'No proof' survey data was faked: census chief Lily Ou-yang

Experts' investigation into government scandal reveals low inconsistency in polltakers' answers

PUBLISHED : Tuesday, 26 March, 2013, 12:00am
UPDATED : Tuesday, 26 March, 2013, 4:50am

A task force set up to probe allegations that census officers have been faking answers to speed up their work has concluded that there is "no sufficient evidence to show the existence of systematic fabrications".

But the chairwoman of the five-member group, Commissioner for Census and Statistics Lily Ou-yang, did not say clearly whether the statement meant that any census officer had ever fabricated answers.

She would say only that she could not rule out the possibility that some officers "have not followed the guidelines".

The task force was set up in January, following a series of reports that quoted current and former officers as saying that they had been fabricating data since as far back as the 1990s.

The officers said they were forced to do so because the regulations were too tough.

Announcing the results of the investigation yesterday, the commissioner said the task force had looked into three surveys at the centre of the controversy.

For the first, it commissioned a private research company to interview by phone 2,821 people from 900 households who had earlier been interviewed in the General Household Survey.

It then matched the new set of data to the previous set and found only a 2.5 per cent inconsistency in the results.

Task force member Vincent Kwan, also director and general manager of Hang Seng Indexes, said one reason for the inconsistency could be that the interviewees did not remember what they had said in the previous survey.

"For example, only one answer in a questionnaire consisting of several questions was inconsistent with the previous survey. The rest of the answers were all consistent," Kwan said.

The task force also asked its officers to look at the Labour Earnings Survey and Annual Earnings and Hours Survey, finding a 4.1 per cent inconsistency in the first and none in the second.

"The existence of inconsistencies does not necessarily mean that the data collected was erroneous … but we do not rule out the possibility that individual officers have not followed strictly our guidelines," Ou-yang said.

"For example, if [the officers] visit a household and see a person about 60 years of age, [the officers] may think that the person does not go to work any more … and skip some questions."

The task force gave a list of suggestions to ensure the quality of the data collected, including establishing a committee to safeguard the quality of the surveys.

The department will also conduct a comprehensive review of the existing fieldwork management system, covering such aspects as the time sheets and officers' workload.

University of Hong Kong statistics specialist Paul Yip Siu-fai said the inconsistency was too small to make an impact on the overall accuracy of the survey.