(Reprinted from HKCER Letters, Vol.33, July 1995)
Response by the Census & Statistics Department
The note by Dr Suen brings out relevant issues relating to the study of the labor market situation in Hong Kong. There are, however, points which require clarification so as to give readers a more complete picture. These points are related to official statistics on unemployment compiled by the Census and Statistics Department through the General Household Survey (GHS).
Dr Suen suggests that researchers be allowed to have direct access to individual records collected in the GHS to facilitate the conduct of analytical research, thus making better use of the data collected. While this suggestion is made with good intentions, there are, however, other factors which need to be considered.
In conducting surveys, official statistical agencies, including this department, are obliged by statutory requirements to keep records pertaining to individuals in strict confidence. These agencies pledge that individual records will not be divulged to other parties not involved in the surveys concerned.
It has been argued that confidentiality is maintained if record identifiers, such as name, geographical location, etc, are purged. Despite the purge of record identifiers and use of carefully designed techniques, such as masking and random perturbation of microdata, the fact remains that there is always a risk of unauthorised data disclosure if microdata files are released publicly.
The disclosure of even a single individual record from an anonymized microdata file will have an adverse consequence on the trust of data providers towards an official statistical agency. Notably, this will diminish their level of cooperation in surveys conducted by these agencies. They may either refuse to respond, particularly when a survey like the GHS is conducted on a voluntary basis, or they may provide less than satisfactory responses to the questions posed, thus putting in jeopardy the reliability of the data collected.
Like many other official statistical agencies in the world, this department has to be prudent in dealing with this matter. It is not only necessary to actually exercise extreme care in maintaining the trust of data providers in respect to data confidentiality but also important to be seen to act as such. The release of anonymized microdata files, no matter how carefully designed against unauthorized data disclosure, may give data providers a wrong perception about the official statistical agency's pledge to keep individual data confidential.
In order to maintain the trust of data providers and to uphold statutory requirements, it is the practice of this department not to release GHS microdata files, even anonymized, to researchers. This, however, does not in any way restrain researchers from conducting analytical research. On the contrary, the department is prepared to produce additional detailed tabulations for them in order to facilitate the conducting of research, provided certain requirements on data confidentiality and data reliability are met, or to run statistical programs for them. If deemed necessary, on-site access to microdata files for analytical studies may also be considered. This mode of operation, albeit not totally satisfactory to researchers, is considered appropriate for guaranteeing data confidentiality on the one hand and for meeting the need to have access to microdata for the conducting of research on the other.
Dr Suen points out that in the regular press releases and in the Hong Kong Monthly Digest of Statistics ( HKMDS), only the seasonally adjusted (but not the unadjusted) unemployment rate is reported. He therefore submits that "public opinion and policy makers have focused almost exclusively on the seasonally adjusted unemployment rate."
It should be noted that seasonal adjustment of the unemployment statistics compiled by this department refers only to adjustment for seasonal variations in the proportion of first-time job seekers (FTJS) in the labor force. The meaning of adjustment of the unemployment rate is explicitly given in the footnote to Table 2.1 of the HKMDS and in the Quarterly Report on General Household Survey. In fact, Dr Suen is quite aware of why and how the unemployment rate from the GHS is adjusted. In a regular press release for the layman public, however, it does not seem appropriate to include such technical details.
In adjusting the unemployment rate as mentioned above, policy makers and other users will have an unemployment rate series which is more comparable over time than an unadjusted series for understanding changes in the labor market situation. It is believed that for the use of policy makers, the direction and the magnitude of changes in the unemployment rate are of particular significance. Anyhow, quarterly unadjusted unemployment rates are available in the Quarterly GHS Report.
It is agreed that, as Dr Suen rightly pointed out, the more common seasonal adjustment methods would preserve the mean of the original series, and if the original series contains peaks and troughs that occur regularly in particular seasons, the adjusted series would remove such patterns. However, as explained above, seasonal adjustment of the unemployment rate currently made is just intended to remove seasonal variations in the proportion of FTJS in the labor force, at least at the time when it was first introduced.
In fact, alternative methods of using seasonal adjustment packages like X11-ARIMA or X11 (which incidentally is used to adjust other time series data such as the monthly index of retail sales as published in the Report on Monthly Survey of Retail Sales) have been studied by the department. However, the seasonality pattern of the unemployment rate derived using such methods appears to be somewhat unstable. Moreover, results of the unemployment rate seasonally adjusted using X11-ARIMA or X11 are difficult to explain to the layman public. Given such considerations, there is hesitation in changing the method of adjustment in preference to maintaining the present method.
It is considered that the existing method, with its imperfections, is still a valid method. Nevertheless, the entire subject of seasonal adjustment of unemployment data has been studied by the department, as the method has been used for a long time (ever since the series was established). Dr Suen's views are therefore a timely contribution to its research.