The name is absent

245

It is difficult to recognize any discernable patterns between the maize production data series.
Comparisons between the MAFF and the CSO data range from 75 to 108 percent. Comparisons
between the two CSO series vary from 88 to 121 percent. The MAFF estimates for maize production
in 1989 and 1990 are substantially greater than preceding or following years.

As with maize production, comparisons between the series show little similarity for the maize
sold estimates. It appears that the MAFF series of maize sales and the CSO series listed in its
"1992/1993 Final Crop Forecast" are the official estimates of sales.

Beginning with the 1993/94 agricultural season, the separate crop forecasting surveys of the
CSO and MAFF were merged. The survey method is that used by the CSO as outlined in the previous
section.

D. Publishing agricultural estimates

Statistical organizations have a choice of either publishing data as collected or using them as
indicators in an estimation process. Governmental organizations rely on survey data as indicators for
setting their published estimates of agricultural production. This method allows the survey statisticians
to adjust for anomalies in the survey data. The result is a smoother set of survey estimates which show
reasonable (believable) changes over time. If the organization publishes "raw" survey data (or data
with little review) the users are forced to make their own interpretation of how questionable data are
to be used. For instance, most agriculturalists would find it difficult to imagine a set of circumstances
that would result in the numbers of commercial farms (or the average area planted to maize)
fluctuating as much as is indicated by the CSO data. The graphs in the appendix illustrate the dramatic
amount of variation in the commercial farms data series. Normally averages for a group of farms
change slowly over time unless there are severe political, economic, or climatic events.

Agricultural statisticians agonize over how much they should adjust survey data. On one hand,
they do not want to be accused of "manipulating" data, but they are also reluctant to publish data
series that show inconsistencies. The CSO has had difficulty in choosing just how much intervention
they wish to impose in reporting their survey results. It appears they should have been more willing
to control for the variability caused by both sampling and non-sampling errors in the commercial
farms surveys. If the CSO statisticians had written technical notes about problems and the procedures
they followed in correcting them, it would have provided both a consistent data series for the many
users who just "need a number" and a basis for further study for the in-depth analysts. However,
simply expending more time writing technical notes will prove ineffective in correcting the problem
without a commensurate improvement in salary structures that provide office staff sufficient incentive
to scrutinize data and make necessary corrections.

After reviewing data series such as those published by the CSO and the MAFF, external
reviewers often face the dilemma of deciding whether to state that the data series are too fraught with
errors to be usable or to be more understanding and merely warn data users that the series contain
many unexplained inconsistencies. The first alternative is seldom taken because data are so badly
needed and are deemed to be "the best there are." Fortunately, an understanding of statistical
organization and training exists in Zambia that can be used as a basis for reviving the system if the
government decides to move toward fewer, more accurate statistics.

<<   250   251   252   253   254   255   256   >>

More intriguing information