Election survey stories

Although scientific social surveys are about much more than elections, it’s a historical fact that their general acceptance was born in their prediction of Franklin D. Roosevelt’s victory in the United States presidential election of 1936. The litmus test for survey quality is its election track record. Success in picking winners is but normal; the real challenge is in being close to the official result.

There’s always a first time. Mine was prior to the 1981 election for the Interim Batasang Pambansa. I designed a province-wide survey with the same sample size in every municipality, to give their results equal accuracy. The aim was for our candidate to win an overwhelming number, if not all, of the municipalities.

The local pros who reviewed the survey findings found them plausible and unsurprising. Their explanations for the variation in voters’ preferences across towns were not based on socio-demographics, but on the local histories they knew. It was fascinating. Our candidate won the election easily, as expected. It was on his own merit, not due to the survey, which only served to verify the instincts of the provincial veterans. I didn’t try to recommend anything for the campaign; I never got to know what adjustments, if any, were made.

The 1987 senatorial election. In 1987-88, Social Weather Stations (SWS) was in partnership with Ateneo de Manila University to do two opinion polls per year, for two years, under a Ford Foundation grant. After the new Constitution was ratified, in February 1987, it was natural to start by looking toward the probable outcome of the first election, set for May. An important issue for our project’s steering committee was timing: do we field the survey as early as possible, thereby producing data more useful for the contenders, or as late as possible, thereby more likely to predict the results and enhance the project’s reputation? The committee decided that earlier was better.

Our March 1987 poll looked into prospects of several dozen candidates, of winning any of the 24 senate seats at stake in the May election. It found that the first 24 included 12 candidates from the slate of then President Corazon Aquino and 12 from other parties. The results were shown to all contenders.

When I presented the survey to Paul Aquino, President Cory’s campaign manager, he vehemently told his staff: “Do you know what this means, people? It means we cannot sleep anymore until election day!” They intensified their efforts, including getting the president to personally accompany the candidates on field sorties, doing what became known as “Cory magic.” The ultimate outcome of the May 1987 election was 22-2 in favor of the administration. (The two oppositionist survivors were Joseph Estrada and Juan Ponce Enrile.)

The biggest comeback was that of Santanina Rasul, who was about 50th place in the survey, but nevertheless won a seat. In March, she had polled some 2 percent of the vote, from the 4 percent who already knew who she was—in other words, her conversion rate was already a very high 50 percent. The campaign apparently introduced her to very many voters, and enabled her to climb by 26 ranks into the winners’ circle by May. (Of course, our critics claimed that the survey was wrong and/or that the administration had cheated, since how could 12-12 in March possibly become 22-2 in May?)

Choosing a candidate, Case 1. There was a political party that commissioned SWS to gauge who among three persons, let’s call them A, B, and C, would be able to defeat the incumbent provincial governor in the next election. We did a poll that tested scenarios of each of the three challengers versus the incumbent and other potential contenders.

SWS found the strongest one to be A, who was an incumbent congressman of that province; he had a very good chance to win, but would have to sacrifice his current seat in order to enter the race. The next strongest, B, could also win, but by a smaller margin. The third candidate, C, was likely to lose. The party chose to field candidate B; he ultimately defeated the incumbent.

Choosing a candidate, Case 2. There was another political party with three persons, let’s call them X, Y, and Z, all wanting to run for governor in the same province. The party asked SWS for a poll about which of them should be its candidate. Again, SWS tested various scenarios of the three, versus other potential candidates. It found X as the most popular, and then Y, and then Z.

Although all three had pledged to support the one who ranked first in the poll, they ultimately broke the pledge, and all entered the race, against each other and other contenders. Candidate X won the election. I don’t remember anymore whether Y came out ahead of Z.

There is much room for candidates to improve their standings. In the current national election races, there is still much room for laggards to improve. Keep an eye on the fill-up rates in the preelection polls for senators and party list; they are still far from complete. Many votes are still available, even without campaigning for voters to change their minds.