I refer to Tim Hamlett's Sunday Morning Post column of May 16 headlined, 'Number-crunching leaves credibility crumpled'.
Mr Hamlett was under the impression that the 'original figure for the number of mainland illegitimate children, discovered by the old-fashioned methods which the Government happily accepts for other purposes', that is, a figure of about 20,000, was discarded due to political expediency. This accusation is totally unfounded.
When the survey was being planned, difficulties were already expected for asking questions about 'children born out of wedlock'. It was therefore decided, right at the beginning of the survey, that only a half sample of respondents would be asked the question using the conventional 'Direct Questioning Method' (DQM), while in the other half sample the 'Randomised Response Technique' (RRT) would be applied.
The actual outcome from the interviews has been that in using the DQM, the enumerators noticed that respondents often felt embarrassed or uneasy, or showed a perfunctory attitude when directly asked to respond to the sensitive question.
This came as no surprise as it was well known that sensitive topics would meet with a lot of underreporting in surveys.
Hence, although the 'old-fashioned method' (of direct questioning) would work for many other topics, it might not work for such a sensitive topic and our experiences in the field on this occasion fully indicated this.
Therefore, the decision to employ the RRT right from the start of the survey was a correct one.
In using the RRT, nobody except the respondent knows which question is being answered and hence the answer provided by the respondent does not carry any meaning to other persons, including the interviewer.
Under such circumstances, the respondent will feel comfortable and unpressured in providing the true information. The RRT fully guarantees the anonymity of the respondent and removes very effectively psychological difficulties so that he/she will be prepared to answer the survey questions truthfully.
Finally, let me clarify that the results now available are based on responses from the independent, random sample for March and the first half of April. This sample is adequate for deriving an independent estimate which is representative of the entire population.
ALVIN W. K. LI for Commissioner for Census and Statistics Tim Hamlett replies: I did not say that any figure was discarded for political reasons. What I said was that where two different methods produce widely different figures it is the honest policy to report both.