Explaining Employment Trends in 1 South Africa: 1993-1998 Haroon Bhorat Introduction Probably the key economic challenge facing the democratic government is howto encourage an environment that sim ultaneously raises the prospects of employment for new entrants, while also creating jobs for the significant number of currently unem ployed individuals. The point of departure for this paper is to analyse employment trends in the economy since 1993. In particular, the paper will try to determine m ore system atically what the m agnitude of the employment shifts have been, and where they have been occurring over the last 4 or 5 years in the dom estic econom y. We will therefore also be indirectly be assessing recent claim s in the m edia an elsewhere of, for example, job losses totalling 500 000 since 1993. We are imm ediately hamstrung by data availability, but an attempt has been m ade to be both creative and thorough with the sub-optimal num bers we have at our disposal. T o this end, the paper will first consider the trends that have been observed from 1993 to 1998 in the form al sector, exam ining in particular the m ovements in employm ent across sectors and broad skills categories. In the second instance, we use the household survey data for the period 1993 to 1997 to consider total em ploym ent shifts across a wider range of covariates. Formal Em ploym ent Trends Since 1993: Is the Data Saying Anything New? In order to analyse form al employm ent trends, and indeed total em ploym ent shifts, the data sources suggest in the first instance that the Quarterly Bulletin of Statistics (QBS) would contain the primary source of employment information. The QBS labour m arket information has more recently been derived principally from the estim ates calculated in the Survey of Total Em ployment and Earnings(STEE). The first difficult y wit h t he QBS dat a though is t hat t he employment numbers do not include Agriculture. Hence, one is always left with a crucial segment of aggregate employment excluded. This is presumably a function of the instrument that is being used by Statistics South Africa(SSA) to collect employment and other labour m arket data. Secondly, and more fatally, the data in m any instances is not plausible and in fact renders the QBS a very poor source of employment inform ation. A simple exam ple will indicate the poor quality of this data 2 . Given that the QBSconstantly revises its em ploym ent num bers from one issue to the next, there has to be a process of always sifting out the final‘revised’ employment numbers produced for any given year. Hence, taking the QBSissues for December 1999 and March 2000, the estim ates for non-agricultural employm ent between 1997 and 1998 are produced below. 1 Director, Development Policy Research Unit, School ofEconomics,UCT. All comments can be directed to the authorat Bhorat@ hiddingh.uct.ac.za 2 Iamindebted to Dudley Horner for this example. 2