(2ND UPDATE) – Pulse Asia released its final survey before the 2022 elections, showing former Senator Ferdinand “Bongbong” Marcos Jr. still leading by a mile against Vice President Leni Robredo. Some Marcos supporters have already taken glee with the poll, reverberating their chant “May nanalo na.”
The 23-year-old research firm has correctly predicted the narrow victory of Gloria Macapagal-Arroyo in 2004 and the landslide victories of Benigno “Noynoy” Aquino and Rodrigo Duterte in 2010 and 2016.
Mrs. Arroyo led by six points against Fernando Poe Jr. in the final survey conducted between Apr. 26 to 29, 2004. Mr. Aquino was 19 points ahead of Joseph Estrada and Manuel Villar, who both notched 20 percent in the Apr. 23 to 25, 2010, poll. And Mr. Duterte solidified his frontrunner status in the Apr. 26 to 29, 2016, survey, getting 33 percent over Manuel “Mar” Roxas’s 22 percent.
Mr. Marcos Jr.’s survey rating has been unprecedented, becoming the first candidate to obtain majority voter preference in Pulse Asia’s history. In its February survey, he got a whopping 60 percent over Mrs. Robredo’s 15 percent.
‘Flaws’ on Pulse Asia survey
But Dr. Romulo Virola, former lecturer at the University of the Philippines and one of the Ten Oustanding Researchers conferred by the National Research Council of the Philippines, wrote in a Facebook post on Monday, May 2, that Mrs. Robredo could have overtaken Mr. Marcos Jr. had Pulse Asia rectified its flawed polling methodology.
In three sample models reweighting the results based on updated population figures and voter statistics, he argued that Pulse Asia may have overrepresented older voters – where Mrs. Robredo is polling higher – and underrepresented the younger ones – where Mr. Marcos is preferred by a near supermajority.
“[T]he biggest source of possible bias of the PA (Pulse Asia) survey in favor of Marcos is the underrepresentation of the young voters in the PA sample of respondents. And if this is corrected under certain assumptions, this alone will be sufficient to turn the tables around in favor of Leni,” he said. “[It] reminds us of the 2016 race for vice president.”
Assuming that 55% of respondents aged 18 to 41 years old preferred Mrs. Robredo to be the next president, she would overtake the lead over Mr. Marcos Jr. by just a percentage point, 40 to 39 percent.
This was also his assessment of the February survey, thinking that young voters born after the 1986 People Power Revolution – the landmark event that ousted Marcos Jr.’s father, Ferdinand Marcos Sr., from his two-decade dictatorial rule – could be “less cooperative in responding to PA surveys.”
The revered statistician then made a bold prediction, contrary to the Pulse Asia survey results: “[A]s a former teacher of probability theory and user of its concepts in my everyday life, including in the mind game of bridge, plus what Google Trends has been indicating for some time now, I dare say that the chances of Leni becoming our President in June are quite good!”
(In response, Pulse Asia said that if they followed Mr. Virola’s model, “a larger sample from the younger and more educated groups may actually inflate Mr. Marcos’s support.”)
Emergence of Google Trends
Google Trends has been an alien term in previous elections but has emerged this year as an alternative source of prediction, away from traditional polling and conventional wisdom.
Mrs. Robredo has consistently led Google Trends, obtaining 55 percent against Marcos Jr.’s 24 percent, a complete reversal of the Pulse Asia results.
Google Trends has become an unlikely indicator as traditional polling faltered in some countries.
It has accurately indicated the victories of the past four U.S. presidents (including Donald Trump in 2016) and three Canadian Prime Ministers, as well as the election results in Greece, Spain, Germany and Brazil. It recently predicted the reelection of Emmanuel Macron in France as he got 57 percent interest over his opponent’s 43 percent. Mr. Duterte, meanwhile, led this indicator in 2016, exhibiting high interest among online users in his populist candidacy.
“Google’s prediction is a departure from the surveys conducted by the different polling agencies, including SWS. Only time will tell if traditional surveys are more accurate than Google, or vice versa,” Andrew J. Masigan wrote in his column for The Philippine Star in March.
May 9 will be a litmus test for polling, a determination of whether survey veterans like Pulse Asia and Social Weather Stations need to reassess their methodology, which has been in place for more than two decades now, or whether online behavior is simply different from the ground, considering that trolls have permeated the internet landscape for years now.
Arnold Gamboa, a tech expert, explained that Google Trends only records the number of Google searches each candidate receives, which may connote interest. The question of whether that interest translates into votes is difficult to determine.
“What the big data show, for instance, in this current trend, is that 55% of people are more interested to know about Leni and what’s happening around her than with Marcos (24%),” he explained in a Twitter thread on Monday, May 2. “Majority of the time though it’s a positive search and not a negative search as shown in previous elections where Google Trends successfully predicted the winner.”
But political experts have cautioned the public from relying on Google Trends as an election indicator. Mr. Virola noted that it only looks for searches, unlike traditional surveys that specifically ask respondents whom they’re voting for.
OCTA Research Fellow Ranjit Rye said the results there should not be confused with polling data because they are completely distinct from each other.
“When you see a spike in a particular topic, this does reflect that a topic is somehow popular or winning. Only that for some set of reasons, it appears to be many users are performing a search about a particular topic. So Google Trends or the data coming from Google Trends can be used as a data point but it is not the same as what is derived from polling data,” he told One News’s The Chiefs on Apr. 27.
Litmus test
Still, Google Trends cannot be completely dismissed, given its track record of predicting winners in other countries.
Results of this year’s Google Trends and traditional polling firms have been divergent, with the former favoring Mrs. Robredo, the de facto opposition leader, and the latter Mr. Marcos Jr., son of the late ousted dictator. That’s why comparisons are being made on which is the better predictor of this seemingly dynamic race.
May 9 will be a litmus test for polling, a determination of whether survey veterans like Pulse Asia and Social Weather Stations need to reassess their methodology, which has been in place for more than two decades now, or whether online behavior is simply different from the ground, considering that trolls have permeated the internet landscape for years now.
Conventional wisdom has long been anchored on the science of surveys, which campaigns use as tools to calibrate their campaigns. A shocking turnaround would be an earthquake for the world of political science and polling.
As for the candidates’ fate, experts say that Mrs. Robredo has a small chance of pulling an upset, considering that Mr. Marcos Jr. still holds a double-digit lead. But as the saying goes, it ain’t over until it’s over.
The Edj View 
One thought on “A Tale of Two Predictors: Polling At Edge This May 9”