Previous Page Table of Contents Next Page


By G. Groenewold and U. Tietze

Macro-level secondary data analysis

The macro-level secondary data analysis consisted of a comparison between the geographic distributions of fisherfolk at the time of the last three population or fisheries censuses to assess the change that had occurred in (a) the geographic distribution of fisherfolk and the number and type of fishing craft and gear, (b) the geographic distribution and fish catches/production, (c) the level of exploitation of aquatic resources in terms of “under exploited”, “fully exploited” and “over exploited” and (d) the state of the coastal environment, including the level of water pollution, destruction of mangroves, etc. Regarding (c), reference was made to inshore and offshore as well as pelagic and demersal resources. It was assumed that the last three population or fisheries censuses would span a period of 20–30 years, depending on the year when the last census was carried out.

Based on the compilation and analysis of secondary data available from population or fisheries censuses and from key informants and experts, (e.g. information on the level of the exploitation of aquatic resources and on the state of the coastal environment), a comparison was made with reference to particular geographical areas within the country: whether and to what extent increases/decreases in the number of fisherfolk populations had affected (a), (b), (c) and (d) above.

Regarding the geographic areas to be distinguished, there was an attempt to consolidate data at the largest administrative unit level, i.e. the coastal State in India and Malaysia, the region in the Philippines, Tanzania and Senegal and the district in Bangladesh. In a particular state, region or district where the focus group discussions and the representative household survey were carried out later, it was attempted further disaggregate data at the next lower administrative level: the district in India, the province in the Philippines, etc.

In addition to the change in the number of fisherfolk, change in socioeconomic, health and demographic characteristics was described and documented. This included the following:

1 Infant mortality refers to deaths occuring between ages 0 to 12 months. Child mortality refers to those occuring between ages 0 to 5 years exactly (0–60 months).

The sources of data for socioeconomic, health and demographic characteristics and their development over the last three population censuses were: population and housing censuses, labour force and employment surveys, world fertility survey (WFS), demographic and health surveys (DSH) and contraceptive prevalence surveys (CPS).

With regard to the socioeconomic, health and demographic characteristics, there was an attempt to extract data for fisherfolk from the above-mentioned surveys in relation to the overall population. This was to be done at the same level as the compilation and comparison of the number of fisherfolk, e.g. at the state, regional and district level. If this was not feasible, it was to be done at the national level.

The overall objective of the macro-level secondary data analysis was to describe and document the extent and nature of the demographic, health and related socioeconomic transformation over the last 20–30 years in coastal fishing communities. This was to be related to the variation in the number and type of fishing technology, fish production, the level of exploitation of marine aquatic resources and the state of the coastal environment.

Focus group discussions

The focus group discussions were to compare the demographic, socioeconomic, health, production and environmental data and information compiled in the course of the macro-level secondary data analysis with the actual awareness, knowledge and perceptions of a selected fishing community. Subjective explanations and perceptions of objective trends and developments were thus provided by the groups, revealing what fisherfolk and non-fisherfolk perceived to be the demographic socioeconomic and environmental changes that had occurred in the past and were unfolding now.

In addition, the focus group discussions were to yield qualitative information on the attitudes of fisherfolk and non-fisherfolk towards human fertility, family size, family planning etc., in relation to the natural resource base. The discussions were also to examine concern for the level of exploitation of aquatic resources, the state of the coastal environment and other significant attitudes and perceptions. These include fisherfolk's and non-fisherfolk's views of developmental needs and constraints, personal ambitions and perspectives etc. (see Annex 4).

To understand whether and how occupation and environment influenced attitudes and views of fisherfolk, it was decided that a comparison would be needed with non-fisherfolk, that is, nearby farmers. Therefore, two villages were selected in a district where, later, a household survey was to be carried out.2 One of the villages was a fisherfolk village, where the majority or all the households earn the major part or all income from fishing. The other village was inhabited by predominantly farmers and located near the fishing village. As the prime interest of the project is to learn about the lives of fisherfolk the number of topics covered in focus group meetings of fisherfolk was more comprehensive than in focus group meetings of farmers (see also: Annex 4)

2 Throughout this document, the terms “village” and “community” will be used interchangeably, meaning the human settlements where the research took place. Therefore no reference is intended to e.g. “village” as an administrative category, where applicable.

