As for the pilot survey the success of the full-scale survey will depend on procedures, the questionnaires employed and, finally, coding and analysis. Recommendations for modifications in these three areas are presented below.
Selection of the sample, including female-headed households, and the execution of the field work, seems to be the most important aspect of procedures.
In order to ensure statistical relevance for the various subgroups in the sample, it would seem that the survey should not cover less than 200 respondents. If the entire population is above 1 000, the sample should be increased accordingly. The sample should be distributed approximately with 50% of respondents being practising farmers, 25% former and 25% potential.
There seems no reason to change the single stage random cluster sampling procedure.
How to ensure the inclusion of female-headed households? Amongst the first two categories, practising and former, the population will include a certain number of female-headed households. The random selection of the sample should ensure that an adequate representation is included. Once in the field it seems likely that the enumerators will encounter two types of female-headed households, which, in this context, will be referred to as “permanent” and “transitory”. In the first case the man is either dead, frequently or permanently absent, and residing far from the household. In this case, the woman should be interviewed. The second category, the “transitory” female-headed households are those where the man is not involved in the day-to-day running of the household but close enough to be able to exercise his authority at brief notice. In these instances the woman is in fact not the decision-maker and should therefore be interviewed about “Current Situation and Outlook”. The rest of the interview should be held with the man.
The survey team must have the time to track down “absent” farmers because a procedure which inteviews only those farmers who are immediately available on the farm is likely to introduce considerable bias in replies.
The selection of potential farmers for interview should also be randomized. One solution could be to systematically, in each village visited, interview one potential farmer for each three former and/or practising farmers, provided that potential candidates are identified by farmers and/or local government staff.
The pilot survey revealed only two informal groups as owners/managers of fish ponds. However, this need not mean that such groups seldom consider or attempt to start tilapia farming. Pond culture as a group activity will most likely be carried out on land not assigned to any individual household. In order to ascertain whether such attempts have been made, it might be necessary to contact village headmen or other authorities, to establish: (i) whether suitable land/water is available on grounds not allocated to any particular household and (ii) if any informal group (or individual) has expressed interest in using it for tilapia-farming in ponds. If that is the case, and only then, the enumerators should attempt to trace the individual /groups who have expressed interest in being interviewed. Consequently, an interview concerning the proposition: “if land/water were made available would you be interested in tilapia culture?” would be pointless unless such an interest had already been recorded.
The composition of the interview teams naturally influences the result. The nature of questions are such that most academically qualified individuals will be able to carry out the survey after having spent a day or two to become familiarized with the purpose of the survey, and having had the opportunity to discuss questions with project staff.
This is true for the Fish-Farmer part of the survey. The Current Situation and Outlook interview needs to be carried out by someone with fish-culture expertise.
Given that the staff of the Department of Fisheries have a vested interest in fish-farmers' success (and the farmers know this), it is preferable that the Fish-Farmer Survey not be carried out by DOF staff. If it is, results may be in favour of “positive” results.
Likewise, the presence of foreigners, “muzungus”, may lead farmers to believe that the results of interviews will in some way influence the selection of future aid recipients. Therefore the fewer non-locals participate in the interview teams the better.
It is useful to break the questionnaire now used in the pilot survey for the “Current Situation and Outlook” into four, as follows:
| - | practising farmer: | intermediary harvesting |
| - | practising farmer: | batch harvesting |
| - | former farmer: | intermediary harvesting |
| - | former farmer: | batch harvesting |
In the pilot survey, as discussed above, the attempt to diagnose the efficiency of the culture system relied upon establishing as accurately as possible the productivity (measured in terms of kg/ha/year) of the ponds. It is suggested that this be continued, but that the questionnaires addressed to those employing an intermediary harvesting strategy, be modified to reflect any information available on the production levels in such systems.
Also, the “productivity level” approach should be complemented by one focusing on the extent to which available feed and organic fertilizers are adequately used. If such information were obtained it would indicate in how far it would be at least technically feasible to produce more fish from the ponds in any given period.
The questionnaires used in the Fish-Farmer Survey may be improved by modifying them to accommodate information regarding:
details of batch and intermediary harvesting
origin of the pond
extent of multiple use of water in ponds
The above, together with some additional information on nature of feed and fertilizers used, and also on water renewal will allow for greater precision about (the unlikely) pollution effects than in the pilot survey.
8.3 CODING AND ANALYSIS
A few modifications are suggested with regard to coding and analysis. First, it would facilitate the analysis if information recorded under “other” was registered on diskettes. Second, the categories:
did not know
refused reply
was not asked
should be added to all questions. This need not take more than one line, or box, at the end of each question. The enumerator will enter a standard code (preferably numerical) in the box when applicable.
It will save considerable computer time for those carrying out the analysis if the computer has a hard disk of some 20 MB. The use of diskettes increases the search time considerably, particularly when a search is programmed and accesses more than one data-base in the same programme.
