PROVIDE Project Technical Paper 2005:1
higher than those reported in the IES. For certain occupation groups, such as clerks, plant and
machine operators, elementary occupations, and unspecified occupations, the average LFS
wage is more than twice that of the IES (also see Table 1).
February 2005
Closer inspection revealed that large outliers in the LFS data are often the cause of the
large income differences. This also explains the large difference between the overall IES and
LFS labour income (R32,405 compared to R53,091). In at least 8 observations the LFS figure
was exactly 1000 times higher than the IES figure, which clearly points to data capturing
errors. In many other instances the figure from the one survey was exactly 10, 12 or 100
times the figure from the other survey. This discovery necessitated looking at records with
large differences individually. A new person-level labour income variable, inclabp_new,
which essentially selects the ‘more appropriate’ of the two reported labour income figures,
was created. Section 7.1 in the appendix explains how this was done. This variable was later
renamed inclabp_old when inclabp_new was scaled to match the household-level inclab
variable. Section 4.2.4 explains how the scaling was done to create inclabp_new. The
discussion of inclabp_old and inclabp_new and Figure 1 is also continued in section 4.2.4.
Figure 1: Comparing LFS 2000:2 and IES 2000 average wages and employment figures
I I inclabplfsorig
I I inclabpold
—⅛— Weighted obs (LFS orig.)
—Ж— Weighted obs. (OLD)
I I inclabpiesorig
I I inclabpnew
—X— Weighted obs. (IES orig.)
—∙— Weighted obs. (NEW)
12
© PROVIDE Project