Surveys validate surveys

A survey does not reach all voters, but only a sample of them. The science of statistics owes much to democracy, because elections have a system to count the votes of the full population, validating the accuracy of the votes counted in a sample. In the survey profession, worldwide, success in predicting elections is the litmus test of quality in sampling.

It is a test hurdled many times by Social Weather Stations, which has the longest experience in election surveys. (See “The challenge of election surveys: SWS experience in the Philippines,” Second International Conference on Public Opinion Polls, “Public Opinion Polls in a Changing Society,” Cairo, Egypt, November 2009, on www.sws.org.ph.)

No media-company sponsor in 2019. Unlike in 2016, when BusinessWorld and TV5 were our sponsors, there was no media company that expressed interest in partnering with SWS for election surveys, for the first time since the 1992 season.

I’m not aware of any media-sponsored election surveys at all. Can the big companies, especially those with election advertisements, explain? Unfortunately, an election survey is too expensive for us to do on our own account.

Our 12-12 score in March 1987. Back in 1986-87, SWS partnered with Ateneo de Manila University to do four national public opinion polls, in a joint project sponsored by the Ford Foundation. After doing two in May and October 1986, our steering committee met in early 1987 to decide the timing of the third round, bearing in mind that the election of the new 24-member Senate would be in May.

To maximize success in predicting the winners, a survey should be done as close to election day as possible. On the other hand, to be most appreciated by political contenders, it should be done as early as possible. After discussion, the committee set the survey for March 1987.

We presented the March 1987 survey finding to many political groups—that the leading candidates were split into 12 Cory candidates and 12 from the opposition. Alarmed, Paul Aquino, the administration’s campaign manager, told his people then and there: “Do you know what this means? It means we cannot sleep any more from now until the election!”

In the 1987 campaign, President Cory Aquino was brought into play, to raise the hands of her candidates across the nation. In what became known as “Cory magic,” 22 of her candidates won; only Erap Estrada and Juan Ponce Enrile survived. Critics of our survey made the laughable claim that the 22-2 score was proof that it had failed.

Look across many samples. Validation of a sample survey does not even require a population count; what it needs are other, preferably many, surveys. What “95-percent confidence” really means is that, from say 20 sample surveys, 19 of the resulting 20 sample proportions, plus/minus the margin for sampling error, will match the true population proportion.

For instance, the SWS quarterly surveys on Gainers/Losers and Optimists/Pessimists are validated by the Bangko Sentral’s quarterly Consumer Outlook Surveys. SWS Joblessness rates match the total Unemployment plus Underemployment rates of the government’s Labor Force Surveys.

SWS’ quarterly Self-Rated Poverty trends—meaning changes from one time-point to another—have regularly anticipated the trends periodically found from applying the official poverty line to the

triennial Family Income and Expenditures Surveys (see “Consistency in poverty trends,” Opinion, 4/13/19).

SWS congratulates the Bohol Poll of Holy Name University, in Tagbilaran, for successfully anticipating the results of the 2019 gubernatorial election!

Contact mahar.mangahas@sws.org.ph.

Read more...