Page 43 - 2000 (5) Agricultur Census 2000 Balochitan Province

which the sum total of the
sampling errors plus all other
errors is very low. In a census of
agriculture, one may visualize
that the chances of occurrence
of non-sampling errors begin
with planning of the census and
end with the publication of data.
In other words, there are
almost endless chances of the
occurrence of non-sampling
errors. lt is due to this reason
that in a census, it pays to give
more attention to the control of
non-sampling errors than to the
reduction of sampling errors,
and quite often some increase
in the latter is to be tolerated
for the sake of reducing the
former. But, it is very difficult
as well as expensive to
measure non-sampling errors in
quantitative terms and specially
to obtain a comprehensive
measure of all the errors that
might have crept into the census
data at various stages. Only a
very elaborate checking and
evaluation of the census
operation can yield some usable
measure of non-sampling
errors, but most countries find it
difficult to spare manpower and
funds for this type of evaluation
after having spent huEe sums of
rnoney on the census itself.
In Agricultural Census 2000 the
best that could be done was
to contro! and rninimize non-
sampling errcrs through careful
planning, elaborate training
and intensive supervision.
Comprehensive editing of
census data for consistency and
relational credibility was another
measure adopted to reduce such
errors.
(3) Consistency and Reasonable-
ness of Data
. lt is evident from the foregoing
paragraphs that, under the
circumstances, it is not possible
to develop a comprehensive
measure to adjudge the quality
of the census data. The
sampling error is one indicator
with limited utility while the non-
sampling errcrs are of multiple
nature and very hard to estimate.
Thus, it becomes necessary to
consider other indicators for
judging the quality of data.
An important indicator in this
connection is the internal
consistency and reasonableness
of the data. This is an extremely
useful measure and it requires
very little extra census data for
its application. The data are said
to be consistent when the values
obtained in related but separate
items bear a logical relationship
with one another. For instance,
one may compare the cropped
area with cultivated area or the
existence of a tubewell on the
farm with area irrigated by
tubewell and so on. In a census
of agriculture, there are a
XXXIX