Back to Plea Homepage
Back to African Studies
Center Homepage
This data set resulted from a survey done as a collaborative research effort between the Tanzania Fisheries Research Institute and Michigan State University under the auspices of the Programme on the Lakes of East Africa. This inquiry was done in two phases: from January to July of 1993 on six fishing beaches and from June to November of 1994 on an additional five beaches. The six beaches surveyed in 1993 were: Mbarika in Kwimba district; Kayenze and Kageye in Magu District; Guta in Bunda district; Esegere in Tarime District; and Kakobe in Sengerema District. The 1994 beaches were: Kigona in Bukoba District, Busisi in Sengerema District, Mchangani / Mikuyuni in Geita District, and Rubiri in Muleba District.
Formal surveys of boat owners, management and fishing crews, riparian households, and fish processors and traders were conducted. In addition focus groups and in-depth interviews were held with the same populations during the approximately two weeks that the research team stayed on each beach. The subjects covered by the survey included the ownership and operation of fishing boats and gear, fish marketing, the basic socioeconomic situation of the respondents, household nutrition, issues pertaining to women and family relations, and opinions about the fishery.
A codebook has been written to make the data set accessible to any interested person with a basic knowledge of statistical analysis. This first section is a general description of the data. Section two is an abridged translation of the survey instrument on which the names of the variables have been superimposed. This is to allow users to locate particular variables within the context of the relevant question. The third section is the actual survey instrument used. The fourth section is a set of tables describing some selected variables and their associations. Finally, a listing of selected variables with their descriptive labels and frequencies is provided.
The beaches where interviews were conducted were selected randomly following two dimensions of stratification: size of beach and ecological zone. Our intent with selecting by size of beach was to differentiate between: (1) small, isolated, beaches where a few fishing boats landed; (2) larger beaches which, while still isolated, had 20 to 30 fishing boats landing; and (3) the large beaches which were more integrated into the larger national and international fish market. The number of fishing boats, however, did not turn out to be a very good proxy for the differences in centrality that we were looking for. For example, Guta is a highly centralized beach from a marketing perspective, but we only found 25 boats there.
The other dimension was five lake ecological zones which represent basic limnological differences (Kudhongania and Cordone 1974).
Fisheries Department census data were used to randomly select 15 beaches, one of each size in each ecological zone. Data gathering on nine of these beaches has been finished; the other six will be done in the future. At times we would arrive at a beach and discover that it wasn't the size it had been selected to represent. In these cases we traveled clockwise along the lake shore to the first beach that was the right size.
The discrepancy between the eleven beaches surveyed and the fact that only nine of the fifteen randomly selected beaches were done came about in the following ways. First, Kayenze beach was not one of the randomly selected beaches. Kayenze is a large and important beach and when we selected the small, nearby beach of Kageye we decide that the best use of our resources would be adding Kayenze. We divided the household interviews and trader interviews between the two beaches which were very much a part of the same community. The second discrepancy arose from one of the beaches we selected randomly, Nkome beach in Geita District. Upon arrival we discovered that this beach was really three beaches at some distance from one another. The largest beach, which dealt mainly with Nile perch fishing, was Mchangani, while the other two were smaller and focused on dagaa fishing. Because of the problem of under-selection of dagaa boats, discussed below, we decided to treat Mchangani, and one of the smaller beaches, Mikuyuni, as a single beach for purposes of boat selection. For logistical reasons we did the household and trader interviews in Mchangani. Three trader interviews (type 14) were done in Nkome center.
When we arrived on a beach we did a census of the fishing boats. If there were less than 25 fishing boats we selected all of them for our sample, if there were more we selected 25. Boats that did not fish were included in the census but not the sample, as were boats that had been in a state of disrepair and not been used for more than a year.
The census form collected basic information about the boat such as the type, length, whether or not it was in working order, whether it was used for fishing or some other task, and the registration number. It also collected simple information about the crew: names, what they did on the boat, whether or not they were married, and where they lived. This information was then used for the random selection procedure.
Individuals were selected to be interviewed on the basis of their roles in the fishing industry. The management of the fishing boats was divided into four roles. The owner holds title to the boat. A renter is someone who pays a fee for control of the boat and gets all the income flow from ownership in return. A manager is someone who oversees the operation of the boat but stays on shore. An operator is an overseer who goes out fishing. Crew are all other fishers.
A management interview was conducted with the person who knew the most about the operation of the boat. Often two management interviews were done for one boat, mainly with both the owner and a renter. In a few cases the owner was not interviewed, but an attempt was made to interview the owner regardless of how much he or she knew about the boat operations. Two crew members from each chosen boat were selected randomly.
Trader interviews were conducted with a sample of traders who bought their fish on the beach. The selection mechanism for small-scale traders was to interview as many as possible during the hectic fish-buying time each morning. In cases where there were large scale traders linked to the international market an attempt was made to interview all of them.
