SWS mobile survey valid despite sampling challenges, say experts

Despite challenges in the sampling of respondents, the results of mobile surveys conducted by the Social Weather Stations should remain valid, according to experts.

“The report is quite detailed. That is more than valid.... The report is very good,” University of the Philippines Professor Josefina Venegas-Almeda said.

Almeda noted that the latest report, released on March 24, was very clear as it contained not only percentages but actual numbers of respondents.

“The fact is that the report was very detailed and all the tables were shown,” she said.

In its latest such survey conducted in partnership with TV5, only had 806, or 67 percent, of the respondents answered the poll through the mobile phones they were given.

Almeda, who specializes in Applied Statistics and Statistical Computing at the University of the Philippines School of Statistics, told GMA News Online that the margin of error of the survey covers for the discrepancy.

The mobile survey, which was conducted on March 22 or two days after the second leg of the PiliPinas Debates 2016, showed Senator Grace Poe pacing all presidential candidates with 35 percent.

Davao City Mayor Rodrigo Duterte had 26 percent; Vice President Jejomar Binay, 18; former Secretary Mar Roxas, 17; and Miriam Defensor Santiago, who didn't attend the debate, two.

One percent of the respondents said they were undecided or were voting for other personalities.

Out of the 300 respondents for each geographic group, 212 responded in Metro Manila, 190 in Balance Luzon, 221 in Visayas and 183 in Mindanao.

"This gives sampling error margins of ±3% for national percentages, and ±7% in Metro Manila, Balance Luzon, Visayas and Mindanao," SWS explained in its March 24 press release.

Sampling challenges

In an article for the Massachusetts Institute of Technology (MIT) News, science writer David Chandler detailed how the margin of error of polls already covers what is called the nonresponse error.

"Pollsters begin by attempting to reach a certain randomly selected set of people that is representative of the overall population — for example, by generating a list of random phone numbers. But there are two problems: Sometimes nobody answers the phone, and even when someone does answer, they often — and increasingly — refuse to respond," he wrote.

Almeda, for her part, said that conducting the survey via mobile opens up sampling to other challenges.

She cited as potential problems the possibility that respondents might not respond to the questions right away or they might let another person answer the survey questions.

According to the SWS, two or three questions are sent to panel members at 7 a.m. each survey day through the mobile phones given to them. They can submit their responses only until 12 noon.

In its contract with the SWS, panel members agreed they will be the only ones who will answer the mobile surveys; they are not allowed to let other people answer the questions.

But Almeda reiterated that these challenges should already be covered by the margin of error.

Professor Danilo Arao of the UP Department of Journalism agreed that the SWS' margin of error should make for valid findings even if not all of the 1,200 respondents actually answered the poll.

"The sample is still representative and to be fair, SWS was above board when it came to a bigger margin of error," Arao said, when shown the rates of actual respondents per geographical group.

Arao was a former editor-in-chief and head of research of Ibon Data Bank, an independent think-tank.

Before the post-debate survey was conducted, the respondents were interviewed face-to-face on March 8 to 11 and equipped them with mobile phones.

A respondent should be a validated voter is a registered voter whose biometrics have been captured digitally.

A survey was also conducted on March 18 where 883 or 74 percent out of the 1,200 respondents replied.

Survey questions will be sent to the respondents until May. —NB/JST, GMA News