Automated Elections For Better Or For Worse By Roberto Verzola

8/12/2009

Automated Elections: For Better or for Worse?

by Roberto Verzola Secretary-general, Halalang Marangal 0929-856-1930

The official canvassing process (manual) Precinct counting (ER)

~260,000

Precinct returns can be found in Election Returns and the City/muni Statement of Votes

City/muni Statement of Votes (MSOV) 1,628 City/muni canvass (MCOC)

Provincial Statement of Votes (MSOV) 81 Provincial canvass (PCOC)

Precinct

City/town

Provincial

National canvass National

1

8/12/2009

The official canvassing process (manual) Precinct counting (ER)

~260,000

Precinct returns can be found in Election Returns and the City/muni Statement of Votes

City/muni Statement of Votes (MSOV) 1,628 City/muni canvass (MCOC)

Provincial Statement of Votes (MSOV) 81 Provincial canvass (PCOC)

Namfrel Citizens' Count Precinct

City/town

Provincial

National canvass National

2004 presidential elections: How GMA 'won' in ARMM

2

8/12/2009

2004 discrepancies between Namfrel and Comelec counts Expressed as percentage (%) of total votes, for valid comparisons Discrepancy between identical results: 0% Highest possible discrepancy: 200% The worse the cheating, the higher the discrepancy

Th e w orst d iscre p a n cie s, b y re g ion Are a ARM M C. M ind a na o CAR N . M ind a na o W . M ind a na o CARAGA S. M ind a na o S.Tagalog Bicol NCR C.Valley C.Visayas E.Visayas W.Visayas Ilocos

N AM FREL le a d

Cong re ss le a d

D iscre p a ncy

FPJ: 1 9 . 2 % FPJ: 2 0 . 5 % GMA: 9.3% FPJ: 3 . 0 % GM A: 4 . 9 % GM A: 2 2 . 7 % FPJ: 1 . 2 % FPJ: 21.2% GMA: 4.1% FPJ: 10.0% FPJ: 11.9% GMA: 56.7% GMA: 2.7% GMA: 33.9% FPJ: 6.7%

GM A: 3 1 . 3 % FP J: 7 . 0 % GMA: 13.4% GM A: 0 . 8 % GM A: 8 . 6 % GM A: 2 4 . 0 % FP J: 0 . 6 % FPJ: 20.6% GMA: 4.0% FPJ: 10.2% FPJ: 12.4% GMA: 56.2% GMA: 2.1% GMA: 33.0% FPJ: 8.0%

50 .5 % 13 .5 % 4.1% 3 .9 % 3 .8 % 1 .3 % 0 .6 % 0.5% -0.1% -0.2% -0.5% -0.5% -0.6% -0.9% -1.3%

3

8/12/2009

Can the discrepancy be due to the incomplete NAMFREL tally? We looked at each NAMFREL provincial tally grouped by completion rate (<50%, 50-59%, 60-69%, 70-79%, 80-89%, and 100%). We found that most province in each group matched the Congress tally closely. But one or two provinces showed a big discrepancy compared to the Congress tally. No, it is not due to the incomplete tally. These discrepancies are probably due to cheating.

Discre pa ncie s in GM A m a rgin in province s w he re t he NAM FREL t a lly w a s < 5 0 % com ple t e P rovince S. Kud a ra t La na o d e l S. Sarangani Quirino Prov. Davao Or. N. Cot abat o Caloocan

N AM FREL Le a d FP J: 2 9 . 1 % FP J: 1 1 . 1 % FPJ: 22.6% FPJ: 7.2% GMA: 7.8% FPJ: 10.6% FPJ: 17.9%

Cong re ss Le a d GM A: 3 6 . 3 % GM A: 4 6 . 8 % FPJ: 13.9% FPJ: 5.6% GMA: 8.8% FPJ: 9.9% FPJ: 17.5%

D iscre p a ncy 6 5 .4% 5 8 .0% 8.7% 1.6% 1.0% 0.6% 0.4%

4

8/12/2009

Discre pa ncie s in GM A m a rgin in province s w he re t he NAM FREL t a lly w a s 5 0 -5 9 % com ple t e Province M a g uind a na o Agusan del N. Valenzuela Z m b del N.

