The Bilang Pilipino SWS Mobile Survey is a pioneering project to rapidly track the opinions of a statistically representative national panel of voters, deliberately equipped with mobile phones for the purpose. It is a joint undertaking of Social Weather Stations, TV5 Network Inc., Voyager Innovations Inc., Philippine Star, Starmobile, and Smart Communications to generate accurate and up-to-date data for news reports and analyses on the elections. This type of survey project has never been done in the Philippines before.
A panel was initially established by means of standard random sampling. On March 8-11, SWS drew a random sample of 1,200 validated voters nationwide, using its standard statistical procedures, for face-to-face interviews, in their homes.
The sample size consists of 300 voters each in the National Capital Region, the Balance of Luzon, Visayas, and Mindanao. The size is the same in each area in order to give them equal accuracy in the analysis. To obtain national figures from the area samples, NCR gets 12 percent weight, Balance
Luzon gets 44 percent weight, Visayas gets 21 percent weight, and Mindanao gets 23 percent weight.
The respondents were invited by the field interviewers to be members of a panel to receive and reply to two or three survey questions each day, using mobile smartphones provided to them for free by Starmobile. Those that did not agree were replaced by continuing with the random sampling technique.
Thus, this project is not a survey of mobile phone owners. It does not require prior ownership of a mobile phone by panel members. Upon agreeing to join the panel, the member is provided with a new touch-screen phone that is capable of receiving survey questions, recording the answers, and transmitting the answers to SWS. The SWS field interviewer trains the panel member on how to use the phone.
The randomly chosen sample spots were checked for presence of a mobile phone signal, and for electricity to enable recharging of the phones. If either is absent, then the nearest barangay with a signal and with electricity is taken instead.
The survey software for the project is provided by Voyager Innovations. There is no cost for receiving questions and submitting answers, courtesy of Smart Communications. Panel members sign a pledge that they will answer the questions personally, and not assign the phone to anyone else. After participating in the project up to the end of May 2016, they may keep the phones.
The initial survey of March 8-11 obtained the usual background information about the panel of respondents by face-to-face interviews. These backgrounders show that the panel is a reasonable representation of the voting population.
By design, the panel’s sex ratio is 50:50 men and women. From the sampling, the panel got an age distribution of 14 percent aged 18-24, 22 percent aged 25-34, 23 percent aged 35-44, 19 percent aged 45-54, and 21 percent aged 55 and up. Its educational distribution is 9 percent with unfinished elementary, 24 percent with unfinished high school, 52 percent with some college, and 14 percent college graduates. Its socioeconomic class distribution is 5 percent ABC, 81 percent D, and 14 percent E.
The mobile phone technology enables the survey items to be fielded nationwide in a single day, processed the next day, and reported on the third day.
The first mobile survey was done on Monday, March 14. On Wednesday, March 16, TV5 was able to report that 64 percent of voters already knew (i.e., even before hearing about it in the survey) of the Supreme Court ruling that Sen. Grace Poe is qualified to run for president. The survey also found 66 percent agreement and 15 percent disagreement with the ruling.
The response rate of the first mobile survey was 71 percent (848 of the 1,200 original panel members), allowing a national error margin of roughly 3 percentage points.
Thus far, the mobile survey has been run five days a week, from Monday to Friday. For the first 12 mobile surveys, the response rates have been: March 14, 71 percent; March 15, 63 percent; March 16, 64 percent; March 17, 72 percent; March 18, 74 percent; March 21, 66 percent; March 22, 67 percent; March 23, 65 percent; March 24 (Holy Thursday), 61 percent; March 25 (Good Friday), 57 percent; March 28, 62 percent; and March 29, 63 percent.
Despite the falling response rates, the demographic characteristics of the responses have been very stable, throughout these 12 surveys. This means that the attrition of the sample is not accompanied by demographic bias. The way to handle attrition is, of course, to augment the panel periodically; we have requested the sponsors for assistance on this.
On March 21, the day after the second presidential debate, the mobile survey found that 52 percent of voters had seen it on television, and that the best debaters, as rated by the voters, were Poe (39 percent) and Duterte (31 percent).
On March 22, two days after the debate, the mobile survey found the following voting percentages in the presidential election race: Poe 35, Duterte 26, Binay 18, Roxas 17, and Santiago 2. By tracing how the panel members had adjusted their votes between March 18 and March 22, SWS was able to report, on March 24, exactly how many points had been transferred from one candidate to another.
The rapid feedback from the panel mobile survey is enabling us to quantify just how stable or unstable are the attitudes of voters. The race is quite dynamic. It’s an exciting time for election research.
* * *
Contact mahar.mangahas@sws.org.ph.