Fishing household interviews were done with the spouses of a randomly selected sub-sample of the fishers working on the selected boats. After beach 3 the number of fishing household interviews conducted was either one half the number of selected boats or a minimum of ten. It was not until we arrived on the fourth beach that we began to attach fishing households to particular boats. In Mbarika, Kageye, and Kayenze we selected from households identified as being fishing households. At this point we also decided that we would ask household data (Sections VIII and IX) of both spouses when they were both interviewed, previously we had only gathered household data from one of them.
Non-fishing household interviews were conducted with a random sample of households within one half kilometer of the beach that did not have a member of their household involved in fishing. The same number of non-fishing households were interviewed as fishing households. All household interviews were conducted with the female head of household where there was one. Households were defined as either a women or a man living alone in a compound (homestead) or as a women and her children. Thus, if there were two wives in one compound they were counted as two households. However, when households appear in the survey in such questions as "number of students in your household" these households were unavoidably defined as who the respondent see his or herself as living with in one household.
Three other types of interviews arose as the research went on. In Mbarika there were boats that were owned by the village. We interviewed the chairperson of the village committee in charge of each of these boats. These interviews were assigned a unique number because while the were boat owner interviews they could not be considered interviews of fishers. In Esegere there were a group of women who were hired by boat owners to sundry their fish. This is in contrast to the more common practice of fishers sundrying their own dagaa. We designated these women "processing crew" and interviewed them using some questions specifically for them along with sections of the main interview schedule. Starting in Esegere we also decided to begin to interview boat owners who had left fishing recently. We wanted to understand who was leaving fishing and why. This proved a very difficult population to sample, we did not do very many of these interviews and the interview type was not continued in 1994.
Interview schedules were made up of sections which were administered to different types of respondents in different combinations. A total of 691 interviews were done. Respondents included 397 fishers from 167 boats, 95 fish processors and/or traders, 74 fisher's wives, and 104 women from non-fishing households.
SECTION
|
TOPIC
|
POPULATION
|
The owner interview contained Sections I - IX, including Section III which deals with ownership of boats and gear. Many owner interviews dealt with more than one boat selected for the sample and in these cases multiple Section IIs were administered. Other management interviews consisted of Sections III - IX. Management interviews with those responsible for the operation of more than one sample boat were given multiple Section IVs.
Crew interviews consisted of Section X and Sections VI to IX. These sections deal with the particulars of the crew member role as well as information on other fishery involvement, including much the same information as is covered by Section III. The basic household interview was Sections VI- IX, excluding certain parts of Section VI which were only relevant to fishers. Processor and traders were given Section XII, dealing with the details of their operations, in addition to Sections VI-IX. They were also given a shorter Section VI. Women, whether responding to a boat owner's, trader's or a household interview, were given section XI.
Survey research in rural Africa presents many difficulties. The questions that one must ask in order to put responses into computers do not reflect the way people usually think about things. This can cause confusion and difficulties in communication between respondents and interviewers. These problems are compounded when the survey instruments are extensive - as was the case here.
Counting things, which is, after all, what quantitative research is all about, is a crucial complement to qualitative work. While a depth hearing of how affected people experience the issues should be the center of any social science research that is focused at the level of social action, if such a hearing is not complemented by quantitative work, the researcher is unable to get a handle on the magnitude and variation of the phenomena the qualitative investigation is uncovering. Furthermore, the laws of statistics, especially as they can be applied to substantial samples, are powerful tools for making tractable the difficulties of measurement reliability.
We felt that training was a crucial objective of our research effort. Thus, the entire gamut of technologies involved with this survey, from question formation through data entry were participated in by people who were new to the processes involved. However, supervision was as careful as could be. Whenever possible, which was at least 95% of the time, survey instruments were checked by the team leaders when the respondent was still available to clarify discrepancies. More than half of the data entry was done twice by two different people to minimize errors.
Our respondents were, for the most part, very cooperative; we experienced only two outright refusals to continue with an interview. A remarkable response considering that these interviews lasted for at least 45 minutes and some of them, where the respondent owned and operated several fishing boats, were as long as three hours. The vast majority of respondents, reflecting the hospitable nature of Tanzanians, welcomed us, wished us well, showed great patience with us, and were very pleased that outsiders were taking an interest in their fish related activities. However, neither the respondents nor the interviewers had very much previous experience with this type of questioning. The survey required respondents to think about their activities in a ways which is very different from the ways they normally would do so.
Unless the problems with the data gathering process are explicated there is a real danger that the numbers will introduce a false precision to the analysis. In this section those areas where we experienced the most difficulty are discussed. A potential user of this data should understand, however, that problems with reliability extend to every question. In spite of willing cooperation and close supervision there is no doubt that many questions were asked and answers given with less than complete mutual understanding. The appropriate general rule should be that while formal surveys, and the counting power they provide, are a crucial part of the research effort they should never be allowed to stand alone without qualitative work to clarify interpretation and support conclusions. None of the data supplied here, or any other survey data for that matter, should be used without being supplemented by a solid theoretical and experiential understanding of the issues and the area concerned.