N AM FREL GM A: 2 9 . 7 % GMA: 26.9% FPJ: 18.1% GMA: 31.0%

Cong re ss GM A: 4 8 .3 % GMA: 29.8% FPJ: 17.3% GMA: 28.7%

D iscre p a ncy 18.7% 2.9% 0.8% -2.2%

D iscre p a ncie s in GM A m a rg in in p rovince s w he re t he N AM FREL t a lly w a s 6 0 -6 9 % com p le t e Province

N AM FREL

Cong re ss

D iscre p a ncy

Ba sila n Benguet Kalinga Bat aan Manila Tarlac Quezon Cit y S. Cot abat o

FPJ: 5 2 . 3 % GMA: 20.1% GMA: 16.6% FPJ: 36.6% FPJ: 11.6% GMA: 10.8% FPJ: 2.7% FPJ: 19.4%

GM A: 2 2 . 8 % GMA: 23.4% GMA: 18.7% FPJ: 35.6% FPJ: 11.8% GMA: 9.5% FPJ: 4.4% FPJ: 22.5%

7 5 .1 % 3.4% 2.1% 1.0% -0.22% -1.3% -1.7% -3.1%

5

8/12/2009

Discr e p an cie s in GMA m ar g in in p r o vin ce s w h e r e th e NAMFREL tally w as 7 0 -7 9 % co m ple te P rovince La na o d e l N . Z am bales Surigao del S. Pasig Cit y Mandaluyong Cavit e Agusan del S. Negros Occ. Isabela

N AM FREL Ta lly FP J: 1 8 . 4 % FPJ: 29.6% GMA: 18.5% FPJ: 13.8% FPJ: 6.8% FPJ: 6.3% GMA: 20.2% GMA: 23.2% FPJ: 18.4%

Cong re ss Ta lly FPJ: 2 . 3 % FPJ: 25.3% GMA: 21.4% FPJ: 12.4% FPJ: 6.6% FPJ: 6.3% GMA: 20.1% GMA: 23.1% FPJ: 18.8%

D iscre p a ncy 16 .2% 4.3% 2.9% 1.4% 0.15% 0.04% -0.06% -0.13% -0.40%

Discr e p an cie s in GM A m ar gin in pr o vin ce s w h e r e th e NAMFREL tally w as 8 0 -8 9 % co m ple t e P rovince Ta w i-Ta w i Rom blon Iloilo Cagayan Nueva Ecija Malabon/Nav. N. Sam ar E. Sam ar Bukidnon Mindoro Or. Pasay Cit y Las Piñas Makat i Cit y

N AM FREL Ta lly FP J: 5 4 . 3 % FPJ: 10.4% GMA: 56.0% FPJ: 7.1% FPJ: 40.4% FPJ: 29.6% FPJ: 32.9% GMA: 1.0% GMA: 10.2% FPJ: 16.2% FPJ: 14.1% GMA: 1.0% FPJ: 6.4%

Cong re ss Ta lly FP J: 1 8 . 5 % FPJ: 4.0% GMA: 57.3% FPJ: 6.1% FPJ: 40.9% FPJ: 30.2% FPJ: 33.4% GMA: 0.5% GMA: 9.8% FPJ: 16.5% FPJ: 14.5% GMA: 1.4% FPJ: 5.8%

D iscre p a ncy 35.9% 6.4% 1.2% 1.0% 0.5% 0.5% 0.5% 0.5% 0.5% 0.4% 0.36% -0.42% -0.59%

6

8/12/2009

Discr e p an cie s in GMA m ar g in in p r o vin ce s w h e r e th e NAMFREL tally w as 1 0 0 % co m p le te P rovince Sulu Guim aras Biliran Aurora Bat anes N. Vizcaya

N AM FREL FP J: 2 9 . 3 % GMA: 60.5% FPJ: 0.1% FPJ: 34.9% GMA: 27.5% FPJ: 10.7%

Cong re ss GM A: 1 2 . 0 % GMA: 63.5% GMA: 1.2% FPJ: 34.9% GMA: 27.4% FPJ: 10.9%

D iscre p a ncy 41 .3% 3.0% 1.3% 0.02% -0.17% -0.22%

Discre pa ncie s in a ve ra ge vot e s pe r pre cinct , by re gion Are a ARMM Ilocos C. V isa ya s E. V isa ya s C.Mindanao Bicol CARAGA S.Tagalog NCR N.Mindanao W.Visayas S.Mindanao CAR C.Luzon W.Mindanao

N a m f re l AV P 99.8 147.0 136.7 138.2 132.9 147.3 146.0 149.8 134.2 143.4 153.4 144.4 148.5 154.5 135.5

Cong re ss AV P 153.4 167.7 151.0 146.6 140.7 152.1 148.1 150.6 134.8 143.2 152.6 143.4 147.2 152.8 133.3

