EXECUTIVE SUMMARY

The existing situation regarding the operation of the MIS project in the TLMP was surveyed and reviewed. The MIS was in fact found to be two quite separate projects with similar sampling, data collection and data presentation techniques, but with quite different data input, data storage and data processing methods.

There was found to be a potential wealth of data available, scanning the period 1989 through 1994. This information however was fragmented and (as above) stored in 2 different systems. In relation to the information available, insufficient analysis and reporting had been done (at least as regards the reports made available to me in Moshi, at MDB and at TLMP).

The physical processes of sampling, initial data recording, and secondary data recording was observed at the 3 major markets plus 2 other markets. The various information presentation formats were reviewed.

At the markets visited, both buyers and sellers were interviewed for their opinion of the MIS. Most were indifferent and some were dismissive, stating that the MIS was inaccurate, not timely, and irrelevant.

The 2 computer systems were both examined with regards to data input, data processing, data storage and information presentation.

The fragmented data from both systems was collected together and translated into one single dbase-format database; this was then examined for data quality. The findings pinpointed weaknesses in sampling - the systems were attempting to produce accurate numbers for 15 Grade-Sex Combinations (GSC's) with net sample sizes which were large in terms of the data collection work involved, but which were small in relation to the need without any proper data smoothing technique; to complicate matters, the systems were using population-representative sampling techniques, with a resulting lack of data for many of the 15 GSC's on most market-days.

In order to get more accurate pricing, and to open the possibility for smaller net sample sizes, I recommended that sampling was changed from a population-representative basis to a sex-weight-grade representative basis; the downside of this of course is losing possibly important information on population trends from this point onwards. I feel that the wealth of information unearthed in the 5 year analysis conducted in this report may compensate for this impending loss of population trend information (or may enable specialised focused searches to continue).

This composited data was also examined for seasonal fluctuations and for time trends during the period 1989-1994 in animal weight, animal quality, grade, price, unit price, sex, age, origin, buyer, yardage and number sold. These trends were graphed using Harvard Graphics, and the graphs are presented with this Report. The trends should be of significant use for the overall TLMP Project Team and for any other relevant parties. Of particular interest was the fact that certain producers and traders supplying the Arusha and Moshi markets have over the period 1993 - 1994 greatly increased animal quality in response to market forces and price incentive.

The composited data was also examined to determine whether there were useable practical relationships between price, weight, sex and grade on individual market-days (i.e. a potential for data smoothing). There was found to be a strong exponential relationship between price and edible weight for each sex on each market-day, regardless of grade; this strong relationship was suspected to extend across the sexes to all data points, but the sampling was such that this was difficult to confirm.

From this exponential relationship, a methodology (log-log best-fit - LLBF) was developed for data processing, which effects data smoothing in a pragmatic way, without making any assumptions that buying or selling behaviour on the current market-day is necessarily influenced by the behaviour last week or last month; this LLBF technique takes into account the non-linear price vs weight relationships, and puts emphasis on live and edible weights (and thus animal quality), with grade taking a smaller (but still important) role than it has up till now.

Because of the non-linear price vs. weight relationships which were observed to occur in practice, I recommended that the presentation format be changed radically from a price per kilo basis to a price per benchmark animal basis; 21 benchmark sex-weight-grade combinations were proposed, based on average values for the 3 markets for the period 1989-1994. I estimate that this type of presentation should also be more understandable and more accepted by our major target group - the buyers and sellers at the 3 major markets.

In addition to the work as above on sampling, data processing and information presentation, it was found that the existing physical systems had weaknesses in initial and secondary data recording, and that both computer systems had weaknesses in data input and in data storage.

Practical and detailed recommendations were made to improve initial and secondary data recording; a computer program was written in Clipper (a 4GL programming language) to improve data input, data storage, data processing (using the LLBF Technique), and data presentation (using the Benchmark Animal Information Presentation Technique).

Recommendations were also made as to which computer hardware and software should be procured for the Project.

Seminars on the findings were conducted for the TLMP staff and training of the Dar TLMP MIS Staff was effected.