Voters 'more co-operative' with exit polls, says HKU's Robert Chung
Voters were more co-operative with exit pollsters in this month's Legislative Council election than they were in 2008, generating largely accurate results, a leading pollster said.
Dr Robert Chung Ting-yiu, director of the University of Hong Kong's public opinion programme, told a radio interviewer that more than 60 per cent of the voters approached by university pollsters on September 9 agreed to answer their questions - about a 10-percentage-point increase from four years ago.
"Voters were much more co-operative," Chung said. "But it still didn't reach the level of eight years ago, before the controversies about exit polls became heated. It was about 60 to 70 per cent [in 2004]."
Exit polls became controversial in the 2008 Legco election when some polling organisations were accused of letting political parties use their data to help their campaigns - spotting constituencies where their candidates were in danger of losing and needed extra support.
This year, the pan-democrats urged voters to take part only in exit polls carried out by HKU, after three out of five organisations allowed to conduct polls were found to be linked to the pro-establishment camp. They were the Hong Kong Research Association, Hong Kong Society Monitor and the Association of Community in Hong Kong.
Some voters claimed they lied to pollsters to confuse their results. But after preliminary analysis, Chung said he believed the results of the HKU polls this year were largely accurate.
All 20 of the candidates they initially deemed to have an "extremely high" prospect of winning did win, he noted. Seven out of nine determined to have a "relatively high" prospect of winning eventually won.
HKU pollsters made initial predictions at 11.30pm and in the small hours of September 10, after voting ended, based on exit poll questionnaires from about 10,000 respondents.
Chung said it would take further analysis to determine the polls' real reliability.
"Four years ago, we did exit polls outside 120 polling stations, but after a long filtering process, data from more than one-tenth, or 14 stations, had to be eliminated," he said. "We suspected some data from these stations were not trustworthy ... this year we are still analysing."