‘Answers are part of the question’
Understanding an opinion poll starts with understanding its questions.
Sometimes, the poll has questions that are open-ended, which means answerable in the respondent’s own words, with no guidance from the interviewer. These words, whether few or many, are meant to be recorded verbatim. This type of question is easy to frame, but its answers are not easy to summarize.
Most of the time, however, the poll’s questions are of the multiple-choice type, which means that the allowable answers have been preframed by the questionnaire designer. Such questions cum answers need more time and care to design, but then they are relatively easy to implement in the field. The answer-choices of the respondents are likewise easy to summarize.
Even when a survey question seems very simple—seemingly answerable by either “Yes” or “No”, for example—the survey user should not take the choice of answers for granted, but see what the questionnaire actually directs the interviewer to prompt. Suppose the choices are “Yes”, “No”, “Can’t Choose”, and “I don’t know anything about that issue”? The choices make a big difference.
In the case of a question asking for vote-preference for only one candidate out of a list, it is crucial for a survey user to know the entire list, rather than just the names of those who turn out to be the leaders.
In case the name Fidel V. Ramos is in the list, for example, it is bound to get a few mentions, and will thus affect the point-spreads between the more plausible candidates. In case “None of the above” is in the list, as told to, or as read by, the respondents, it will also make a difference, and will also affect the point-spreads.
In a single survey, the respondents may be asked separate voting questions, each with a separate or distinct list. This type of survey will obtain different results for the distinct lists. It may produce differences not merely in the spreads, but even in the rankings, of the candidates common to the distinct lists.
Smart politicians are the ones who commission, privately, such multilist surveys, before the filing date of candidacies. They do it precisely to learn how the lineups will affect the odds in the election.
Even after the filing deadline, such information still matters, since it shows the consequences of possible pullouts of any candidates, i.e., “spoilers,” before Election Day. In 2004, for example, our surveys showed that Raul Roco’s voters had Gloria Arroyo as their main second choice; thus, the revelation of Roco’s illness, late in the campaign, was a critical boost to Arroyo.
Multilist surveys are originally meant for confidential private use, of course, rather than for general public information. When done by SWS, such surveys are, by contract, embargoed for three years, and then become open for public research in the SWS Survey Data Library.
In this way, the raw data of all SWS political surveys before (as well as after) all the elections from 1987 to 2010 may be studied by anyone. There is no need for consent of the original private commissioners.
If the lineup of candidates can matter very much within a single sample of respondents, all the more can it matter across separate samples, even if interviewed in the same period of time. Surveys with different lineups are simply not comparable to each other. The lineups of candidates for president, vice president and senators in the BW-SWS survey of September 2015 are as directed by the sponsor, BusinessWorld. They are for one-time use since the official list of candidates will soon be known.
Politicians know very well how sensitive can be the picture of an election race to the exact listing of the potential candidates. So they are not, in fact, likely to be confused by variation in survey appearances.
Regardless of how they criticize surveys in public statements to the media or to their support groups, professional politicians know how survey numbers operate. They will keep on commissioning their own private surveys, and will stay posted on public surveys at the same time. They are too intelligent to risk losing an election by fooling themselves and basing their campaigns on wishful thinking.
The best way to learn how a specific candidate is progressing over time is to examine successive surveys that use the same list. Usually, a candidate is unsure of who his/her opponents will be, but anticipates alternative scenarios like List A, List B, List C, etc. Then the candidate has to sponsor successive surveys that each use Lists A, B, and C, etc.
To discover not just HOW but also WHY a candidate’s survey numbers are changing over time, it is not enough to be well-versed in the sequence of political events and the statements and actions of all the pertinent players. In the first place, how can one tell if such matters are as critical as political pundits think they are? For instance, how many voters listened to or read the President’s last State of the Nation Address, and what did they think of it? How many listened to or read Mr. Binay’s so-called True Sona, and what did they think of it?
Ideally, the preelection surveys should also include probes into the voters’ awareness of, and reactions to, the said events, statements and actions. This will enable analysts to do statistical correlations of the said factors to the voters’ preferences among the potential candidates. Are the said factors as important as the key personal attributes of the candidates, and their priorities or platforms for governance, as perceived by the voters? What, in fact, are the voters’ perceptions?
So many basic bits of data, needed to analyze an election, are not provided by the official count. The Commission on Elections knows only the location of the votes. It knows nothing about the age, gender, education, religion, social class, party affiliation, etc. of the voters. The latter data can be supplied only by scientific opinion polls.
* * *
Contact [email protected].
