South Africa’s transition to democracy in 1994 engendered expectations of rapid changes in racial and
gender disparities in wages. Twenty years after the transition there are strong disagreements about
whether there have been gains and how these have been distributed. In particular there are debates about
whether wage increases have outstripped productivity increases.
Another legacy of apartheid which makes these debates more difficult is that there is little pre-1994
survey evidence that can be used as effective benchmarks for assessing progress. We show in some
detail that since 1994 there have been notable changes in the quality of the data and that dealing
properly with changes in sampling design, survey instruments and missing values is necessary in order
to identify trends.
We show that real wages have increased over the period – however real median wages of employees
have remained stagnant. The top tail of the earnings distribution has “fanned out”, with larger gains at
the 90th percentile than at the 75th, but both showing good real earnings growth. At the bottom end of
the distribution there seems to have been compression – with the 10th percentile making real gains
relative to the median. Among the self-employed there is no evidence for systematic shifts in the
distribution over the post-apartheid period.
The employment series is bedevilled by a huge break between October 1999 and February 2000, which
coincides with the shift from the October Household Surveys (OHSs) to the Labour Force Surveys
(LFSs). The LFSs find more marginal and informal activity than the OHSs did. However in the
immediate 2000 period they seem to find an excess number of subsistence agricultural workers (around
800 000) which disappear in later surveys.
The much higher level of employment is confirmed in the total hours worked series. Many of the
additional workers enumerated in the LFSs (particularly in subsistence agriculture) work lower hours
than the norm.
The mean wage reported in the Quarterly Employment Statistics (QES) is almost double the mean wage
from the Quarterly Labour Force Survey (QLFS). We examine various hypotheses about why they are
so far apart. Differences in coverage, missing information, and various types of measurement error all
help to explain parts of the gap, but even assuming some quite implausibly large magnitudes we cannot
close the gap entirely. Undoubtedly this is a matter for further research. Given the discrepancies it is
clear that statistics from the QES and the QLFS shouldn’t be mixed and matched.
With these caveats we use Gross Value Added figures from the national accounts and total employment
figures which we cleaned of the excess subsistence workers to construct labour productivity series.
Looking at the sectoral figures there seems a clear pattern of divergence: sectors that were relatively
low productivity ones in 1994 (Agriculture, Construction, Trade and Services) also show much lower
labour productivity growth, while the remaining ones all seem to have achieved productivity gains. On
average there is little evidence that wages have outstripped productivity or vice versa.