Incre a se p e r p ct 53.6 20.7 14.3 8.3 7.8 4.8 2.1 0.8 0.6 -0.1 -0.8 -1.0 -1.3 -1.7 -2.2

Tot a l Incre a se in p re cinct s vot e s 5,736 307,484 11,402 235,563 16,567 236,985 10,553 88,039 9,077 70,930 12,295 58,822 5,931 12,562 33,080 26,789 29,370 18,007 10,419 -1,425 17,562 -13,333 10,463 -10,835 3,861 -4,948 24,525 -40,709 8,361 -18,680

7

8/12/2009

2004 presidential elections: How GMA 'won' the NAMFREL count

Num be r of pre cinct s not t a llie d by NAM FREL Re g ions w he re GM A w on (lead) C. V isa ya s (1.3M) W . V isa ya s (757T) CARAGA (144T) Bicol (71T) CAR (42T) E. V isa ya s (42T) W . M ind a (31T)

Precinct s not t allied in GMA areas

P ct s not t a llie d 0* 2,757 1,590 599* 814 396 1,417

7,573

Re g ions w he re Ra nk FP J w on (lead) S. Ta g a log (924T) 1 2 N CR (290T) 3 C. M ind a (129T) 4 C. V a lle y (112T) 5 Ilocos (98T) 6 C. Luz on (70T) 7 ARM M (67T) 8 N . M ind a (39T) 9 S. M ind a (16T) Precinct s not t allied in FPJ areas

P ct s not t a llie d 3,048 7,892 4,334 1,329 1,887 4,648 1,727 1,336 1,456 27,657

8

8/12/2009

2007 senatorial elections: how Maguinanao made Zubiri senator

M ag uinda nao 2 0 0 7 Vot er t urnout Vot ers who voted Aquino (GO) Cayetano (GO) Coseteng (GO) Kiram (TU) Lacson (GO) Osmena (GO) Pimentel (GO) Roco (GO) Trillanes (GO) Zubiri (TU) Tot al vot es Tot al vot ers x 12 % of Theor. Max. Z ubiri vot es Z ubiri m argin over Pim ent el

l ga sa ng ng n A n pi a ga an s tu b ua t u lt o n g i lu n hU at a D a d t im S t l a p u r g g Mi Pa Pa So Ba Am 98.84% 96.71% 97.13% 70.55% 77.09% 98.82% 5,212 4,528 10,308 2,392 9,461 8,530 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1,740 1,356 5,977 1,587 1,998 2,025 0 0 0 0 0 0 0 0 0 0 4,732 768 1,240 2,196 8,504 1,671 0 0 0 0 0 0 0 0 0 0 4,991 4,401 10,205 2,316 9,305 8,412 72,342 54,380 123,928 29,802 115,139 113,553 62,544 54,336 123,696 28,704 113,532 102,360 115.7% 100.1% 100.2% 103.8% 101.4% 110.9% 95.8% 97.2% 99.0% 96.8% 98.4% 98.6%

259

3,633

8,965

120

801

6,741

l ta To 93.48% 198,912 0 0 0 86,122 0 0 67,111 0 2,147 195,823 2,260,848 2,386,944 94.7% 98.4% 128,712 20,519

9

8/12/2009

Will electronic voting machines make elections more honest?

Background 6-week research fellowship awarded by University of Oxford's Internet Institute to HALAL secretary-general Roberto Verzola Research conducted Apr 20-May 30, 2008 Research outputs submitted to the COMELEC June 2008

10

8/12/2009

Research Outputs Costs of automated elections Automated elections: electronic voting machines have made mistakes too Double-entry accounting can improve both automated and manual tabulations Catching machine problems through postelection audits of sample precincts

Costs of automated elections $3,000 - $7,000 per machine Unexpected costs −

additional hardware; memory; batteries

−

software upgrades; configuration

−

training; public education

−

maintenance, storage, insurance

DRE (touch screen) much more expensive than OMR (optical scanners) Abalos OMRs: $15,000 (2003 prices)

11

8/12/2009

DRE (touch screen) problems insoluble No software-independent permanent paper record of voter intent Voters cannot check the electronic record Trivial programming to put one thing on the screen, another thing on the printer, and still another on a memory card/stick DREs have been phased out in several states of the U.S. DREs should be considered OBSOLETE!

