The day the opinion polls went awry

Ken Lui - Editorial Board Member, The Journalist

[中文][Oct 2012 - The Journalist] The 2012 Legislative Council Election was a battle royal, whether in geographic constituencies or the 'Super Constituency'. To give viewers a better focus Cable TV and now TV joined hands to commission Hong Kong University's Public Opinion Programme for a daily rolling poll to measure the ratings of candidates in different constituencies.

The findings of opinion poll conducted by an academic organisation commissioned by TV news outlets could not predict the election results. Some candidates lost although the poll suggested that they could be elected. (Ming Pao Photo)
Most candidates claimed “I will probably lose”. A majority of voters said they had not made up their minds. As a result some candidates' daily rating kept going up and down, like a roller coaster. The Journalist compared the results between the opinion polls and the final results, and found significant differences in four constituencies.

The rating of candidates is important information for voters. It may affect the outcome at the last minute. Unfortunately, opinion polling has become less and less accurate. According to such polls in Hong Kong Island constituency, both People Power's Christopher Lau and DAB's Chung Shu-kun each had 5 percent of support, placing sixth and seventh respectively. Going by this poll they should have won the last two seats.

But Christopher Lau lost while Wong Kwok-hing from FTU won. Opinion polls had earlier showed Wong Kwok-hing getting merely 4% of support, not enough to send him to Legco.

In Kowloon East constituency, opinion polls showed People Power's Wong Yeung-tat garnering 7 percent support, safely placing in fifth place and a coveted seat. But it was Paul Tse with 6 percent support who got the last seat.

New Territories West results were equally surprising. According to the polls, Lee Wing-tat of the Democratic Party's scoring 5 percent was in a safe fifth place.

The DAB's Chan Han-pan and Leung Che-cheung, each with merely 3 percent support, should be left out in the cold.

Ironically, the hot pick Lee Wing-tat lost; the two underdogs Chan Han-pan and Leung Che-cheung romped home.

In New Territories East, Scarlett Pong of Civic Force got 5 percent of support. This put her in fifth position in the line-up of 19 candidates.

Christine Fong, Fernando Cheung of Labour Party and Ip Wai-ming of FTU each had 3% of support, so their chances were equal.

The result? Scarlett Pong, Christine Fong and Ip Wai-ming were out. Fernando Cheung got fourth highest vote in that constituency. Gary Fan of Neo-Democrats who got only 2% of support according to the poll (14th place in the list of 19), won the last seat.

Only the election result of Kowloon West matched the results of opinion poll.

Opinion polls showed People Power's Wong Yeung-tat could get the fifth seat in Kowloon East Constituency but he lost. Democratic Party's Wu Chi-wai and Civic Party's Alan Leong of democratic camp were elected. (Ming Pao Photo)
According to one legislator, opinion polls can affect voters' thinking but they have also become less reliable. He said researchers often called voters using landlines between 7pm and 9pm. These are not times when professionals are usually at home. These people generally do not have time to talk in the office, so opinion polls; do not reflect actual support, especially among professionals which are his supporters.

Another point need to notice is that the sampling size was small in the opening phases. For example, on 6 August, only 312 and 289 respondents in New Territories East and New Territories West respectively replied.

Moreover, the maximum sampling error for all the samplings was 5% (when the poll started) to 3% (election day). In the constituencies with a dozen candidate lists, such as New Territories East, the ratings of fifth place list to thirteen place list were about 2% to 4%. Taking sampling error into account, the value of such opinion polls would not be high.

Finally, opinion poll showed around twenty to thirty percent of voters who had not made up their minds. Cable TV took this into account but now TV did not count the undecided voters before ranking the candidate lists. Both stations claimed the results were not affected.

Assignment editor of Cable TV, Lam Miu-yan explained that if they did not count the undecided voters, the sampling size would further shrink which would amplify the sampling error. She also believed that it could reflect the voters' opinion more faithfully.

now TV deputy assignment editor Billy Ko said if candidate A had 7% of support, candidate B had 3% of support and the sampling error was 4%, the difference was not significant. They also interviewed expert to explain why opinion poll was not trustworthy.

He said when they reported the poll results, they did not just look at the data but they also emphasised the trend, included a weekly summary, considered the latest situation and made references to the previous election results.


沒有留言: