IS IT A TIE?: Postscript has questioned some partisans’ calling a “statistical tie” the close ranking of presidential bets Noynoy Aquino (36 percent, Liberal) and Manny Villar (34 percent, Nacionalista) in the Feb. 24-28 nationwide survey of the Social Weather Stations.
Because the SWS survey had a margin of error of plus/minus two percent, Villar followers added two percent to his 34 percent and concluded that the two candidates were again deadlocked (36-36) after Villar dropped by six percentage points in a previous survey.
We echoed the line that if two points are added to Villar’s score, two points must also be added to Aquino’s (not only to be fair but also to maintain the integrity of the mathematical comparison) and their two-point difference stays.
Readers emailing reactions were split 60-40 percent between those who scoffed at the idea of a “statistical tie” and those who insisted on it.
* * *
THE OVERLAP: As we gave more space to the contention that there is no “statistical tie,” let us now print an email saying that there is. Dr. Kenneth F. Smith <kenfsmith@aol.com> said:
“While I do not doubt your knowledge and understanding of Philippine political practices, and while I have always regarded the Social Weather Stations surveys as being of high quality, I also share your apprehension about the misuse and abuse of survey statistical results.
“However, I regret to inform you that in your article — when you state that “If we add two points to Villar’s score as we do in doing equations, we must also add two points to Aquino’s — not only to be fair but also to maintain the integrity of the mathematical comparison. We must allow both sides in the equation to move up or down together….” you misunderstand the concept of a ‘statistical tie’ and as a consequence have misinformed your readers.
“Essentially in assessing statistical significance, we do the complete opposite of what you stated; that is we not only add the margin of error to the lower score, but we also subtract the margin of error from the higher score. If the ranges of scores then overlap, there is ‘no significant difference’ and it is declared a ‘statistical tie.’ If they do not overlap, there is an acknowledged statistically significant difference.
“However, I also agree with you that the term ‘statistical tie’ is probably too technically abstract for most people, and the better way of reporting it for the layman is to describe the race as ‘tight’ or ‘too close to call’.”
* * *
CONDITIONING: Survey reports are not always validated by the election results. Still, they succeed in conditioning the minds of voters whose preferences are not deep-rooted.
We have observed that many Filipinos (1) regard elections as popularity contests, (2) gloss over the candidates’ qualifications and suitability to the position, and (3) hesitate to vote for anyone they think will lose anyway.
Surveys groups enjoying high credibility are able to play with these attitudes of voters. They are able to influence many of them into favoring candidates with high survey ratings and perceived to be likely winners.
This is the marketing value of surveys that some moneyed candidates want to buy.
* * *
HOW MUCH?: Some political operators have admitted to us having influenced a survey to favor a client.
One way, they say, is to commission a survey. The price reported is P3-P4 million. While the survey is dressed up as an objective professional job and the standings are not drastically disturbed, it puts some candidates in more favorable light.
Another way is to insert a rider question. An interested party may pay P300,000 to insert a question seeking to measure an item in the public mind. It may be coincidence, but the party paying for a rider ends up getting favorable treatment in the results.
* * *
TABLE SURVEY: The way it was described, the process — which we call “table survey” because it can be done without rising from one’s desk — is like the preparation of a fraudulent income tax returns.
The taxpayer starts by deciding how much he wants to pay the Bureau of Internal Revenue. Then, he (or his accountant or fixer) computes backwards the incomes to be declared, the exemptions, deductions and the rest of the items that will determine the taxable income left.
If survey field data say Candidate A has 32 percent and Candidate B has 26 percent (or a difference of six percentage points), it is still easy to assist B by allotting him maybe 29 percent and thus narrow the difference to just three points.
The scores in the geographic and economic tabulations are “corrected” slightly to produce the 29 percent for B. Candidate A the true leader is still on top, but B is close enough to A, putting him in contention in the minds of his financiers, followers and the public.
* * *
TWO EDITIONS: An expert using a computer can alter and rearrange the figures in one sitting. Nobody would notice the difference.
But if data are doctored, does not the survey become useless even to the client? The customer can ask for two reports — much like a businessman maintaining two books, one for the BIR and City Hall and another (the true one) for his private consumption.
(Note: This is a general discussion. We are not saying that Social Weather Stations and Pulse Asia, the two most respected survey firms in town, are engaged in a racket akin to that described above.
(They have experts whom you can call and consult if you have questions about their data, methods, et cetera.)
* * *
ePOSTSCRIPT: Read current and old POSTSCRIPTs at www.manilamail.com. E-mail feedback to fdp333@yahoo.com