U.S.: voting machine troubles Search: “electronic voting machines” 2000, 2004 “stolen” U.S. Presidency 42,000 “incidents” in 2004-05 (U.S.) −

Uninitialized voting machines

−

Votes not counted or reversed

−

Wrong winner comes out

−

Multiple voting

−

More votes than voters; negative votes

−

Unauthorized software replacement

12

8/12/2009

U.S.: types of election problems Election problem reported

Count

Registration-related Polling place inquiry Absentee ballot related Others Machine problem Other polling place problem Voter intimidation Provisional ballot Identification related Long lines Other ballot-related Disability access Criminal status related Late opening Insufficient number of ballots Early closing Language assistance Student status Unable to read ballot

15,404 36.0% 7,404 17.3% 4,122 9.6% 3,896 9.1% 2,293 5.4% 2,248 5.2% 1,763 4.1% 1,181 2.8% 1,075 2.5% 1,072 2.5% 794 1.9% 539 1.3% 354 0.8% 208 0.5% 129 0.3% 121 0.3% 111 0.3% 74 0.2% 51 0.1%

Total

42,839

% of total

100.0%

France: EVMs had greater discrepancies than paper ballots By Ryan Paul | Published: July 09, 2008 - 08:32PM CT A study conducted by a researcher in France has uncovered that polling locations which use electronic voting machines exhibit a higher number of discrepancies than those using conventional paper ballots. Unsurprising to those who have followed the problems plaguing e-voting since its introduction, the revelation has fueled renewed calls for greater scrutiny of electronic voting technology in France. The study was conducted at over 21,000 polling stations by comparing electoral registers, which voters sign after voting, with the total vote counts from machines and paper ballots in several elections. Discrepancies were found at almost 30 percent of polling stations that use electronic machines and only at about 5 percent of those using paper ballots. Based on the results, the researcher believes that broader studies are needed to determine the scope of the pattern and the reasons for the discrepancies. The root cause is thought to be technical rather than a result of widespread operator error because the margin of discrepancies increased in later elections when voters were already familiar with the systems.

13

8/12/2009

Ireland: Electronic voting system to be scrapped 23 April 2009--The electronic voting system is to be scrapped, having cost at least €51m to date. Minister Gormley said the figure of €51m does not include the cost of storing the machines. Environment Minister John Gormley said a task force is being set up to oversee the disposal of the voting and counting equipment and the end of storage arrangements. He said it was clear from the Report of the Commission on Electronic Voting that 'significant additional costs' would arise if electronic voting was proceeded with. Mr Gormley said the public appeared to be satisfied with the present paper-based system. It was of paramount importance to ensure public confidence in the democratic system, he added.

Finland: Gov't Rejects Defective E-Voting Results June 9, 2009--Back in February, we found it disturbing that Finland was allowing the results of an election to stand, despite the fact that at least 2% of the votes had gone missing due to e-voting glitches. However, it looks like some sense of sanity has been restored as a higher court has now rejected the election results and ordered a new election. One hopes that the new election won't involve similarly screwed up e-voting machines. In a separate article, we find yet another story of e-voting machines that were "mis-calibrated" in such a way that made it difficult to impossible for people to vote for candidates of their choice.

14

8/12/2009

Germany: e-voting unconstitutional Newsweek, May 23, 2009--After almost two years of deliberations, Germany's Supreme Court ruled in March that e-voting was unconstitutional because the average citizen could not be expected to understand the exact steps involved in the recording and tallying of votes. Political scientist Joachim Wiesner and his son Ulrich, a physicist, filed the initial lawsuit and have been instrumental in raising public awareness of the insecurity of electronic voting. The younger Wiesner said, with some justification, that the voting machines used in Germany are even less secure than mobile phones.

Netherlands: hacked in 5 mins. Newsweek, May 23, 2009--The Dutch publicinterest group Wij Vertrouwen Stemcomputers Niet (We Do Not Trust Voting Machines) produced a video showing how quickly the Nedap machines could be hacked without voters or election officials being aware (the answer: five minutes). After the clip was broadcast on national television in October 2006, the Netherlands banned all electronic voting machines.

15

8/12/2009

Scotland: serious technical failures May 4, 2007--A major and urgent investigation was launched today into the electronic voting failures which disrupted the Scottish Parliament elections. 10 Downing Street said the problems raised serious question marks about the future development of the technology not just North of the Border but across the UK. The inquiry was announced by the Scotland Office who said "serious technical failures" had delayed the announcement of results in several areas. More than 100,000 votes across the country may have to be discounted and problems with the new electronic counting system, being used for the first time in Scotland, also meant several counts were suspended until much later in the day. A spokesman said: "We share the public's concern about the high number of rejected ballot papers."