The following subjects were dealt with in the focus group discussions:

Household survey

The household survey was the main field research activity carried out by the Project. It aimed to explore, describe and analyse demographic characteristics and cognitive patterns of marine fisherfolk with regard to human fertility rates, reproductive behaviour and other demographically relevant factors. The survey also aimed to further the understanding of the cognitive patterns and demographic characteristics of the natural environment in which fisherfolk live and make a living.

In order to support country teams with the design and implementation of household surveys, FAO produced and distributed a number of documents, as well as a software application to the staff of the country teams, including:

A set of household questionnaires was prepared by FAO and field-tested in the study areas by the country teams. A set consisted of one household level questionnaire and a number of individual questionnaires. The household questionnaire, answered by the head of the household, collected information on household composition and certain characteristics of household members, housing and living conditions, income, savings and debt, assets and comprehension of the local environment. The individual questionnaire, answered by each eligible household member, collected information on socio-demographic characteristics, production, reproduction and understanding of the local ecosystem. In general, the eligibility of household members for an individual interview was based on age and marital status selection criteria.

The survey questionnaires adopted the local definition of the household concept. A household was defined as a socioeconomic unit, irrespective of whether the household members were related by blood or marriage. Individuals belonged to a household if they were thus perceived by the head of the household, and if they formed a socioeconomic unit in terms of housing, generation and use of income and consumption. Despite the fact that the concept “head of household” has become controversial, the household survey maintained the use of that concept. It was accepted as such in the rural communities covered by the Project.

3 Based on the ISSA software (Integrated System for Survey Analysis) of the Institute of Resource Development at Macro International Inc., USA.

According to the guidelines, the household surveys in each country were to be carried out in the coastal fishing communities of a particular district or province where coastal fishing is considered to be an important livelihood. Where possible, probability sample designs were to be developed and implemented, making survey results statistically representative for all of the fishing communities in that district or province. It was proposed4 to sample at least five fishing communities for a total of 100 households. Moreover, five farming communities and 100 households were to be sampled in the vicinity of the sampled fishing communities. Thus, in each country, the household survey collected information on about 100 fisherfolk households and about 100 farming households.

In two of the six countries studied, however, namely Bangladesh and Malaysia, sampling of fishing communities could not be carried out. Probability sampling of fishing communities proved problematic because there were only a limited number of agricultural communities located near fishing communities in Bangladesh. In the case of Malaysia, the advice of concerned Government authorities on the selection of fishing communities was sought.

Selection of sample

The Philippines

The household survey was carried out in five fishing villages (i.e. barangays) that were sampled from a sampling frame consisting of 22 fishing villages. All the villages belong to the municipality of Miag-ao, situated on the island province of the Visayas. Near each of the sampled fishing villages, a farming village was identified. All the selected villages were screened to determine the number, type and location of the households. In fishing villages, only fisherfolk households were identified, selected and interviewed. In farming villages, only farming households were identified, selected and interviewed. In each selected village 20 households were sampled. Fishing villages varied in size from 93 to 324 fisherfolk households. As the number of households sampled in each village was not in proportion to the population size of the village the computation of sample design weights was required and to be used in data analyses. Survey results from fisherfolk households are representative for fisherfolk households in all fishing villages of the municipality of Miag-ao. Survey results of farming households in farming communities are only representative for the particular village in which farming households are sampled and interviewed.


The household survey was carried out in five fishing and five farming villages in the northern part of the state of Kedah, near the Thai border on the West Coast of Peninsular Malaysia. The villages were selected according to the perceived responsiveness of villagers and the advice of the Fisheries Development Authority of Malaysia (LKIM) and the Agriculture Development Authority of Malaysia (MADA). Within each village, 20 households were selected. Enumerators were asked to pick a household in the village at random and interview it. In the case of absence or non-availability, a replacement household was identified and interviewed. As in Bangladesh, in fishing villages only fisherfolk households were considered eligible for interviewing and in farming villages, farming households only. Whereas in other countries all (married) women from 15 to 60 years and men from 15 to 64 were eligible for interviewing, because of budgetary constraints, only the head of household, his/her spouse, eldest son and/or daughter were considered eligible. Eventually, a total of 201 households were identified and interviewed. The survey results of sampled households were only representative for the particular village and for its household sub-population. The way villages and households were selected make that survey results cannot be generalized beyond the level of interviewed households.

