Using Text/SMS for an Online Database of Election Returns


Roberto Verzola

Executive Director, Philippine Greens Institute

<[email protected]>

Abstract

A text/SMS-based system to ensure the integrity of election tallies through a database of precinct-level election returns maintained by a nationwide network of volunteers


Background

The results of the 2004 Philippine presidential elections remain under a cloud of doubt for many reasons: 1) during the canvassing of presidential returns in 2004, the administration party blocked opposition requests to compare the provincial totals with the municipal breakdowns; 2) a year later, recorded conversations were made public revealing that winning presidential candidate and incumbent Gloria Macapagal-Arroyo had talked many times to a commissioner of the Commission on Elections (COMELEC) about the election tally in several Mindanao provinces, where results turned out to be highly questionable. Their conversations ranged from references to military officers who refused to condone the cheating to election officials who might have to be kidnapped for refusing to cooperate; 3) the COMELEC commissioner subsequently fled the country, a virtual admission of guilt, though he resurfaced after a few months to face questioning; 4) the administration party resorted to technicalities to prevent a credible impeachment process; 5) the winning presidential candidate reneged on her promise to set up a truth commission to investigate the matter; 6) potential government witnesses like a Philippine Army general were ordered not to testify before investigative bodies and were harassed when they did testify; and 7) the administration continues its efforts to put an end to independent activities by citizens and civil society groups toward determining the truth behind the this election.

A major cause of this persistent doubt is the absence of a credible institution that can tabulate results honestly and truthfully and which the public trusts. For the reasons listed above, the public does not fully trust the canvass conducted by COMELEC and Congress.

For many years, a citizens' group, the National Movement for Free Elections (NAMFREL), has conducted a relatively credible parallel count of national elections. However, its credibility has also suffered serious damage recently (see Verzola, Roberto. “The True Results of the 2004 Philippine Presidential Elections Based on the NAMFREL Tally”. Kasarinlan, Vol. 19 No. 2. 2004). NAMFREL lost its credibility because: 1) it kept quiet about the discrepancies between its tally and the official results; 2) the pace of its count in different regions showed a bias in favor of the winning candidate; 3) it stubbornly refused to release to the public the breakdown of precincts it has counted, data that is essential to the appreciation of its results; and 4) the COMELEC commissioner suspected of rigging the elections was recorded on tape telling the winning candidate privately that NAMFREL had turned “sympathetic” to their side.

The basic election unit in the Philippines is the precinct, where voters go to cast their ballots. Each precinct has around 200 registered voters, and the Philippines has approximately 250,000 voting precincts. Ballots are counted manually by election officials in each precinct right after the voting closes, in full public view. By midnight or early morning after election day, the canvass is usually over. The tally sheets are filled up and signed, and the ballots and tally sheets are stored in metal ballot boxes. These are then padlocked and transported to municipal centers, where another manual canvass of precinct results takes place. The results of the municipal canvass are then tallied in provincial centers, and eventually at the COMELEC central office for the national results. The manipulation of results can happen at the precinct level, but big-time cheating takes place primarily at the municipal, provincial and national levels.

The project

The author has been involved in developing a system to help realize honest and truthful tabulations in future elections by enabling any citizen to determine whether the results in one's precinct have been properly tallied in the subsequent canvassing and tallying up to the national level. A citizen can do this by:

  1. Observing the manual precinct tally to confirm if it is honest and jotting down the precinct results which one has determined to be truthful; and

  2. Monitoring the national tabulation of precincts to ensure that the online results for one's precinct are identical to what one saw and jotted down and then adding one's confirmation to the online results (or questioning them if they are different, and sending in one's own report).

Details of the project

The manual, open counting of ballots at the precinct level is already relatively transparent. Cheating at this level can be minimized if citizens and volunteer watchers simply observe this tabulation process, jot down the numbers themselves, and report any anomaly.

The challenge is to ensure that these precinct results will retain their integrity as official canvassing proceeds from the precinct to the municipal, provincial and finally national levels. The project will ensure this by:

The biggest challenge in developing and implementing this approach is in fact the social, not the technical, aspect of recruiting more than 250,000 volunteers and keeping cheaters out of this army of volunteers. Recruitment is initiated by a group of convenors who are known to be honest, who then invite only those whom they personally know and trust and can be relied upon to send in truthful reports. These will in turn invite others, in a classic tree-like expansion of exponential growth. Invited volunteers submit their personal data by text/SMS to a national database of volunteers. They are trained to send simulated reports using text/SMS, after which they become full members. They can then invite and train more volunteers. The volunteers comprise what we call the Network of Citizens for Honest Elections and Truthful Statistics (NoCHEATS).