Causes inherent to complex technologies Software bugs Hardware problems (e.g., alignment, calibration issues) Environmental stresses Poor or flawed design Human error (election officials, technicians, operators, voters) Malicious tampering (esp. “inside jobs”)

16

8/12/2009

R.P. choice: Smartmatic Venezuelan-owned: 82.5% by Antonio Mujica, Roger Pinate, and Jorge Massa Separate company in the Netherlands, but registered in Barbados Election experience: Venezuela, Curacao, ARMM (but not the SAES 1800) Single largest contract: Venezuela elections Owner died 2008 in a mysterious plane crash

Plane crash: both engines fail May 3, 2008--A small plane, carrying the co-founder of Venezuelan voting machine company Smartmatic, crashed earlier this week shortly after take-off from the Caracas airport, killing two employees of the company, and several others on board and on the ground. The initial reports indicated that the cause of the crash may have been the unusual failure of both engines on the small plane. Smartmatic has been named as a subject in several recent, exclusive, investigative reports here at The BRAD BLOG, surrounding questions of the Venezuelan firm's apparent continuing control over the American evoting company, Sequoia Voting Systems.

17

8/12/2009

18

8/12/2009

The machine: SAES 1800 SAES 1800 for the 2010 is a new machine (Google “SAES 1800” Smartmatic: 15 hits) 13 hits are recent Philippine stories; the other 2 hits are both company brochures by Smartmatic Machine licensed from Dominion Voting Company of Canada, to be manufactured by Kenmec of Taiwan Joint venture company to be formed with Total Information Mgt Corp. (TIM) upon contract approval.

19

8/12/2009

Our ARMM experience Most candidates unopposed Smartmatic used a different machine, different technology (voting pad) Problems swept under the rug One person entering votes for many voters − Transmission problems − Security issues (results uploaded into a machine) Election law says: test first in Luzon, Visayas and Mindanao −

Controlling election cheating Cleansing the voters' list Two-column balanced accounting for election tallies Transparency in count, canvass Post-election statistical audits to ensure the integrity of results Punish the cheats!

20

8/12/2009

Two-column balanced accounting for election tallies Universally-recognized superiority over single-column systems Minimizes clerical errors, delays Facilitates detection of fraud Local pool of expertise (accountants and bookkeepers) already exist Low-cost, no special hardware

Recording a P500 expense: two methods Single entry:

Double entry:

Item A

100

Cash

Item B

150

Item A

100

Item C

250

Item B

150

--------

Item C

250

Total

500

500

------ -----Total

500

500

21

8/12/2009

Sample ballot report Ta b le 1 . Ele ct ion Re t urn, Ba llot St a t us Cr Ballot Type Dr Received 200 Excess 0 Unused 45 Cast 150 Missing 5

Sample vote report Ta b le 2 . Ele ct ion Re t urn, P re sid e nt No. of slot s: 1 Ballot s cast : 150 Vot es Dr Cr Cr in words Available 150 Invalid/Blank 5 five Candidat e 1 80 eight y Candidat e 2 50 fift y Candidat e 3 15 fift een

22

8/12/2009

Sample multi-slot vote report Ta b le 3 . Ele ct ion Re t urn, Se na t or No. of slot s: 12 Ballot s cast : 150 Vot es Dr Cr Cr in words Available 1800 Invalid/Blank 225 t wo hund. t went y-five Candidat e 1 140 one hund. fort y Candidat e 2 135 one hund. t hirt y-five Candidat e 3 130 one hund. t hirt y Candidat e 4 125 one hund. t went y-five Candidat e 5 120 one hund. t went y Candidat e 6 115 one hund. fift een Candidat e 7 110 one hund. t en Candidat e 8 105 one hund. five Candidat e 9 100 one hund. Candidat e 10 95 ninet y-five Candidat e 11 90 ninet y Candidat e 12 85 eight y-five Candidat e 13 80 eight y

Post-election audits using statistical sampling Persistence of machine troubles requires independent audits Science and mathematics of election audits emerging Random sampling of precincts can confirm at a 95% or higher confidence level the integrity of machine results Has been tried in the U.S., now recommended in California

23

8/12/2009

Remember: Moral problems cannot be solved by technological fixes A magician's trick can happen “faster than the eye can see” A basic principle of democracy is to “vote in secret, count in public” Law of “unintended consequences”

Halalang Marangal convenors Wigberto Tañada, former senator Mehol Sadain, former Comelec Comm. Gen. Francisco Gudani, retired Sr. Mary John Mananzan, former St. Scholastica's College president Isagani Serrano, PRRM senior VP Ma. Paz Luna, TOYM awardee Roberto Verzola, HALAL sec.-general

24