:02||pinoyweekly||[none specified]||Arellano HS, Sta. Cruz||Manila||broken PCOS machine|
|5/10/2010 16:36:19||bulatlat||[none specified]||Vicente Lim Elementary School||Tondo, Manila||Defective PCOS machine. problem started at 11:30, stopped operating at 12:42pm|
|5/10/2010 16:39:12||raikun07||[none specified]||Sinait||Ilocos||one machine not functioning, but all other machines work|
|5/10/2010 16:40:09||coderadiophils||Precinct 1175 cluster 215||Brgy. Salvacion, District 1||Quezon City||PCOS machine broke down|
|5/10/2010 16:41:25||workerswatch||[none specified]||Brgy. O’Donnell, Capas||Tarlac||3 PCOS machines fail|
|5/10/2010 16:45||bulatlat||V. Mapa High School, Sta. Mesa||Manila||PCOS machine shut down after 176th ballot. Manual voting ongoing.|
|5/10/2010 16:47:05||henriah||[none specified]||Masbate||Masbate||PCOS machine caught fire|
|5/10/2010 16:48:08||Arlyn Cabrera||Precinct#202||Sagkahan 62-A||Tacloban City||PCOS machine gave out, voters go home|
|5/10/2010 16:49:00||bulatlat||[many]||Oslob, Argao, Mandaue, Consolacion and other precincts||Cebu City||PCOS machine malfunction|
|5/10/2010 16:50:10||Jjaness||Precinct No. 05010A,Cluster 3||Baliwasan Central School||Zamboanga City||PCOS machine not working|
|5/10/2010 17:03:11||bicolanodevil||001a, 002a, 003a||Malinao Elementary School, Malinao||Albay||PCOS machines shut down|
|5/10/2010 17:11:27||Karen Ang||52A 4||Bagong Ilog Elem School||Pasig||BEI didn’t use UV lamp. When I asked why, she said they only use it when PCOS rejects a ballot.|
|5/10/2010 17:26:06||gmanews||[many]||Southern Leyte||Leyte||35 PCOS machines experience glitches|
|5/10/2010 17:27:06||pinoyweekly||[none specified]||Guevarra Elementary School||Sta. Cruz, Manila||3 PCOS machines did not function|
|5/10/2010 18:35:44||kontradaya||[not specified]||[not specified]||Taguig City||PCOS malfunctioned, voters sent home|
5. Parallel Manual Count
Why not have the teachers count the filled-up ballots at the close of voting on Election Day? Their counts can be compared with the count from the PCOS. If the counts are the same, or if different by not more than a COMELEC-set percentage, then the results can be transmitted onward. Otherwise, the COMELEC can decide what to do with this discrepancy. Would you not agree that the manual count will be the more credible figure in this case?
The experience of the random manual edit exercise reveals that this would be difficult to carry out.
Random Manual Audit (from the Report of the Technical Working Group (TWG-RMA) on the Random Manual Audit of the Automated Election System (AES) in the May 2010 National and Local Elections):
The audit process included the comparison of the number of votes counted by Precinct Count Optical Scan (PCOS) machine or the Automated Election System (AES) against the number of votes manually counted by the Random Manual Audit Team (RMAT) for every sample clustered precinct for the positions of President, Vice-President, Member of House of Representatives, Governor and Mayor. The difference between the AES count and RMA count, which is also referred to as the “variance” was expected to be computed for each candidate from a total of 1,145 sample clustered precincts or an equivalent to 1.5 percent of 76,347 total clustered precincts in the Philippines.
The Random Manual Audit in the 1,145 selected clustered precincts did not happen simultaneously. Since the ‘tambiolo’ system of randomly selecting the clustered polling precincts took close to twelve hours, the transmission of the results to the PES in all provinces nationwide was, as a consequence, also delayed. Many of the RMATs waiting for their assignments had gone home. Many local COMELEC offices no longer had their staff on duty. The Provincial Election Supervisors were busy with the canvassing of the votes. There were selected precincts whose ballot boxes had already been brought to the offices of the Municipal/City Treasurers. It was only in a few nearby places, or in precincts were voting ended early that the Random Manual Audit took place on Election Day itself.
