Revelations of the TV5-SWS Exit Poll

The many revelations of last Monday’s TV5-SWS Exit Poll demonstrate why exit polling is valued all over the world, even where official vote-counting is very speedy and reliable. There is so much more to learn about the voters than merely which candidates they voted for.

A clear decision for the presidency, a close fight for the vice-presidency. Although its first findings were telecast at 6:25 p.m., by 10 p.m. the exit poll had come up with the following results, from 45,123 voters interviewed upon leaving 733 out of 802 voting centers (VCs), drawn nationwide with probabilities proportional to the numbers of registered voters at the levels of the region, the province, the city (in Metro Manila), and the VC itself. (See InterAksyon.com posting of May 10, 12:35 a.m. Balances from 100.0 percent are refusals to answer.)

For president: Duterte, 39.3 percent; Roxas, 24.0 percent; Poe, 19.2 percent; Binay, 13.5 percent; and Santiago, 3.1 percent.

The 15.3-point lead of Duterte over Roxas—rather than over Poe, who was 2 points ahead of Roxas in the May 1-3 SWS survey—showed his clear victory.

For vice president: Marcos, 34.6 percent; Robredo, 32.8 percent; Cayetano, 14.9 percent; Escudero, 11.0 percent; Honasan, 2.5 percent; and Trillanes, 2.1 percent.

The mere 1.8-point lead of Marcos over Robredo at that point is not statistically callable given both the sampling error (which depends on the sample size) and the sampling-design effect (which depends on the stages of probability-based drawings).

Thus, the virtual tie of Marcos and Robredo as of May 1-3 held on in the exit poll. It can only be resolved by the full vote count by the Commission on Elections.  It is a tenet in statistics that the (one and only) full population count is what judges the quality of any probability sample count. A probability sample count (and there can be multiple such counts) cannot judge the quality of the full population count. Nonprobability samples do not deserve consideration. The overseas vote is outside the scope of the exit poll.

The voters made their choice for president a little later now.  In the 2016 exit poll, 18 percent of those interviewed said they made their choice only on Election Day itself. Another 15 percent decided only during May 1-8, 12 percent did it in April, 8 percent in March, and 46 percent in February or earlier.

In the TV5-SWS 2010 exit poll, 15 percent said they decided only on Election Day, May 9, and 51 percent said they had decided by February.

But the delay in 2016 was not peculiar to any candidate; it affected them all. Sixteen percent of Duterte voters, 18 percent of Roxas voters, 20 percent of Poe voters, 20 percent of Binay voters, and 19 percent of Santiago voters all decided only on May 9 itself.  Taking time to decide is a characteristic of the voters, not of the list of candidates.

The voters of 2016 ignored the pairings of the candidates even more than in earlier elections. As of 2 a.m. on May 10, when the 2016 exit poll sample had reached 62,485 voters from 785 of the 802 VCs, the Duterte vote percentage had reached 40 points, for a 16-point lead over Roxas.  Of his 40 points, only 13 came from voters of his cocandidate Cayetano; the bulk of 18 came from Marcos voters, and another 6 were from Robredo voters.

Of all the tandems, only Ro-Ro really appealed to the voters.  Of Roxas’ vote points, then at 23, the bulk of 18 came from Robredo voters.

Poe’s vote points, then at 19, were split at 6 points each from voters for Marcos, her cocandidate Escudero, and Robredo.

Of Binay’s 14 vote points, 7 came from Marcos voters, while 2 came from Robredo, and only 1 came from his cocandidate Honasan.  Of Santiago’s 3 vote points, 1.5 came from her cocandidate Marcos.

Big wins for Duterte in Mindanao and Metro Manila. As of 2 a.m. on May 10, Duterte’s lead over Roxas was a massive 48 points in Mindanao, and a large 22 points in Metro Manila.

In the Visayas, Roxas was 4 points over Duterte.  In the rest of Luzon, Duterte was only 3 points over Poe, but 8 points over Roxas. The full Comelec count has the final say on the geographical pattern.

Only an exit poll has more details than geography. As of May 10, Duterte’s lead over Roxas was 26 points in urban areas, but only 2 points in rural areas.

The higher the class, the more the appeal of Duterte: His lead over Roxas was 26 points in class ABC, compared to 17 points in class D, and only 7 points in class E.

The more the schooling, the more the appeal of Duterte: His lead over Roxas was 28 points among college graduates, 19 points over those with some college, 8 points among those with some high school, and 7 points among others.

The younger the voter, the more the appeal of Duterte: His lead over Roxas was 33 points in ages 18-24, 26 points in ages 25-34, 14 points in ages 35-44, 10 points in ages 45-54, and 4 points in ages 55 and up.

Duterte’s lead was 22 points among men, versus only 12 points among women.

Duterte was least supported by Catholics. Duterte led Roxas by 16 points among all voters, but by a below-average 10 points among Catholics.  He led massively by 53 points among Muslims, by 70 points among Iglesia ni Cristos, and by 24 points among other Christians.

Platform now counts for more than personality. In the 2016 exit poll, 57 percent said they voted for president on the basis of platform, while 40 percent said they voted on the basis of personality. This item had been evenly divided at 46-46 in the 2010 exit poll.

* * *

Contact mahar.mangahas@sws.org.ph.