Passing the election test again

In SWS’ view, its survey work has, again, handily passed its self-imposed periodic test of accurately anticipating election results. Our report, “SWS/BW final preelection survey of  May 2-3, 2013: 9 Team PNoy, 3 UNA in top 12; 7 Team PNoy, 2 UNA safe,” announced on May 8 (www.sws.org.ph), proved entirely correct, compared to the latest actual results, official or unofficial. It correctly preidentified all winners and losers.

Let me emphasize that the full population count validates the quality of a sample survey, not vice versa. The honesty of a full count should not be judged according to expectations generated by a sample survey. Those who criticize the full count should consider direct evidence (for instance, a parallel full count), not sample survey results.

A few surprises.  It is precisely because of the good reputation of the major opinion polls that many observers did not expect Grace Poe (5th as of May 2-3) to be first, and Sonny Angara (tied for 10th on May 2-3) to be sixth. But these are minor surprises. The true challenge in senatorial polling is in distinguishing winners from losers.  SWS called 9-3 the most likely result, with 10-2 or 8-4 less likely but possible, and only named seven candidates as safe.

How could Grace and Sonny have improved their standings so much? The “how” question is easier to answer. Voters had room to add one, two, or even three more names to their choices, without having to switch between candidates. As of May 2-3, only 44 percent of those surveyed had a full slate of 12. Their average fill-up rate was only 8.7 names; 6.2 percent had no choices at all.  (In 2010, the TV5/SWS exit poll found only 31 percent with a full slate. The average fill-up rate was 7.9 names; 5.8 percent gave not a single name.)

Answering the “why” question calls for inside information. To explain why voters added more names at the last moment requires knowledge of the last 10 days of the campaign—or what Ateneo Dean Tony La Viña aptly calls the “end game,” as in chess. Not being part of campaign operations, SWS has no expertise on end games. Descriptions of campaign openings and middle games are not enough. What were the end games?

Survey quality. For each election test, SWS rates itself by the number of winners identified, and the correlation of the final preelection survey ranks and the actual ranks of the candidates. In 2013, the rank-correlation is 97.0 percent, across the 33 candidates; all the winners were identified (though this is tentative, until the Commission on Elections issues its final report).

In 2010, the rank-correlation of the SWS May 2-3 senatorial survey with the May 10 election outcome was 98.5 percent; all winners were identified. In 2007, the rank-correlation of the SWS May 2-4 survey with the May 14 outcome was 97.1 percent; 10 winners were ultimately (excluding Zubiri due to his resignation) identified.  In 2004, the rank-correlation of the SWS May 1-4 survey with the May 10 outcome was 98.1 percent, with all winners identified. In 2001, the rank-correlation of the SWS May 4-7 survey with the May 14 outcome was 98.3 percent; 10 winners were identified.

Vote-mixing.  The 9-3 result of 2013 was due to the voters’ desire to put a mixed group of candidates—something like an “all-star team”—into the Senate, rather than due to partisan divisions among them.

From the 12 Team PNoys, the 9 Team UNAs, and the 12 “others,”  voters could choose purely from one group, or take a mix from two groups, or even mix up all three groups. In the May 2-3 SWS survey, the most popular move, taken by 44 percent, was to mix up some PNoy and some UNA candidates. The next most popular move was to mix up the three groups (36 percent). Thus, the top two mixes already satisfied 80 percent of the electorate.

Then came the pure PNoy voters (7.2 percent), the pure UNA voters (2.3 percent), the PNoy-others voters (2.2 percent), the UNA-others voters (1.2 percent), and the pure others voters (0.5 percent); the remainder had no choices at all.

The religious vote was not critical.  As of May 2-3, the ranks of candidates among Catholic voters were the same as the national ranks, since Catholics are so dominant. For weekly churchgoing Catholics (WCC), the top 12 had only slightly different ranks than among all Catholics (AC): 1. Legarda, 2. Binay, 3. Cayetano, 4. Escudero, 5. Pimentel, 6. Poe, 7-8. Ejercito and Villar tied, 9. Aquino, 10. Angara, 11. Trillanes, and 12. Honasan. Enrile was 13th in both AC and WCC.  Hontiveros was 17th in WCC, but 16th in AC. Mitos Magsaysay was 20th in WCC, but 21st in AC.  Casiño was 24th in both AC and in WCC.

The Iglesia ni Cristo endorsed all the eventual Team PNoy winners, except Escudero and Pimentel, and endorsed all the eventual UNA winners, but added Enrile and Gordon. As of May 2-3, before the endorsement, the top 12 choices of INC voters already included Enrile and Zubiri, and excluded Aquino and Angara. Thus the INC solidified the wins of Aquino and Angara.  The net beneficiary of the INC was probably Gordon, since his original INC votes before the endorsement were much less than Enrile’s original INC votes; thus, getting an equal number of INC votes could have advanced Gordon to 13th, over Enrile.

The best means of analyzing the vote would have been an exit poll, but unfortunately the big media networks did not sponsor one this year.  The MBC-DZRH exit poll, reported by Solar News, was a dud because it named Enrile (8th place) and Zubiri (12th place) as winners, at the expense of Pimentel (13th place) and Honasan (14th place).

* * *

Special tabulations for this piece were done by Josefina Mar of SWS. Contact SWS or [email protected].