PROVIDE Project Technical Paper 2005:1
individual reported zero income in the one survey and non-zero income in the other, the
non-zero entry was assumed to be the correct one.
February 2005
Next, variable diffp, defined as the percentage difference (in absolute terms)
between the LFS wage (inclabp_lfs) and the IES wage (inclabp_ies), was constructed.
If this difference was less than 30% the larger of the two income levels was chosen as
the correct factor income. Initially there were about 24,700 workers (unweighted) in the
LFS, and 23,221 in the IES. When considering all people reporting income from labour
in a combined IES-LFS survey, there were almost 27,000 workers. For the majority of
these (approximately 23,000 observations) the difference between IES and LFS wage
income was less than 30%.
The remainder were evaluated record by record, following the basic rule of thumb
that the larger of the two incomes is correct. However, in instances where it was
reasonable to believe that the larger entry was incorrect, the smaller entry was select.
For example, in many instances the one figure was exactly 10, 100, or 1000 times the
other, which clearly suggests an error in the data capturing process. In such instances
the more realistic figure was selected, given the average income of the factor group, the
education level of the respondent, the total household income or expenditure
(unadjusted), as well as the reported income levels of other members of the household.
In all instances where it was still unclear whether to use the IES or LFS data, the one
that would be more successful at closing the gap between total household income and
expenditure was selected. Of the 4000 observations examined there were 95 cases
where the LFS income was larger, but the IES was used, and 59 cases where the IES
was larger, but the LFS was used. For the rest the larger of the two seemed more
appropriate.
The new income variable created was initially saved as inclabp_new, but later
renamed inclabp_old when inclabp_new was scaled upwards to match the household-
level variable inclab. This is discussed in section 4.2.4.
65
© PROVIDE Project