The RMA Reports started coming in slowly. These were brought to the COMELEC Command Center in the Philippine International Convention Center (PICC) during the first days after the May 10, 2010 elections. Later on, after the COMELEC had proclaimed the twelve (12) winning senators, and before the National Canvassing of the votes for President and Vice President commenced in the Batasan Pambansa, the PICC Command Center was dismantled. The RMA Reports were then delivered and received at the COMELEC Project Management Office in Intramuros.
The reception of the Random Manual Audit Reports from the field followed this system:
1. The copies of the RMA Report intended for the COMELEC and the TWG‐RMA upon delivery by the Provincial Election Supervisors (PES) or their representatives to the COMELEC Central Office in Intramuros were received by the Election Records and Statistics Department (ERSD) of the Commission.
2. The Audit Returns (ARs) and the Minutes/Reports of legislative districts of the provinces were noted down by the secretariat of the Office of Commissioner Lucenito Tagle. These RMA reports were then assigned to “Verifiers” gathered from the offices of the Commissioners, ODEDO, EBAD, FSD and the IAO, and supervised by some staff from the Office of Commissioner Gregorio Larrazabal. PPCRV volunteers also assisted in the verification phase of RMA Reports.
3. Verification consisted of the following tasks:
° Checking the completeness of the RMA documents like the ARs and the Minutes/Reports. Note: When there are no ARs, but the data needed can be found in the Minutes/Reports, the Verifier considered such RMA Report as complete. When there are no Minutes/Reports, but the ARs are there, the Verifier is to prepare an improvised Minutes/Reports by using the ARs for the “Per RMA” and the “Per AES” from the data downloaded from the COMELEC website. This would render the RMA Report complete.
° Checking the correctness of the Taras, words and figures of the votes garnered as found in the ARs should tally.
° Checking the correctness of the transposition of the figures from the “Per AES” as found in the ERs and “Per RMA” as found in the ARs should be the same. Note: If any difference were to be detected, the Verifier can correct the variances that are due to clerical and/or mathematical errors. Thirty‐two (32) such variances were corrected.
° The Verifiers then submit their verified RMA Reports to an encoder from the Internal Audit Office (IAO) who will record the received and verified RMA Reports, after which these RMA Reports are passed on to the National Statistics Office (NSO) staff for the encoding of data, the computation of variances and the preparation of summary tables on these variances.
And the results of the random manual audit?
The conclusion of the same Report reads:
Root cause analysis of the variances in a majority of the clustered polling precincts manifested the difficulty of conducting parallel manual and machine counts. The manual count will always be subject to the discretion of the auditor in trying to interpret and/or appreciate voter’s intent. The machine will only count in the way it is programmed to.
If the Random Manual Audit is to make sure of the accuracy of election results and preserve electoral integrity, then it must be clearly pointed out from the beginning that the margin of variance is a computation of the difference between the manual count and the machine count. Hence, it is a test between man and machine. While the tolerance level set for the accuracy of the machines is a test that involves only the machines. The margin of variance indicated by the NSO as equivalent to 1% based on the accuracy rate of 99% (the allowable rate used in statistical analysis) cannot thus be used as the basis of comparison to the 99.995% accuracy rate of the PCOS machine set by the Request for Proposals (RFP) as a requirement for the Bidding Process of the AES technology.
As indicated in the NSO cumulative variances for the five positions subjected to the Random Manual Audit, none failed the accuracy rate of 99%.
Unfortunately, the Report does not contain (1) number of voters who actually voted, (2) number of valid ballots, (3) number with over-votes and (4) number with unusual markings, but which are available in the individual RMA minutes.
I understand that the denominator used to compute for the accuracy rate, as suggested by the Technical Evaluation Committee (TEC), is the number of valid ballots cast based on Automated Election System (AES) supplied by the COMELEC. The RMA process essentially checked the valid ballots count only. This assumes that the PCOS machines could not be held accountable for the rejected ballots; this was the fault of the voter. I would have preferred that the denominator include the rejected ballots as well.