60-30-10 voting pattern: Conspiracy or just math?

Earlier in the week, Ateneo math professor Lex Muga noted what appeared to be a consistent voting pattern across the country: throughout most of the voting period, Team PNoy candidates were consistently getting 60% of the vote; rival UNA candidates, 30%; and independent candidates, 10% – the now-infamous "60-30-10" voting pattern.

Muga's assertion of an "interesting pattern" has flown across Facebook and fueled suspicions of election manipulation, leading the Comelec chairman to announce an investigation.

 

At first glance, the numbers do seem counterintuitive at best and a smoking gun at worst: How can a supposed democracy, comprised of people with varied political inclinations and candidate choices, yield such uniform results across the board?

The question has triggered a prickly intercontinental debate among academics, with New York University science dean Michael Purugganan dismissing the pattern as simply "what we expect" when adding up large sample sizes, and warning other academics against creating conspiracy theories that could "grow into urban legends."

"I frankly am flabbergasted because the 'problem' they describe is completely explainable," says Purugganan.

 

Law of Large Numbers

Purugganan is not the first person with a PhD to challenge the conspiracy theorists.

 

Shortly after Muga posted his findings, Center for People Empowerment in Governance (CenPEG) IT consultant Dr. Pablo Manalastas posted a counter-argument on his Facebook page, saying that the pattern may simply be due to the Law of Large Numbers.

 

Basically, this means that a large number of votes counted over time will tend to average out and follow an expected pattern.

 

"Naturally, the bigger the number of votes canvassed, the closer you get to the true will of the people," Manalastas said.

 

Dr. Purugganan, the Dean for Science at New York University, shares a similar view.

 

"If (Comelec) was reporting results that came in from all over the country, and each canvass was from a large chunk of voters, then each canvass would be very similar to each other and close to the national result," he said.

 

Moreover, Purugganan said that the pattern can be explained by statistical analysis and is likely not the result of fraud.

 

"I do know that (it has been observed) that the Team PNoy average is stable, but the order of senators can be different in different regions.  That is hard to try to explain as a conspiracy," he pointed out.

 

He also said that the data which Muga analyzed doesn't quite fall neatly into the 60-30-10 pattern.

 

"(For example), the range from canvass to canvass is 54% to 64% (for Team PNoy). That is a huge 10% spread!" Purugganan said.

 

The small details

 

The real key to determining whether or not fraud has been perpetrated, according to Manalastas and Purugganan, is to look at the numbers up close.

 

"The more correct indicator of a conspiracy is if we get the same 60-30-10 figures in a precinct-by-precinct comparison, provided that the precinct figures were used to get the national canvass," Manalastas said on Facebook.

 

But in a follow-up interview, Muga said that the same 60-30-10 pattern is apparent even down to samples of as small as tens of thousands of voters.

 

"We are now computing at areas involving less than one million senatorial votes. For example, (we're looking at) Basco, Batanes, with about 61,000 senatorial votes and we still get the same pattern," he told GMA News.

 

"But that's still such a very large number!" Purugganan countered. "If you were to go into the low thousands (and the pattern is still there), then that would certainly raise my eyebrow."

 

Sample size matters

 

The lynchpin of the argument over the 60-30-10 pattern is the sample size being analyzed: to find out whether or not the bigger picture is an accurate reflection of the small details, analysts need to look up close – but how close?

 

"How many votes do you need to count before you get very little change from sample to sample? It doesn't take many votes to really get a good estimate of the national average, as long as each group of votes that are counted are really representative of the country as a a whole," Purugganan explained.

 

He said that statisticians would need to run a "statistical power test" to determine whether Muga's observations hold water.

 

A simple explanation from the Indiana University website says that "statistical power analysis estimates the power of the test to detect a meaningful effect, given sample size, test size (significance level), and standardized effect size." 

 

Back to the ballots

 

For his part, Muga said that the only way to cast aside all doubts concerning the election results would be to gain full access to Comelec data.

 

"Kailangan natin ang lahat ng data (from Comelec) para masagot natin with finality  kung (Law of Large Numbers nga) lang ba ito," he said.

 

GMA News asked Muga if he has seen any data so far that doesn't conform to the 60-30-10 pattern.

 

"Sa ngayon, wala pa. (Pwede mo) tanungin sa iba kung may nakita sila. Pero kailangan talaga namin ang mga data ng Comelec," he concluded.

Other patterns

Purugganan says that there is much that we can learn from Comelec's election data.