Surveys Useful But Can Be Misleading Say Pollsters

[Lite Beer]

Surveys useful but can be misleading: pollsters

By Kornchanok Raksaseri

The Nation on Sunday

The results of opinion polls could be misleading while also indicating trends and keeping people aware of the upcoming election, academics say.

In a seminar hosted by The Nation's sister publication Krungthep Turakij, academics and conductors of opinion polls discussed the advance of poll surveys in Thailand with a more considered process. But they said people should not make political decisions based only on survey results.

Assumption University's Abac Poll director Noppadon Kannika said readers should pay attention to the "margin of error" indicated by pollsters conducting the surveys.

"If the survey shows two political parties having close popularity, it is difficult to say which party will actually get more votes," he said.

Noppadon gave an example that from a survey party A might beat party B by 40 to 38 per cent. But with a 3per cent error margin - a statistic that indicates the possibility of a varied result, meant party A might actually have a popularity rating from 37 to 43 per cent, while party B might have a popularity rating from 35 to 41 per cent. There was also the chance that party B might beat party A, if it was actually more popular.

Pichai Ranatilaka Na Bhuket, deputy dean at the National Institute of Development Administration (NIDA)'s School of Social and Environmental Development, said readers should consider the sampling of each poll. if the sampling was not proportionate with the actual population, the poll result was likely to be unreliable.

"Polls will not change the decision of those who have made up their minds," he said. Poll results might only influence people that could not make up their minds very close to the voting time.

Suan Dusit Poll's Sukhum Chaleysub said: "Thais follow trends. They think if many people would vote for a particular candidate, they assume the candidate is a good choice."

Noppadon said poll results could affect people's decisions. There are people who like to back the winning side, while others would vote for the underdog.

Pachitchanat Siripanich, director of NIDA, said she was glad more and more advanced surveys were being conducted in Thailand, as the polls and publication of the results helped keep people aware of the July 3 election. Polls should give some information to people although people's opinions changed all the time.

Kiatanantha Lounkaew, director of Dhurakij Pundit Poll Centre, said opinion polls emphasised the importance of information and were an important part in Thailand's democracy development.

He disagreed with the ban of publicising poll results seven days before the election day.

Sukhum felt the same: "The closer the election, I think the poll results should be publicised and updated daily."

He also noted loopholes in survey results that might hide the real thinking of people being polled.

"Was it really that the respondents haven't made up their mind or that they didn't want to tell us [how they would vote]? We have to put 'decided but preferred not to answer' in the choices as well as 'undecided' to get the most accurate result," he said.

"The poll results may mistakenly make political parties and politicians optimistic by saying a large proportion of voters have not decided when actually they had low popularity," Sukhum said He said more deliberate poll surveys would be a better way to judge the election result.

"When respondents answered whom they would pick as their choices, we have to rate the answer considering other answers from the same respondent. For example, if a respondent picked Candidate A, but said he or she "might" go to vote, we will rate the point for the candidate as 0.5 instead of 1," Sukhum said.

Noppadon said his centre also weighted its calculating and invested more in getting a large sample of respondents and trying to get the most accurate result.

He said the public should be aware of three kinds of polls: "slave polls" or "servant polls" conducted by or for politicians, unreliable polls conducted without scientific methods and mobilised polls conducted by people pretending to survey but actually trying to frame political opponents.

-- The Nation 2011-06-05

[jaltsc]

It doesn't matter what the poll results are. The statistical results can be manipulated to make the results appear favorable to either side.

One of the standard texts on statistics, "How To Lie With Statistics" was written by Darrell Huff in 1954. "There is terror in numbers, and nowhere does this terror translate to blind acceptance of authority more than in the slippery world of averages, correlations, graphs, and trends". "The secret language of statistics, so appealing in a fact-minded culture, is employed to sensationalize, inflate, confuse, and oversimplify." Statistics are rife with opportunities for misuse, from "gee-whiz graphs" that add nonexistent drama to trends, to "results" detached from their method and meaning, to statistics' ultimate bugaboo--faulty cause-and-effect reasoning.

DOES ANY OF THIS SOUND FAMILIAR? Yes, it might take the Thai electorate some time to approach 1954 reasoning.