Monitoring public opinion - CHASING THE WIND By Felipe B. Miranda

In an age where democracy has gained much respectability, served at least by the lip if not by the heart of most of those who actually govern, one can anticipate an exuberant interest in discovering the political opinions, sentiments and attitudes of the people, a politically correct democracy’s demos. If the people, yes, the mostly ordinary people, somehow got to be enthroned as a polity’s "sovereign people", then an industry can develop which regularly probes the hearts and minds of those who provide democracies and quasi-democracies their requisite governance imprimaturs.

Like mostly any other thing in life, the monitoring of public opinion can be done responsibly or irresponsibly. In the immediately previous article of this columnist, the irresponsible conduct of public opinion assessment was discussed and the present article will now focus on the more pleasant topic of responsibly looking into public opinion.

Once a concern or issue has been identified on which a public opinion probe is contemplated, the first challenge is clarifying as much as possible what one understands by it and how to make the actual respondents to a survey understand it similarly. If one is probing into national security threats, for instance, it is important to conceptually clarify what one understands by the idea of national security. One might be more interested in its comprehensive sense such that not only armed threats but also a people’s poverty status and their political marginalization, among other things, become an integral part of the idea. Or alternatively, one could narrowly consider national security to exclusively focus on some specific group whose challenge to a regime’s political legitimacy is sustained through armed rebellion. Depending on how one finally formulates the survey concern, certain implications ensue which are unavoidable for designing an appropriate sample and the requisite questionnaire.

The next challenge has to do with establishing the relevant population for any given survey interest and the appropriate sampling design for determining how survey respondents from this population might be chosen. If, for instance, one’s focus has to do with the personal or direct experience of child-bearing, it would be illogical to consider a population primarily of males from which the sample respondents would be drawn. It would be equally strange if women who have or had children were excluded from being respondents. However, women who have never had children might be included in the sample if the surveyors’ concern for child-bearing included a projective interest (i.e. how this kind of women anticipated what or how the actual experience of child-bearing might be.)

Sampling design which concerns itself with such problems as the legitimate survey population, sampling frame and sample determination, has one fundamental concern. Where it is impractical to study everyone in the entire population and one must study it through a much smaller number of sample respondents, the question of sample representativeness arises and is primarily addressed by employing some randomizing mechanism which diminishes the biased selection of survey respondents. Randomly drawn probability samples (i.e. samples resulting from following a selection process where each unit of the population has a known – not necessarily equal – probability of being chosen as part of the sample) are the only ones for which a survey’s confidence levels and margins of errors might be legitimately computed.

The issue of sample size is for most people a counterintuitive matter. Many people find it hard to understand that 1200 or 1500 or 2000 respondents can effectively reflect the preferences or characteristics of 100 million or 700 million or more people. Most find it difficult to accept that 300 or 400 respondents may be sufficient to capture the sense of 8 million Metro Manilans. Still, given randomly drawn, representative samples of fairly small sizes, general demographic surveys and the more specialized electoral polls across the years have largely proven reliable in correctly reading the properties or directions of relatively large, even huge populations.

Sample size is actually determined by two major considerations. One has to do with the financial and material resources available to conduct and finish a study of any given population. Money and time limitations set constraints on the number of people who might be interviewed or included in any survey sample. The other consideration has to do with what an expert or informed judgment might hold to be the level of precision necessary for any specific study.

In an electoral contest between two particularly well-matched candidates, a sample of 10,000 eligible voters could be mandatory should there be insistence on a survey where the margin of error in reading possible electoral outcomes is at most plus or minus 1 percent. On the other hand, if a deluded UP full professor somehow figured in an electoral mismatch with a nationally prominent film actor, the sample could be as small as 100 respondents and the resulting survey data could be read with a generous error margin of as much as 10 percentage points either way.

Both surveys would do their respective jobs quite well of predicting which candidate would probably emerged the winner or loser. However, it would be a criminal waste of resources if the 10,000 sample were used to read the electoral probabilities for the nutty UP full professor and the nationally prominent actor. On the other hand, it would be false economy for someone to insist on using a sample of 100 respondents in calling the very close contest between the two hypothetically well-matched electoral candidates.

Questionnaire design is crucial in running a good survey probe. The main challenges here are to avoid loading biases into the questions used and to maximize the clarity of what is really being asked. For instance, when asking people about their opinions regarding possible American participation in Philippine politics, it makes a lot of difference whether the term used in the controlling Pilipino question is "pakikialam" (intervention) or "pakikilahok" (involvement or participation). The first automatically triggers primarily negative reactions, the second allows for a wider range of reactions, from that which may be fully positive to that which is completely negative.

Operational definition of many survey concerns helps clarify the question. In the probe considered above, the respondents’ answers would be less ambiguous if American involvement or "pakikilahok" were explicitly clarified as referring to American participation in joint Filipino-American military exercises of a given magnitude, perhaps even in some explicitly identified Philippine area, for some specific period of time.

In the case of multilingual societies like the Philippines, questionnaire design also automatically relates to the effective and reliable translation of the original or linguistically controlling questionnaire into versions which would be respondent-friendly, given their particular ethnic or regional group. Cebuanos normally prefer to be interviewed in Cebuano, Tagalogs in Tagalog, Ilonggos in Ilonggo and Pampangos in Kapampangan. (Incidentally, Pilipino – as distinguished from Tagalog – is increasingly popular among the citizenry and may yet become a truly national lingua franca.)

The quality of the actual interviews impact much on the ultimate value of any given survey or public opinion poll. Given the current status of public access to telephones, national surveys have to rely mostly on personal or face to face interviews. Field interviewers have to be provided good training and must have competent and conscientious field supervisors and managers to undertake their fieldwork well. A survey can only be just as good (or just as as bad) as its fieldwork and its fieldworkers. The need to maintain fieldwork quality at the highest possible levels is often reflected in the normally higher costing of the more responsibly done public opinion surveys.

Data encoding, processing and analysis are then brought to bear on the data generated by the field interviews. These tasks call for the most competent and dedicated people to collaborate in ensuring that survey data are transformed into information packages that allow for faster and easier communication with target audiences, whether they be the general public, media or some group with rather specific, perhaps even partisan, needs. Tables, charts and analytical reports issue from regular people who may have basic production skills and those who, having substance expertise, undertake the preparation of analytical reports and, ultimately, their presentation to a public opinion survey’s varied audiences.

Up to this point, this columnist has offered an overview of the concerns and processes underpinning the responsible monitoring of public opinion in a democratic or democratizing society. A final dimension of this challenge needs to be addressed in his next column. One must look into the ultimate values which must be demanded of anyone who presumes to undertake the monitoring of the public’s opinions, sentiments and attitudes. Lacking a strong commitment to these values, survey researchers and pollsters can easily transform into, at best, run-of the mill business people and, at worst, predatory hustlers and quick-draw hired guns.

Adrian Cristobal, in one of his invariably elegant newspaper columns about a decade ago, speculated on whether some people who monitored public opinion in this country might be more precisely addressed as "pollstitutes." He might have been wrong a decade ago, but his neologism may gain currency given current times.