Damn statistics

User Rating:  / 2

It was the great journalist and writer Mark Twain who classified lies into three kinds: “There are lies, damn lies and statistics.”

I remember Twain after hearing my favorite outlier political analyst, Professor Antonio Contreras of De La Salle University, once again take issue with giant survey provider Social Weather Stations for coming out with another of those strange polls that he says are skewed to favor certain candidates. Contreras is talking about the SWS’ Sept. 2-5 “best leaders” poll, which had Malacañang’s candidate for president Mar Roxas in surprise second place with 39 percent, after regular survey-leading independent bet Senator Grace Poe with 47 percent.

The third major presidential candidate, Vice President Jejomar Binay of the United Nationalist Alliance, “slipped to third, up by a negligible percentage point,” with 35 percent, according to the newspaper BusinessWorld, which has exclusive rights to publish the SWS survey first. And right there, according to Contreras, lies the survey company’s grievous sin.

Adding up the percentages gathered by SWS gives a grand total of 121 percent, not including those racked up by the other possible presidential candidates like Davao City Mayor Rodrigo Duterte. This is because, as Contreras has pointed out, the survey was rigged to show results that are different from polls that would show an ordinary (and infinitely more logical) total of 100 percent.

SWS’ insistence on asking its respondents to name three candidates is clearly the culprit here. I will let Contreras explain, in his own words:

“This is a survey about the presidential elections, where there is only one name written on the ballot,” Contreras said. “It is misleading, to say the least, to keep on administering a survey that asks respondents to write up to three names.

“It is not only unscientific. It is also structurally designed to benefit some candidates and put [others at] a disadvantage.”

“Any statistician worth his/her degree knows that in order for surveys to have sense, the design and the conditions for administration of the survey must mimic, or if not approximate, the phenomenon being measured. In fact, the questionnaire must even appear like a sample ballot.”

That should be clear, even to SWS. But apparently, simulating an actual election is not the goal of the survey-taking company here.

* * *

I guess the more important question is not who the preferred candidate of the people in an election is, if it were held during the time the poll was conducted, but who benefits from conducting such a loaded survey. Contreras is not one to pussy-foot on this matter, either:

“This type of survey is advantageous to Poe and Roxas and not to Binay. Binay has a loyal base, people who thick and thin will stick with him. Hence, they will most likely write only his name. The Poe and Roxas voters are from the same demographic group and are most likely anti-Binay. Thus, there would be a significant number of respondents who would write both Poe and Mar, without Binay.”

In other words, because the voters who would most likely choose Poe or Roxas come from the same group, they would include both their names if they were allowed to put down more than one name in a survey. And because Binay’s likely voters are not voting for either Poe or Roxas, they will probably not choose someone else besides their chosen candidate.

Of course, this is Contreras’ theory, which cannot really be gleaned from the results of the SWS survey. But even if the survey company insists on conducting its strange survey method, it can easily find out if that theory holds, the professor explains.

“There would have been more meaning that can be drawn from the survey if SWS would also provide us the percentage of voters for each candidate who also wrote the name of the other candidates,” he says. “It would not have been an extra effort. All they had to do was to cross-tabulate.”

But the news media and even the candidates themselves are not helping to make sense of the SWS survey, which they report as if it was just a regular poll taken and interpreted the regular way. This, Contreras says, eliminates any possible value that may be derived from what is an important tool to judge public opinion at a given time.

“As it is, this SWS survey is one that would not even pass as a defensible undergraduate thesis,” Contreras says. “That it is now used by many politicians, opinion makers and media people without question, to my mind, makes SWS guilty of a serious offense not only against the science of statistics, but also against democratic politics.”

Lying through statistics is not new, of course. Just like losing even after winning the surveys isn’t new.