PRE-ELECTION survey results for the May 2022 polls released by Pulse Asia (PA) may have undersampled young voters and oversampled older voters in their survey, according to a retired government statistician.
In a public social media post, former National Statistical Coordination Board Secretary-General Romulo A. Virola said based on his estimates, the PA survey may have undersampled the 18-24 age group by 46 percent and the 25-34 age group by 12 percent.
In the same vein, Virola said, the PA survey results may have also oversampled the 55-64 age group by 46 percent and 65 years old and over by 38 percent.
“Our short analysis about the distribution by age of the PA sample respondents is something else. And it could very well spring a surprise when the actual results of the May 2022 elections come out if in fact, there is a youth/young vote, and/or many of the older voters would/could not vote due to Covid-19 or for other reasons,” Virola said. “The Pulse Asia 18-23 February survey could be way off.”
Pulse Asia: No such thing
PULSE Asia President Ronald D. Holmes denied to BusinessMirror that the company over or undersampled in its election surveys. He said, nonetheless, that they will be validating their results his week.
Virola explained that ideally, sampling to be used in surveys should mirror the distribution in populations.
This means, if for example, the share in the population of 18 to 24 year olds is 19.65 percent, the same share should be used in a survey sample. In the PA survey results, the share is 10.58 percent of this age group.
Nonetheless, Virola said, it was possible that cost constraints may have played a role in increasing or decreasing the share of certain demographics in a survey.
Virola said it is the call of Pulse Asia to determine if cost savings and other considerations are worth taking given the possible impact on the validity and accuracy of their results.
“Of course there are cost considerations—a constraint on sample size—that’s why you can’t use all the stratification variables you want to include [in the survey]. But when reporting the results, a post stratification is done,” Virola told BusinessMirror in a mix of Filipino and English.
“There needs to be a deeper study of whether age or other potential stratification variable like sex is indeed a determinant of voters’ preferences and the budget for the survey. I suppose Pulse Asia’s resources are not unlimited,” he added.
Data crunch
Based on Virola’s data, the share of respondents aged 25 to 34 years old in the PA survey was at 21.92 percent, lower than the share of the age group in the country’s population at 24.97 percent.
For ages 35 to 44 years old, these account for 21.5 percent of PA survey respondents but as a share of the Philippine population, this age group only accounted for 19.8 percent.
The age group of 45 to 54 years olds represented 17.42 percent of PA respondents but accounted for only 15.52 percent of the Philippine population; 55 to 64 years old, a share of 16.13 percent in the PA survey but only 11.04 percent of the country’s population; and 65 and over, 12.46 percent of PA respondents but only 9.01 percent of the entire population.
As a percentage of Comelec voter registration, Virola said, the same under and over sampling was observed. The Filipinos aged 18 to 41 years olds accounted for 47.55 percent of PA respondents but account for 56.32 percent of Comelec registered voters.
The data also showed that Filipinos aged 42 to 57 years old accounted for 28.7 percent of the PA sample but account for only 25.53 percent of the voting population.
The 58 years old and over Filipinos, who comprise 23.75 percent of the PA sample, only accounted for 18.15 percent of total Comelec voters.