The small and medium enterprise (SME) sector in South Africa has been the focus of attention since the first democratic elections in 1994. Not only does the sector offer the opportunity to enhance entrepreneurship amongst previously disadvantaged communities in South Africa, but it is also seen as one that has the ability to absorb relatively more labour per unit of output than large scale enterprises. One possible reason for the relatively higher labour absorption of the SME sector is that they pay relatively lower wages per worker.
In order to investigate whether this is indeed the case and whether this has resulted in relatively better performance by the SME sector in the manufacturing industry, we present data that offer a breakdown of key economic variables (value added, employment, wage bill, etc.) in the manufacturing industry by four size groups of enterprises: small (employing 1-19 workers); medium (employing 20-49 workers); large (employing 50-199 workers); and very large (employing more than 200 workers).
The data are presented for four points in time, spread over the period 1971-1996. Although the results are not as accurate as they might have been if we had time series of annual data, the analysis of the changes over the discrete time intervals gives us some idea of the economic performance of the different size groups of firms.
We start with a discussion of the data set in Section 1. Section 2 gives the broad descriptive picture of the role of SMEs in the industrial structure of South Africa and its changes over time. Finally, in Section 3 we turn to an analysis of the wage-employment trends in the different size groups of firms, based on the decomposition model developed by Mazumdar (2000), which has been used in an earlier paper for the South African manufacturing sector as a whole (Mazumdar and van Seventer 2002).
Section 1: Data
While a previous analysis of real wage decomposition for South Africa made use of an extensive industry database consisting of 30-year trends on an annual basis covering about 46 industries in the South African economy, this database is not endowed by a size class distinction. For our purposes here, we have to settle for less perfect data, recently made available in an unpublished format by Ntsika (1999). Although Ntsika has tried to cover all sectors in an attempt to bring size class differences in the South African economy to the surface, we limit our analysis to the manufacturing industry. The data shown in Table 1 are, according to Ntsika (1999), drawn from various issues of the Stats South Africa Statistical Yearbook. This cannot be correct as the last Stats South Africa Yearbook was published in 1995, while the more recent South African Statistics 2000 publication - which resembles the Statistical Yearbooks very closely - does not offer size class information. More likely, the information shown in the next table is drawn directly from the manufacturing census publications for the relevant years, which suggests that several other manufacturing census, such as the one for 1985, were not employed.
It should also be noted that the data shown in Table 1 are reported in 1995 constant prices, while the original manufacturing census is only reported in current prices. This means that an implicit deflator must have been employed; which one, however, is not clear. The other issue to note is that, probably as a result of employing a sub-industry specific deflator, the data set is no longer consistent. This can be attested in the last five rows of the table, where we sum the individual entries of each sub-industry for each size class and subtract the manufacturing totals shown at the top of the table. In the last row, it can be seen that even for the sum of all size classes, the sub-industries do not sum to total manufacturing.
Since we do not know what deflator Ntsika has employed, we use the TIPS South African Standardised Industry Database (see www.tips.org.za) to construct a deflator for the relevant years and relevant sub-industries in order to arrive at current values. Since it is unlikely that our deflator is the same as the one used by Ntsika, value added and wages and salaries at current prices also turned out to be inconsistent. We enforce consistency with the South African Standardised Industry Database by employing the biproportionality method (see Miller and Blair, 1985: 276-294) to a matrix consisting of size class dimensions per sub-industry for each year in two rounds.
Starting with the variables in constant 1995 prices, we let the sub-industry totals add up to the relevant counterparts of the South African Standardised Industry Database, while maintaining as much as possible Ntsika's proportions across sub-industries and across size classes. We then apply the South African Standardised Industry Database deflators to reconstruct values at current prices, followed by another round of the biproportionality method. The end result is a set of value added and wages and salaries data points for the four selected years in current and constant prices (see Table 1).