On Election Polls
by Dr. Romulo A. Virola ¹
Secretary General, NSCB

As the 2004 election fever rises, election pollsters proliferate and public interest in election polls intensifies. Candidates who rank high on a particular poll, understandably, consecrate its legitimacy; on the other hand “losing” candidates are quick to cast aspersion on the reliability and validity of the survey results, even if deep in their lonely hearts, they know that the survey only tells the truth, no matter how terrible. But what to do, if different surveys show conflicting results?

Read PART II - PART III - Part IV

Thanks to the science of statistics, well-designed surveys speak gospel truth. Yes, even if many people cannot understand it, less than two thousand respondents can speak the same voice as the 84.2 million Filipinos living in the 41, 969 barangays of our country! Haven’t you heard about the taste test? Don’t you know that you need not finish off a bowlful of dinuguan from a neighborhood carinderia to check if it is as yummy as the stew from mother’s kitchen? Science of sample surveys! Pure and simple!

But while some surveys have been properly designed, there are those that malign the statistical profession and desecrate statistics to the realm of lies and damned lies. The fourth of the ten Fundamental Principles of Official Statistics states that the statistical agencies are entitled to comment on erroneous interpretation and misuse of statistics. In this light, let me now share some views towards a more critical discernment by the public, including the politicians, of what polls truly say.

1. Election polls, when conducted systematically and objectively, using sound statistical techniques, are indeed, powerful tools for measuring the pulse of the voters. The statistical theory behind these polls guarantees their usefulness. In fact, past successes of legitimate pollsters have been impressive, especially to those who have little understanding of the science of polling. On the local front, the Social Weather Station, the Pulse Asia and the Krystal Research Association, Inc. have had convincingly successful track records in predicting election results. Practical and thinking politicians have been known to use the results of sound surveys to guide them in making election-related decisions. The foolish ones, of course choose to ignore the signs and go wayward, or rich, sometimes.

2. However, election polls and pollsters have also failed the public many times in the past, basically, for two major reasons: unsound methodology and/or deliberate intent to release to the public fabricated or pre-determined outcomes.

   Towards rationalizing the conduct of surveys by or for government agencies, the NSCB has put in place the Statistical Survey Review and Clearance System, a mechanism under which proposed surveys are evaluated on technical matters. In the review, the NSCB is assisted by a Technical Committee on Survey Design whose chair is Dr. Isidoro P. David, retired Chief Statistician of the Asian Development Bank. Surveys that pass the review are given NSCB clearance numbers that appear on the top right corner of the questionnaire. However, surveys conducted by the private sector do not fall under this system.

   A well-documented example of a failure of polls to predict the results of an election primarily due to unsound methodology is the Literary Digest faux pas in the 1936 US presidential election. The Literary Digest had sent out 10,000,000 ballots to its subscribers, most of whom were telephone and automobile owners. On the basis of 2.3 million returned ballots, it predicted that Alfred Landon would win over Franklin D. Roosevelt. In the four previous elections, the Literary Digest had predicted accurately the winner using the same approach; but in the post-Depression 1936, there was a high correlation between income and party preference, which was not there in the previous elections. Gallup, on the other hand, correctly predicted FDR as the winner and went on to continue a successful string of accurately predicting the winners of the US presidential elections except in 1948.

   The other major reason , of course, needs no elaboration as the failure had been known and expected from the very start especially by those who conducted the polls – the illegitimate pollsters.

   It must be added, however, that the use of sound methodologies by the best statisticians does not guarantee 100 % success. Random errors, which are a necessary component of statistical surveys in contrast to complete enumeration of respondents or censuses, will contribute a certain frequency of failure, the extent of which can be scientifically estimated.

3. Thus, it is desirable that the public be protected from unsound election polls and from illegitimate pollsters. While the trained eye may be able to distinguish between valid and invalid election survey results, the general public will benefit from some guidance on how to deal with election polls. Toward this end, we support the call ² of the eminent UP Economics Professor Solita Collas Monsod, Mareng Winnie to many, for the Philippine Statistical Association ( PSA) to rise to the challenge. As the country’s one and only association of statistical professionals, the PSA is in the best position to provide expert advice on how the public should manage the poll results released by various sources. This subject has been discussed before in the PSA, particularly an accreditation process wherein the PSA will maintain a list of accredited pollsters. Unfortunately, a decision was made not to push the idea. In this regard, it may interest the readers to know that the Makati Business Club Alert came out with 7 Rules for Reading Surveys which should be a useful start in enlightening the public about election polls. These rules have been reproduced by at least two columnists in nationally-circulated newspapers. The Philippine Daily Inquirer (February 6 issue, Tyranny of surveys) has raised similar concerns and has called on the Comelec to seek the assistance of the National Research Council of the Philippines and come up with a system of accreditation “to impose sanity and order in the national mania for polls”. But we believe that it is the Philippine Statistical Association, whose current president is Dr. Jose Ramon Albert, which should take a more responsive role in protecting the public from these polls, which, with the campaign period just around the corner, have truly become weapons of mass distraction.

Read PART II - PART III - Part IV

Reactions and views are welcome thru email to the author at ra.virola@nscb.gov.ph.

_______________
¹ Secretary General of the National Statistical Coordination Board (NSCB) and Chairman of the Statistical Research and Training Center (SRTC). He holds a Ph. D. in Statistics from the University of Michigan in Ann Arbor, USA and has taught mathematics and statistics at the University of the Philippines. He is also a past president of the Philippine Statistical Association.

² In her keynote speech during the Annual Conference of the Philippine Statistical Association (PSA) held on 12 November 2003 at the Sulo Hotel, Quezon City.

Posted 09 February 2004.