The type of interview where we had the most problems with sampling was the traders and processors. It was impossible to do an initial census and then select respondents from this population because it changed every day. The nature of their work was such that they would be waiting for fishing boats one minute and then suddenly have to rush to buy fish because competition with other traders was often intense. What we ended up doing amounted to interviewing those traders who were available. We began doing the standard version of the trader section on beach 2. Our goal was to interview 10 traders on each beach, however on the smaller beaches, #s 2, 6, and 7, this did not prove possible. There were fewer traders, and those who did come where much more likely to be passing through quickly to see if there were fish that day rather than sitting and waiting for the boats to come.
There is a built in sample bias on the larger beaches. We made a point of interviewing those traders who were agents of the fish processing factories. In addition, traders who were involved in processing fish on the beach were more easily available for interviews than traders who left the beach with their fish after buying them. When using trader data, agents should be weighed at 1/3 and processors at 1/2. A third problem with the trader sample was that inclusion of secondary trading activities in the survey data was inconsistent. This was particularly important in regards to sales of swim bladders, which, particularly in 1994, was the most profitable part of many processors and traders businesses. By this time almost all of them were engaged in selling the swim bladders, but, because in occupied a relatively small amount of their time, they did not always mention it as a separate activity in the survey.
There are many variables which involved the respondents estimating quantities. These estimates should be treated as having high error terms. Certainly, biological data about fish catches should not be inferred from any estimates fish quantities. They should be used for comparisons only. Additionally, errors in estimates of fish quantities will be autocorrelated. This is because fishers on larger beaches who sell their fish to larger traders and agents are much more likely to have used a scale to weigh their catch when they sold it.
Several questions ask respondents to make comparisons with their current situation and that of a specific time in the past such as "in the past year" or "compared with 5 years ago". The actual times in these questions should be taken as representing no more than the respondent's description of secular trends that he or she perceives.
In addition, the basic idea of comparisons does not translate easily, the Swahili words for "more" and for "many" as well as that for "little" or "less" are the same and depend on the context to get the differences across. Particularly in comparisons of items or situations which don't change a great deal, and are either very uncommon or very common both currently and in the past, answers such as "very little" or "much more" may actually mean "the same".
An example to clarify: one of the questions in Section IX on nutrition asks if R eats more, the same, or less chicken than five years ago. The vast majority of these respondents eat chicken only on rare occasions. But many of them would respond "much less" when "the same" would be more appropriate. On the other hand, when asked about foods such as maize or fish which form an important part of the diet and the availability of which does vary a great deal, answers such as "more" or "the same" are much more likely to mean just that.
A similar problem arises in questions which take the form "in the past week" or "in the past month". Regardless of the time frame in the question responses will tend to reflect the situation in the recent past to a greater extent than the less recent past. "In the past week" will often mean "yesterday" and in "past month" will become "past week". When probed, the respondent will then almost always respond by extrapolating the original answer. Thus the week will simply be seven yesterdays.
The most serious example of this problem is in the section where the respondents are asked about who their customers were in the past week. Most respondents would list only those to whom they sold the day before, or one or two other customers from the recent past. This remained true despite repeated attempts to push for more data. These answers will still give a good aggregate picture of who is buying fish on that beach, but one which will be somewhat less useful for getting a picture of who that particular fisher sells to.
A slight problem arose from selected boats leaving the beach before we had a chance to interview the fishers. This happened particularly in Kigona and to a lesser extent in Mchangani and Guta.
Our sampling strategy had an inherent bias against selecting dagaa fishers. Nile perch fishers must sell their catches every day, and so they will tend to appear everyday at recognized landing beaches. Dagaa fishers, on the other hand, will often sell their sun-dried product at the end of the "dark", the 16-21 day period in which the moon is not too bright to fish. This means that many more dagaa fishers than Nile perch fishers will camp in isolated areas by themselves or with a small number of colleges. Our strategy of basing ourselves on known landing beaches, therefore, systematically under- counted dagaa fishers.
RETURN TO TOP
This data set is available for scholarly use. Copies of the
data set can be obtained from Doug Wilson. The
data set will only be released after receipt of a statement declaring that
the data set will not be used in any way that threatens the confidentiality
of the respondents. The data set is stored in SPSS-PC+ but may be exportable
to a different format.
Any use of the data from the data set should be cited as follows:
Wilson, D.C. and M. Medard 1994 Programme on the Lakes of East Africa Socio- economic Data Set for Lake Victoria, Tanzania Tanzania Fisheries Research Institute, Mwanza Station / African Studies Center, Michigan State University
Acknowledgments: The investigators gratefully acknowledge the support of: the U.S. Department of Education's Fullbright-Hays Doctoral Dissertation Research Abroad Program; the John T. and Catherine D. MacArthur Foundation; the Michigan State University African Studies Center; the National Science Foundation Dissertation Improvement Grant #INT 9320235; the Tanzania Fisheries Research Institute; and the Social Science Research Council's International Predissertation Fellowship Program. Without these institutions this research would not have been possible. A great debt of thanks is owed to Joyce Frederick, Ramadhani Mhekela, and Bellarmine Zenge for perseverance in data gathering and Charles and Olivia Mkumbo for the same in data entry.
Tanzania Fisheries Research Institute
P.O. Box 475
Mwanza, Tanzania
Michigan State University
African Studies Center
100 Center for International Programs
Michigan State University
East Lansing MI 48824