242
checks, it is easy to understand why data inconsistencies are routinely seen in the data sets
(Zambia/CSO, noncommercial). 16
6. Summary
It is obvious that the CSO, as the premier data collection agency for Zambia, has devoted
much time and effort to the continuation of agricultural data series. It is unfortunate that more effort
was not spent in assuring the quality of the data. It appears the foundations of the survey organization
may be sound, but the current activities are sorely in need of reevaluation and redirection.
C. Comparison of MAFF and CSO estimates
One method of assessing data accuracy is to compare several independent data series.
Sometimes two data series will rise and fall in the same general pattern but have minor differences.
In these cases, a close comparison may reveal that one series uses a slightly different definition or may
contain a bias that accounts for the variations. However, at other times, when more serious differences
exist in one or both of the series, little comparability may be present. It is difficult to compare the
MAFF and the CSO data series because of differences in methodologies. While both the CSO and the
MAFF call their surveys "crop forecast" surveys or exercises, the MAFF series is derived from a
grassroots listing of data from all farms. The CSO uses a sample-survey approach. The intent of both
is to provide early-warning information during the growing season. In contrast, the CSO's commercial
and noncommercial surveys are meant to provide a more comprehensive view of agricultural
production without the time constraints of the forecast surveys.
The "final" estimates of the CSO crop forecast (CSO/CF), the MAFF crop forecast
(MAFF/CF), and the combined commercial and noncommercial agricultural surveys (CSO/AS) were
compared. While it would have been preferable to use basic estimates such as "numbers of farms" or
"total land in farms" for comparison, they were not available, so maize was chosen as the crop
estimate series which was likely to be most accurate. Data for "area planted," "production," and
"sales" were compared. Table 8.6 lists the data and figure 8.1 provides graphic comparisons of the
data sets.
Comparisons of maize area planted shows the MAFF/CF to be generally higher than the
CSO/CF. In the eight years of comparative data, only in 1984 and 1987 were the MAFF estimates
lower (90 and 92 percent respectively) than the CSO. There seems to be no real connection between
the CSO/AS data and the CSO/CF data. For the four years, comparisons range from 91 to 126
percent. Special note is made of the 1988-89 MAFF estimate of maize planted, which is substantially
higher than any other estimate in any series. The 1,021,000 hectare figure is about 40 percent greater
than the MAFF estimates for the years immediately preceding and following.
16 However, questionnaire size is not the sole or even the most important reason for these problems. According to CSO
management, the problems encountered during the summary phase stem mainly from the lack of an incentive structure that
encourages personnel to properly edit data before summaries are undertaken. An officer who goes to the field to collect data
is eligible for an allowance in addition to salary. No such incentive is given to an officer in Lusaka who is responsible for
scrutinizing data to detect and eliminate data errors. It is generally the case that data processing personnel are left alone to
resolve inconsistencies without proper guidance from the statisticians—the outcome being inadequate editing and short-cut
methods employed to speed the summarization process.