Prometheus bound ratios, transparency and lemons



The recent buzz on the supposed 60:30:10 party ratio found in the midterm senatorial elections has generated various reactions from different sectors and spawned data analysis by IT experts and other academics. The said pattern was initially observed by former Comelec IT officer Ernie del Rosario and was demonstrated by Ateneo Prof. Lex Muga in a Facebook post.

An explanation for the observed pattern was later given in terms of the law of large numbers. Since the Comelec reports posted in the transparency server aggregates the data at a national, and therefore large, numbers, the said explanation is an acceptable one. One should also appreciate that when we talk of percentages in this election, one percent of the total number of registered voters is already a number that is just a bit larger than half a million.

When drilling down up to the precinct levels, at least for those precincts that we have data for, the ratio seems to vary widely. Differences in voter preferences seem to already appear at this level. For one to check the consistency (or not) of the said ratio, we really need to have data for all electoral precincts. Some sectors are proposing this route in order to eliminate doubts on the elections insofar as the ratio is concerned.

Yet there is a caveat in this proposal: it assumes that the initial Comelec report, and the succeeding transmissions to the transparency server, are correct. It also would imply that the initial report has already given a fair sample of the national voter preference in the initial burst of transmissions (Footnote: this random sampling might be even a side result of the [random]transmission failures nationwide).

Garbage in, garbage out
Any data analysis, especially of large data sets such as electoral returns and voter preferences, should start with the appreciation of the accuracy and veracity of the information being processed.

What is not being stressed in these various data visualizations and mathematial calisthenics in the internet is the fact that the data we are crunching has not been demonstrated by the Comelec to be accurate nor reliable.

The initial impressions and patterns were based on the data originating from the transparency server. This transparency server has itself shown problems just a few hours after the polls closed when the PPCRV were reporting improbable numbers for the national counts. The supposed double count and format mismatch was corrected onsite by a script rewrite by a Smartmatic technician.

The transparency server as being implemented by the Comelec is really not tranparent at all. We assume that it is a correct copy of the results in the Comelec ladderized tally but we have no assurance for that. With the digital signatures disabled, the canvassing and transparency servers receive the data from the PCOS machine without verification, authentication nor acknowledgement. Therefore, the data can be incomplete, and even different, from what is being canvassed in the official ladderized system. In 2010, out of the 60,000 or so election returns that have transmitted its data, around 13,000 were found to be incomplete.

What the transparency server does is only to set trends or patterns which may or may not be there because we simply have no way to verify despite our best minds at working on it. The only way would be to check the ballots themselves.

Why then do we have to use the same lemon of a system for the 2016 elections? With a clear thirty percent failure rate in terms of functionality, the Smartmatic-Comelec Automated Election System used in 2010 and in 2013 remains opaque, slow and untrustworthy.

Chairman Brillantes had to perform procedural acrobatics to save face and make it appear that the elections were proceeding as he had announced. He was bitten by his misplaced faith on the machines and Smartmatic. He had to make ad-hoc adjustments and improvisations in order for the counting to proceed and the winning candidates to be announced. We were expected to trust the AES system as a black box when there were obvious problems with the whole system itself—compounded by the lack of trust with the Comelec itself.

The core problem is still the fact that the Comelec surrendered its core function to a foreign and privatized company. It has now become an apologist to an untrustworthy system that is not only a lemon but is as opaque as the blinders it puts on each commissioners’ eye.


Comments are closed.