IT may take a while before the Philippines can maximize the benefits of using big data such as Google Trends, especially for elections, according to local statisticians.
Philippine Institute for Development Studies (PIDS) Senior Research Fellow Jose Ramon G. Albert told BusinessMirror that the country’s digital divide remains a major obstacle to the use of big data.
However, Albert said, studies on the use of big data such as those obtained from Google Trends have shown that social media data has been used for nowcasting and forecasting exercises in other countries.
“Right now, (there are no examples of big data use) for the Philippines but there have been instances in other countries where there is no tech divide (that this has been used),” Albert said.
Key to the use and reliability of big data is access to stable and high- speed internet connection as well as technology such as laptops and mobile phones that have the capability to connect directly to the internet.
Albert said this also requires that people know how to use the technology that they have access to. Knowledge and access are important if the country will be able to maximize big data.
“If not, then the signal you are getting is only from those with access and knowledge plus trolls,” Albert warned.
Given this, Albert said, it is difficult to disregard traditional data sources such as surveys. However, it is important to ask whether a data source is fit for an intended purpose.
Trolls
In a public post on social media, former National Statistical Coordination Board Secretary General Romulo A. Virola said traditional and new data sources, such as big data, can both be “infiltrated by trolls from all sides of the political spectrum.”
However, Virola admitted that big data, specifically that of Google Trends, can be “easier to manipulate” in favor of a certain individual.
Virola said trolls working for one side of the political spectrum can game the data and increase searches for the politician they represent to give that candidate an online boost.
Nonetheless, Virola said, the increase in searches online could also indicate greater interest with the intention of voting for a particular candidate.
For traditional surveys, Virola said trolls could also game these data sources through the commission of “huge nonsampling errors” which can happen when barangays or households have been compromised by trolls.
“This can happen if the list of sample barangays/households has been compromised because the trolls got possibly unauthorized information on the identity of the sample barangays/households, with or without knowledge of the survey firm. The trolls can then try to ‘influence’ how the respondents should respond to the survey questions, similar to vote-buying,” Virola explained.
Virola earlier said Pulse Asia election surveys oversampled lower socioeconomic classes and excluded higher-income classes which may skew the actual results of the May 9 polls.
In these surveys, former senator Ferdinand “Bongbong” Marcos Jr. was the leading candidate for President. The survey showed 56 percent of respondents favoring Marcos over Vice President Ma. Leonor Robredo with only 24 percent.
Online, data from Google Trends showed Robredo was the top choice, earning 52 percent of the searches over a 7-day period. Marcos Jr. accounted for only 22 percent of the searches.
