THE Bangsamoro Autonomous Region in Muslim Mindanao (BARMM) elected its first regular parliament on Sept. 14. As it does after every automated election since 2010, the Commission on Elections (Comelec) is expected to announce in the coming days the result of a random manual audit. This is usually reported as an “accuracy rate” — a single number meant to show that the counting machines worked well. Expect the number to be above 99.9 percent. Expect most people to read that and simply feel reassured.
That reaction deserves a second look. A closer look at how this particular audit’s random draw was set up — the kind of review Namfrel regularly does as one of Comelec’s accredited citizens’ arms — shows that many citizens, and even some officials who quote the number, may not fully understand what it really measures.
Here is what the audit actually does. Under the election law, Comelec must pick a small number of precincts at random after the voting, then have a team open the ballot boxes and count the paper ballots by hand. That hand count is then compared with what the machine had reported. This is a spot check, not a full recount of every vote. For the BARMM election, there are 5,212 precincts in total. The plan is to check only 21 of them, picked by lottery. Before the election, Namfrel asked Comelec to raise this to 32 — one precinct for each of the region’s 32 new parliamentary districts. The audit committee accepted several other fixes Namfrel suggested, explained further below. But it turned down the request to check more precincts, saying it did not have the authority, at that point, to change a number Comelec had already approved.
This is the part every citizen should pay attention to. Whether the audit checks 21 precincts or 32, out of 5,212, the result is the same in one important way: The check could easily be off by more than 16 percentage points. To put that in perspective, an election survey with that kind of error margin would almost never be published without a strong warning attached. A check this rough cannot tell citizens whether something went wrong in one specific town or district. At best, it gives a very general, regionwide picture — useful but nowhere near as solid as the confident number that eventually gets announced.
Then there is the number itself: “99.995 percent.” This figure has been attached to every Philippine automated election since 2010, including this one in some form. It sounds exact, and in a way it is — it simply means allowing for one wrongly read mark out of every 20,000 checked, the same way “one typo in a 20,000-word essay” would count as 99.995 percent accurate. What nobody has explained in 16 years and six elections is why 20,000 was chosen at all. There is no public record of Comelec saying how confident it wanted to be, or how many mistakes an election could safely have, before settling on that particular number. It was simply written into a bidding document for the 2010 elections and has been repeated ever since — even in three of the last five elections, when the actual result came in below that mark, with no visible change to how the number is explained or used.
There is a second problem specific to BARMM, and it matters. At the turnout expected in this election, one precinct produces only about 2,900 marked ballots — far short of the 20,000 needed to fairly compute that “99.995 percent” figure. To reach 20,000, the audit would have to combine the ballots from about seven different precincts, using seven different machines, into one blended result. So, when “99.995 percent accurate” is eventually announced, it will describe the combined performance of those seven machines together — not the one machine that actually counted any single voter’s ballot.
None of this says whether the results of the BARMM election were correct or not, and it is not a criticism of the machines or the people who ran them. It simply points out how a well-known number is calculated, and what it can honestly be used to prove. Comelec’s own audit teams have said the same thing in the past: The random manual audit checks whether a machine can read ballot marks correctly. It was never meant to prove that an entire election’s results were accurate. Those are two separate questions, and treating them as the same does not serve the public well, even if it feels reassuring the moment the number comes out.
To its credit, the audit committee acted quickly on several fixes Namfrel pointed out before the election — publishing the random numbers used to pick the sample, correcting a computer script that had errors, and removing a shortcut that made part of the “random” selection not truly random. These are real improvements, and they should make the next audit more transparent. What is still missing — checking more precincts to match BARMM’s own 32 districts, and finally explaining, in public, the reasoning behind whatever accuracy number Comelec uses — can still be done before the 2028 national elections. No new law is needed. What is needed is the will to explain a number the public has been asked to trust for 16 years, instead of simply repeating it.
