Creating a 2000 IES-LFS Database in Stata

PROVIDE Project Technical Paper 2005:1

Table 8 continued...

From Table 8 it is clear that the numbers of true missing values are quite low. The only
sections that contain more than ten missing values in certain variables are the two sections on
monthly housing costs. Changing missing values to zeroes is justifiable given the small
number of true missing values. It is better to rather lose information content of a few true
missing variables by changing it to zeroes than lose the entire observation due to the adding-
up restrictions in Stata. As explained previously missing values in expenditure categories
from domworkerh.dta and homegrownh.dta were also changed to zero given that these were
optional section in the questionnaire.

The only other anomaly in Table 8 is variable P2003Total. Answering this question was
not optional, and hence the large number of uncoded observations is strange. However, as
explained below, it was established that this was a result of a coding error. Since this is a sub-
total it could simply be recalculated.

Another concern relates to questions 3.1 and 3.2 in the housing section. The questionnaire
requests respondents to answer either the section labelled monthly housing cost if rented or
the section labelled monthly housing cost if owned. In 21 cases households answered both
sections. It does not seem highly unlikely that some households own property (excluding
holiday homes) as well as rent property. The list below shows the reported values for each of
the questions in these two sections. There appears to be no duplication (apart from record
number 26015 - levy is reported twice). The error is small enough and is unlikely to affect
results in any great deal, and is consequently ignored.

41

© PROVIDE Project

<<   43   44   45   46   47   48   49   >>

More intriguing information