Eds note: This is Part III of Romulo Virola's analysis of Pulse Asia's survey conducted in February this year.

Part I of my attempt to analyze the Pulse Asia (PA) February 18-23 survey, vis-à-vis other data sources did not disclose anything that I consider serious misalignment of the PA sample whether we compare with the regional distribution of (1) the voting age population as projected by the PSA or (2) the registered voters from COMELEC.

• Statistically Speaking: Some questions on data and surveys for the Halalan 2022

But Part II discovered that PA underrepresented the younger ages and overrepresented the older ages again whether we compare with the PSA voting age population or with the COMELEC (sourced from CNN Philippines) registered voters. Thus, we concluded that if there is a youth/young vote and/or if the older voters would or could not vote in May 2022 for some reason including the COVID 19, the final election results might be different from what the PA survey had predicted.

• Statistically Speaking: Gen Zs, Millennials underrepresented; 'Boomers' overrepresented in Pulse Asia Feb survey

For Part III, we follow up the investigative analysis by comparing the educational attainment of the PA sample respondents with the educational attainment of the voting age population from the PSA Census of Population 2015, noting from Parts I and II the seeming congruence of conclusions whether the comparison is made with the COMELEC registered voters or with the PSA dataset.

Pulse Asia (PA) overrepresented those who did not reach college representing more than 78 percent of the PA sample and almost 70 percent of the PSA 2015 Census of Population and consequently underrepresented those who did:

• PA overrepresented those who did not go to college (at most vocational graduate) by 12 percent; and underrepresented those who had at least some college education by 28 percent;

• PA heavily overrepresented those who had a vocational education by 92 percent, almost double their share of the voting age population, although the overrepresented group represents less than 3 percent of the voting age population; (please take note however, that the comparison is with the 2015 voting age population structure)

• Also overrepresented by 33 percent are those who are high school graduate (no vocational, no college) representing 30 percent of the voting age population.

In summary, the Part III findings of overrepresentation those who did not reach college and underrepresentation of voters with at least some college, plus the findings in Part II of underrepresentation of the youth/young voters and overrepresentation of the older voters support the possibility of a significant sampling bias in the PA survey of 18-23 February 2022!

Of course, we are not certain if educational attainment and age would be critical factors in our final choice who to vote for on May 9! My gut feeling however, is that both age and educational attainment might be!

Without presuming that my analysis is correct, let me now offer some friendly but professional warning/ advice/ suggestions to Pulse Asia!

• There is probably still time (before your last pre-election survey) to take a good second look at your sampling scheme!

• Totally removing the AB socio-economic classifications from your sample may have contributed to the possible bias and therefore may have not been a good, although apparently practical, and even sampling theory-based move!

• In Part II, a colleague asked: “how do we account for the people who said..this time boboto na ako after getting burned for not voting in 2016?” which leads to the question: Does the sampling frame used by PA cover these “new voters”?

• Lastly, is considering a radical change in the sample selection (maybe even a 100 percent sample rotation?) still logistically possible, if only to guard against possible infiltration by systematically biased respondents? Or if at all possible, to increase the probability that their indicated individual survey preference will be reflected when they vote on May 9, 2022?
Remember the Literary Digest debacle in 1936?

Reactions may be emailed to ravirola@yahoo.com or sent via Facebook.

Endnotes:

Romulo A. Virola is formerly with the Actuarial Research and Development Group of the GSIS, and a former Professorial Lecturer in the Graduate Programs of the Statistical Center/School of Statistics and the Department of Mathematics (now the Institute of Mathematics) of the University of the Philippines . He retired in 2012 as Secretary General of the then National Statistical Coordination Board (NSCB) of the Philippines or NSCB (now part of the Philippine Statistics Authority). He finished his BS (Mathematics) from UP, and MS (Actuarial Mathematics), MA (Statistics), and PhD (Statistics) from the University of Michigan, where he was a Fellow in its Sampling Program for Foreign Statisticians under the late Prof. Leslie Kish, author of the pioneering “Survey Sampling”, considered by many as the bible in the field. He used to write/co-write the Statistically Speaking articles posted on the NSCB website from 2004 until his retirement. The author thanks his former colleagues in the National Statistical System (NSS) of the Philippines particularly from the NSCB, Jay Mendoza of IOM, UN Migration Agency and some FB adviser-friends-former classmates for the assistance and support in sustaining the preparation of posts for the Statistically Speaking v.2.0. We also thank FB friends and readers who contribute their insights towards a better appreciation of the issues and concerns raised by Statistically Speaking v.2.0.

This article was originally published on the author’s Facebook page on April 5, 2022. Since then, Pulse Asia has come out with 2 other surveys on April 6 and March 2, and the PSA released on March 28 an Updated Projected Mid-Year Population for the Philippines Based on the 2015 POPCEN Results: 2020-2025.

Sources:

https://psa.gov.ph/content/updated-projected-mid-year-population-philippines-2020-2025

https://psa.gov.ph/population-and-housing/node/166426

https://www.pulseasia.ph/february-2022-nationwide-survey-on-the-may-2022-elections/

RELATED STORIES

• Vote-rich provinces: Who won where in 2016 and some prospects for 2022
• Marcos tops Pulse Asia's final survey among pres'l bets; Pacquiao overtakes Isko

Disclaimer: The views in this blog are those of the blogger and do not necessarily reflect the views of ABS-CBN Corp.