PROVIDE Project Technical Paper 2005:1
large numbers of missing values. These are in actual fact not true missing values, but rather
values that are “uncoded”, i.e. values that should have been coded as zeroes but were coded
as missing because the specific section did not apply to the respondent.28 Missing expenditure
values only occurred when respondents indicated that they did not produce any goods for
home consumption, nor did they keep any livestock. Thus, these missing values should
rightfully have been coded as zeroes and are changed accordingly.
February 2005
Next, a series of checks are performed (part 4) to see whether there are any coding
inconsistencies in the data. The first type of inconsistency checked for is one where
respondents indicate that they have no livestock but nevertheless report expenses (this error is
later referred to as “miscoded” - see footnote 28 and section 3.2.5). The second inconsistency
checked for is double counting of expenses. For some inexplicable reason many entries are
clearly duplicate entries. All entries where the same expense appeared two or more times
under the same hhid and the same produce or livestock code were identified and duplicate
expenses were changed to zeroes. A sub-do-file replace.do was created to make the necessary
changes (see Figure 8).
Part 5 of homegrown.do calculates the implicit prices of produce and livestock sold by the
household. These prices are needed to value produce and livestock consumed by the
household. As mentioned previously the value of home produce consumed was never
captured on the expenditure side of the IES 2000. Where Hoogeveen and Ozler (2004) used
actual market prices for 2000 to value home consumption, the approach here makes use of
actual sales data to calculate implicit prices. It is believed that this represents a more accurate
valuation of produce and livestock in rural areas where general market prices seldom prevail
due to separation from formal markets. The database contains data for quantity of sales and
value of sales. The implicit price is the value divided by the quantity. Since the price is
calculated for each observation, there is considerable variation in the implicit price. After
closer inspection and correspondence with various specialists in the field it was decided to
use the median price of produce and livestock sold to value home consumption. The study on
home production for home consumption (PROVIDE, ____-b) gives a detailed analysis of the
process followed to calculate implicit prices.
As mentioned in footnote 25 sales figures of commercial farmers should not have been
included in the database. Thus, in part 6 of the do-file respondents that clearly had production
and sales levels that are only possible when operating at a commercial scale were identified.
The value of sales was changed to zero, but care was taken to keep home consumption figures
28 This anomaly is also found in various other parts of the IES 2000 database. See section 3.2.5 for a more
detailed discussion.
33
© PROVIDE Project