10 December 2009

Labour Force Survey 2009

A total of 33,000 housing units were selected for the initial sample for the comprehensive Labour Force Survey 2009.

Of these, 1,494 households were subsequently found to be unoccupied, non-residential or demolished and excluded from the survey.

How were non-residential or demolished housing units included in the initial sample?

Following the practice in previous years, the housing units would have been selected from the National Database on Dwellings (NDD), a register of all residential dwelling units in Singapore.  The NDD, which is maintained by the Department of Statistics, is updated monthly from administrative data to provide a comprehensive sampling frame for conducting household surveys.

The NDD should be able to differentiate a non-residential housing unit from a residential housing unit (unless the change of use was illegal) and should have up-to-date information on whether a housing unit exists or has been demolished.

Secondly, 28,487 households responded to the survey; 3,019 households didn't.

Altogether, 4,513 (1,494 + 3,019) housing units in the initial sample did not participate in the survey.  That's 13.7 per cent of the initial sample.

The housing units in the initial sample were chosen by systematic sampling in a stratified design.  Collectively, they were supposed to fairly represent the resident population, within certain sampling errors.

As 13.7 per cent of the housing units in the initial sample did not participate, this could have significantly distorted or otherwise affected the conclusions of the Labour Force Survey, over and above the usual sampling error.  For example, HDB households might have been over-represented if the non-residential or demolished housing units were less likely to be HDB flats.  Or perhaps, certain categories of households might have been less inclined to respond than others.

Unfortunately, this non-sampling error cannot be quantified.

Thirdly, the problem with the non-sampling error resulting from households being unwilling or failing to respond cannot be addressed by tighter supervision of the data collection process or other administrative procedures.

What can be done?

The Labour Force Survey was conducted under the Statistics Act (Chapter 317), which empowers the relevant authority to collect data on economic activities.

The Act requires every person to whom a requisition for particulars and information has been issued to furnish particulars and supply information, and provides for penalties for refusing to furnish particulars and supply information within the time allowed.

In spite of this, 3,019 households — almost one in ten of those surveyed — did not respond.  Perhaps, they did not understand that they had to respond.  Could more have been done to inform or advise them?