The question of surveys continues to puzzle people, primarily because of the suggestion that aside from opinion polling, there may be other ways to gauge the standing of candidates (see my previous column and entry on this). Taking a look at Philippine elections: 7 reasons why a Leni Robredo win ‘very possible’ includes this, which helps explain why talk of Google trends and other things, is getting a serious hearing:
On March 6, 2022, Dr Guido David, a Professor of Mathematics at the University of the Philippines, twitted this image from the Marcos camp, and wrote: “Surveys can be wrong at predicting the true voter preference. We have seen too many upsets, especially in recent times, where the electoral outcomes defied the industry’s prediction.”
“At least 6 surveys/polls point to a BBM (Marcos Jr’s) victory by a large margin. Each of these surveys is based on a sample size of between 1,500 to 2,400 respondents only. Statisticians generally concur that this sample size is enough to represent the voting population. Except for one of the polls, most were done in January. Thus, these polls may have largely failed to capture changing preferences in February.”
David added: “We believe it is possible to get a better ‘pulse’. Using more than 50 million social data points, our analysis reveals a much tighter race. A Leni victory over BBM is very possible, with only minor tactical changes needed from Leni’s campaign team.”
On the other hand, the widely-noticed (and commented upon) lack of enthusiasm and verging on anemic attendance at the Marcos rallies, the delirious enthusiasm and respectable, even impressive, attendance at the Roredo rallies, not to mention recent endorsement by local government figures, makes the Marcos lead in the polls, somewhat counter-intuitive.
My column addresses this; as well as another question being discussed, which is, will the President finally endorse someone, and who might that candidate be? The recent pointing to the President, by some governors who have endorsed Robredo, to me is misleading. They are using the President as speculative cover when the real calculus is which presidential candidate is popular in that governor’s locality.
This week’s #ProyektoPilipino on the other hand, also discusses surveys, but from a civics perspective.
For additional readings, first, a slide shared on Twitter by Wilson Chua, illustrating his over-all point that the surveys are a delayed snapshot of a dynamic trend that FaceBook and Google more quickly and accurately capture (as for Chua’s methods, etc. see the desx Leni Robredo leads in Social Media, Google Trends and Voters’ Enthusiasm over Bongbong Marcos & other candidates).
Second, an entry by Cebuano journalist Max Limpag on the technology of online sentiment as a predictor of electoral outcomes: Google searches “accurate predictors” of presidential election results, researchers say:
Google Trends predicted the winners of the past two presidential elections in the Philippines. Except for former President Donald Trump’s loss in 2020, Google Trends predicted the winners of the previous four presidential elections in the United States and the past six in Canada since the company’s search data were made available in 2004.
Google Trends is a portal that allows people to explore large data on search requests. Its tracking of search interests for presidential candidates, according to the paper “Google Trends as a Predictor of Presidential Elections,” are “accurate predictors of election outcomes.”…
The paper by the Universidad Rey Juan Carlo researchers said there have been many studies on the impact of digital media on campaigns. The studies have one common theme, “that the volume of keywords searched on the internet or contained in chats on social networks such as Twitter are revealing the current and future thinking of a significant amount of the population.”
The paper said Google Trends “has predictive capacity in short periods of time and even under a high-volatility scenario such as the Greek referendum in 2015 on the approval or rejection of conditional financial aid from the European Union.”
There is also “strong correlation between potential voters’ searches and election results” in the national elections in Greece and Spain and with the German elections of 2005, 2009, and 2013.
The researchers predicted that the winner of the 2020 US election would be “the one that arouses the greatest interest among the population of the United States of America and generates the most Google searches.” That missed the mark – Trump lost, even though he does not accept it, despite getting more search interest.
But the researchers added an important caveat, “as long as these searches are related with positive news and the environment remains stable.” In analyzing the 2016 elections that Trump won over Hillary Clinton, they said that “conflicting communication techniques used by Trump led to a greater internet search, but generated a negative bias.”
Third, an interesting entry putting the case forward for what Wilson Chua and others claim: Pink Storm Rising (click on the link to see the charts and graphs):
Our models show the contest is closer than what the surveys predicted… And in fact, by early March, Engagement scores show Leni and Kiko forging ahead of BBM and Sara Duterte. Note: the use of sentiments has been shown to be better at predicting poll wins.
To explain the difference, we look at two other charts; cumulative positive sentiment and cumulate negative sentiment. Leni has a higher positive, and BBM has a higher negative.
The simple explanation is just people who are exposed to Leni or Leni’s propaganda, LIKE her more than BBM. A sector of the public doesn’t like BBM, despite his saturated online presence and higher virality. This new situation may see the BBM Sara camp join the debates. They have nothing more to lose.
Fourth, a valuable observation by way of John Nery, on the apparent shared skepticism of Marcos and Robredo supporters when it comes to surveys: Dynamic race, static survey?
Jason Haw, a PhD epidemiology student at Johns Hopkins I’ve followed since the pandemic began for his facility with numbers, suggests three “hypotheses” to explain the “paradox” of a surge in the ground for Robredo that is not reflected in the surveys. I have been wrestling with exactly the same ideas since the Cavite rally, but without Haw’s clarity of thought.
First, it is possible for millions of Filipinos to attend Robredo’s rallies, and yet for Robredo to still lose. If say 20% of all the 65.7 million registered voters cast their ballot for Robredo, that is 13 million votes.
Second, it is also possible that the surveys, this time, are wrong. Haw raises the possibility of “some kind of Hawthorne effect” that skews the findings, because the survey respondents “are observed by family and neighbors when these interviews are done.” I am reminded of that Japanese study, conducted in Metro Manila early last year, that concluded that as much as a third of President Duterte’s popularity rating may be due to “social desirability bias.”
Third, it is also possible that the surveys are, in fact, right. Marcos enjoys majority support, because (as I have argued before) he and his family have been seeding disinformation on digital and social media for years.
Thank you to those who are contributing to Patreon and thus helping provide the resources required to keep producing my newsletter and podcast.
Consul: Abigail Salta
Praetors: Carlos v. Jugo, Ramon Rufino, Arbet Bernardo
Aediles: Steven Rood, Willi, Cleve Arguelles
Quaestors: Joseph Planta, Giancarlo Angulo, Annie Inojo,
Sam Chittick, Patrice P
Become a patron of Manolo Quezon is #TheExplainer today: Get access to exclusive content and experiences on the world’s largest membership platform for artists and creators.