On election day, after the voting, the canvassing starts in each of the 250,000 precincts. The volunteers, at least one per precinct, observe the canvass and then text the results and other observations to build a national database of precinct level election returns. While the project relies mainly on its trained volunteers to maintain the database, any citizen with a cellphone can also confirm online precinct entries or even send in their own reports. This can be useful to cover precincts where we have no volunteers, or to flag cheats who have managed to infiltrate our network of volunteers. We can then screen questioned reports and compare these with official precinct returns.

What is different with this project

Most computerized election systems rely on automatic counting machines that process ballots at high speed to generate electronic totals. These are then automatically communicated to central databases for the national totals. People are excluded from the system. As a result, the system becomes much less transparent. There is no way to relate the final results to the numbers that ordinary voters actually saw at their voting centers.

The main feature that distinguishes this system from other computer-based systems is that this system relies on an army of people rather than an army of counting machines. It makes the entire tally more transparent, not less, by making available to any citizen not only the outputs (totals) of a computerized tally but also the inputs (breakdowns). This is what the Philippine Congress refused to do in its 2004 canvass when it refused to open municipal breakdowns for comparison with provincial totals, and this is what NAMFREL refused to do in its 2004 Operation Quick Count when it released only the national total of precincts tallied and refused to breakdown this total by region and province.

The system starts with citizens who saw the individual ballots (input) being counted, and who confirmed that the precinct totals (output) are accurate. It relies on such citizens in every precinct to make sure that the citizens' copy is identical to the online precinct results (input). If the public can do this for every precinct in the country, and they can also download the entire database for independent analysis, to reconstruct for themselves the municipal, provincial and national totals (output); then they can truly ensure for themselves that the election tally has been honest and truthful. The key is not to replace people with machines, but to use technology to make the process more, not less, transparent to people.

Implementation details

The system can be implemented with standard hardware and software, and with some programming that is relatively standard. The current crop of advanced desktops should be able to handle the processing load easily. One desktop running free software (Linux/GNU, MySQL, and glue code in PHP, Perl or Python) will handle the bulk of database operations. Another machine will handle the incoming reports and a third will handle the response to incoming queries by text/SMS and over the Web. There are some but minor challenges in structuring the databases to optimize for speed and in designing the user interface for friendly text/SMS interaction.

Because text/SMS messages may not exceed 160 characters, between five and ten messages are needed for one complete precinct report. This report consists of the precinct number, the sender's full name and zip code, the total votes cast per position contested, and then for each position contested, the candidates names and votes received.

Incoming text messages are saved unmodified to a text file. A parallel text file of information about the reporting mobile phone, extracted from the volunteers’ database, is generated. These two files comprise the system's audit trail, and they are saved regularly to data CDs and made available to the public. For security reasons, the volunteers’ database is not available to the public. 

As the text messages arrive, they are parsed to build a database of field reports, one record per complete precinct return sent in from a particular phone. A report about the same precinct sent in from another phone is stored on a separate record. A flag marks a complete record (i.e., all messages comprising a complete precinct report that has been received). Complete records are reviewed for wrong or inconsistent entries and texted back to the sender, who may send changes, or a final confirmation. Once a report is confirmed, it becomes read-only. The database of field reports is optimized for replying to queries.

At regular time intervals, the database of field reports is scanned for confirmed records. If conflicting reports exist for a precinct, their credibility index is computed (reports from volunteers carry a greater weight than reports from the public; reports with more confirmations are more credible; etc.). The report with the highest index will be used.

The database of precinct election returns is then updated. This database is structured optimally for repeated computations of municipal, provincial and national summaries of election results. These summaries are stored on separate files, and may be downloaded by the public anytime.

Because precinct details will remain disaggregated, those who personally witnessed the canvassing in their precinct may, at any time, confirm that their local precinct returns and the online precinct returns are identical. This is the key to the credibility and usefulness of the system.

Future possibilities

The potential of the network for other types of quick, nationwide information gathering is enormous. Depending on the inclinations of the volunteers, the network can, in between elections, direct its attention to corruption in government, opinion polls, survey taking, etc. with exciting possibilities for democratization as well as fund-generation for the network itself. It must be ensured, however, that none of these other activities will undermine the main mission of the network to keep elections honest and statistics truthful.

A system which empowers ordinary citizens to ensure the honesty and credibility of national and local election tallies is a worthy legacy to leave our children and grandchildren. It may even set a good example for the rest of the world.