WP09/2009

The comparability of Labour Force Survey (LFS) and Quarterly Labour Force Survey (QLFS)

Stellenbosch Working Paper Series No. WP09/2009

Author(s): Derek Yu

JEL Classification: J00

Keywords: South Africa, Household survey

Download:  PDF (369 KB)


Abstract:

Statistics South Africa (Stats SA) has been collecting labour market data since 1993 with the October Household Survey (OHS), which was conducted annually between 1993 and 1999, as well as the Labour Force Survey (LFS), which was a biannual survey introduced in 2000 to replace the OHS. In March 2005, consultants from the International Monetary Fund (IMF) were appointed to revise all aspects of the LFS. All documents, processes and procedures relating to the LFS were reviewed, before a report on the findings was presented to Stats SA in June 2005. At the end, it was decided to re-engineer the LFS, and this took place in October 2005. Moreover, consultants were appointed in 2006 to help improve the survey questionnaire, sampling and weighting, data capture and processing systems. Eventually, Stats SA came up with a decision that the LFS would take place on a quarterly basis from 2008, i.e., the Quarterly Labour Force Survey (QLFS) was introduced to replace the LFS. The comparability issues between the OHSs and LFSs have been discussed thoroughly by Burger and Yu (2006), Casale, Muller and Posel (2005), Wittenberg (2004) and Yu (2007), focusing on changes in the sampling frame, inconsistencies in the questionnaire design, changes in the methodology to derive labour market status, trends in numerous variables (e.g., demographics, educational attainment, labour force participation rates, unemployment rates, earnings, etc.), oversampling of informal sector workers in 2000, overestimation of the earnings of self-employed in the OHSs, and the continuous improvement of the questionnaire by Stats SA. Therefore, this paper rather focuses on the comparability between LFS and QLFS, so as to assist researchers and policy makers when they try to analyze or compare both the LFS and QLFS data. As only four QLFSs have taken place at the time of writing, trends in variables will not be the focus of this paper. Instead, this paper will mainly look at the changes in questionnaire design, sampling method, derivation of new variables (i.e., underemployment status and unemployment status), a new methodology to capture the formal/informal status of the employed, as well as the drastic changes in methodology to capture labour market status. With regard to the latter, it is found that there is no longer a clear distinction between strict and broad labour market status in the QLFS, and this makes it difficult to derive long-term trends in the labour force participation rates (LFPRs) and unemployment rates under both strict and broad definitions.