4 See Annex 3: Sampling guidelines.


The household survey was carried out in villages selected in three districts along the south and southeastern coast of Bangladesh: Laxmipur, Chittagong and Noakhali. Poor road and transport infrastructure, as well as the difficult-to-meet condition set by the Project that the chosen agricultural village must be close to the chosen fishing village, were some of the main reasons why a probability sample of villages and households was not accomplished. Therefore, five fishing villages and five agricultural villages were selected by judgment. In each village, 20 households were selected and interviewed. In fishing villages, only fisherfolk households were listed, selected and interviewed. In agricultural villages, only farming households were listed, selected and interviewed. The selected villages appear to vary in size from 37 to 350 households. A total of 200 households were selected and all were successfully interviewed. The purposive selection of villages makes that findings from the household survey are only representative at the level of a selected village and the sub-population of households within that village.


The household survey was carried out in coastal villages in the Thane district, north of Bombay in the State of Maharashtra. It is estimated that about two million people live in the 1 682 villages of the district. In 82 of the villages, fisherfolk households were the predominant inhabitants. The idea was to develop a sample design that would generate representative survey data for all the fishing villages in Thane district. More specifically, from the list of 82 fishing villages, five fishing villages were randomly selected. Five farming villages were identified as the nearest to each of the five sampled fishing villages. A fixed number of 20 households were sampled within each of the villages (after the total number of households and their locations were determined). Because the population sizes of the villages varied, i.e. from 200 to 3 000 households, sample design weights were computed and applied in the analysis. The weights compensate for the fact that, due to the method used to sample unequal sized villages, data taken from households in small villages are over-represented in the database, whereas data from households in large villages are under-represented. Survey results are representative for all fishing villages in Thane district.


The household survey was carried out in the southern part of Bagamoyo district in the coastal region, north of Dar-Es-Salaam. Adequate and appropriate sampling frames were lacking on district and village levels. Moreover, road and transport infrastructure was so poor that the household survey was confined to one sub-district in the south of Bagamoyo district.

The sub-district consisted of six fishing and five farming villages. The goal was to develop a sample design that would make survey data from fishing villages representative for all the households in fishing villages in the sub-district. Two fishing villages were randomly sampled from the six villages in the sub-district. Farming villages were selected on the basis of how closely they were located to the sampled fishing villages. Independent from the aforementioned sampling, a third fishing village was sampled to pre-test the questionnaire. In the analysis, interviews obtained from the test population in this village were added to those of the other four villages, introducing sampling bias. The objective was to sample and interview a fixed percentage (10%) of households in each of the five villages, making the design a self-weighting one. A screening operation in the selected villages was carried out to determine the location and number of households. From the 2 423 households counted in the five villages, 242 households (10%) were sampled. Eventually, the eligible household members in 225 households were successfully contacted and interviewed, resulting in an enumeration response rate of 93 percent. Merging data from the test population with data of the sampled population resulted in survey results that are, strictly speaking, not representative for the sub-district in Bagamoyo district.


The household survey was carried out in communities located in three regions north of Dakar: Louga, Thiès and Saint-Louis. A probability sample design was developed to generate representative data for fishing and farming communities in the three regions. The recommended sample size of 200 households was increased to 500 households to enable, statistically speaking, more meaningful analysis and generalizations.

All the communities in the three regions were classified and grouped into the two strata of fishing and farming communities. The fishing communities consisted of 20 villages and 61 “district de recensement” (DRs), or census blocks; the farming communities consisted of 25 villages and 72 DRs. For the purpose of census taking, communities were subdivided into DRs (various sized spatial units consisting of the number of households that can be handled by a census interview team on census day). Communities in each stratum were geographically sorted by their DRs. The 1988 census reported on the population size by DR, making it possible to create a cumulative population size list for the communities in each stratum. Subsequently, using the systematic selection method, a total of 20 DRs (ten in each stratum) were sampled.

After the 20 DRs were sampled, a quick screening operation took place in each DR to determine its current (1996) population. The population of the DRs appeared to have changed dramatically from 1988 to 1996. Therefore, raising factors or weights were computed with the help of the 1996 population count of the sampled DRs. Application of such weights in analyses ensure that overall selection probabilities of households at the time of the survey are the same, irrespective of the size of their DR. The 20 (10+10) sampled DRs were located in five fishing villages and five farming villages. Then, in each DR, a fixed number of 25 households was sampled, bringing total sample size to 20×25=500 households. Eventually, 449 households were successfully located and interviewed (231 in fishing villages and 218 in farming villages), yielding an overall enumeration response rate of about 90 percent. Survey results are representative for the fishing communities in the three coastal regions of Louga, Thiès and Saint-Louis.

It follows from the above that analyses of survey data across countries are affected by the following methodological differences:

Previous Page Top of Page Next Page