• Tidak ada hasil yang ditemukan

CHAPTER 3. METHODOLOGY

3.5 Sampling

3.5.1 Sampling techniques and procedure

Multi-stage sampling was employed to select the sample for this study. The first stage was purposive selection of Goromonzi and Murewa districts. This was done as fodder seed technologies were introduced and promoted in these areas. The sampling frame consisted of all the farm households in both districts. The sampling selection of sample farmer households was at 95.0 % Confidence Level. In stage two, wards within the districts were purposively selected with the assistance of ward-based extension staff, based on areas for forage seed production. A total of 10 wards were selected (5 from each district) for the study.

Number of households within the ward was based on ward household population according to ZimSTAT (2013). In the third stage, random sampling was then used to select households for the household survey.

In the study, the target population was the smallholder farmers in Goromonzi and Murewa districts. As the study could not cover all farmers in the districts, a sample was selected for in-depth study. The sampling technique used was probability sampling as each member of the population had an equal chance of being selected into the sample.

Levy and Lemeshow (2013) define a sample as a small part or sub-set of anything selected from a population and designed to show characteristics of all other elements in the population. A good sample has to meet certain predetermined criteria and in the study, a sample was drawn from smallholder farmers in Goromonzi and Murewa districts. For example, if a sample is drawn from smallholder farmers and the attribute to be measured is normally associated with commercial farmers, then accuracy is compromised.

However, sampling error can occur as a result of random fluctuations inherent in the sampling process. The

ultimate test of a sample design is its representativeness of the characteristics of the population it is supposed to represent (Singh and Masuku, 2014). The main idea behind sampling is to enable the drawing of conclusions on the entire population using selected elements of that population through in-depth studies of the sample.

The advantages of sampling are that it possesses the possibility of better viewing, more thorough investigation of missing, wrong or suspicious information, better supervision and better processing than is possible with complete coverage (Bryman and Bell, 2003). Quality of a study is often better with sampling than with dealing with a whole population such as in a population census, as a census more costs more.

Sampling also provides much quicker results than does a census and therefore, is time saving. It is the only possible process when dealing with an infinite population. As described by Bryman and Bell, (2003); Singh and Masuku, (2014), sampling techniques are divided into random (probability) and judgemental (non- probability) sampling.

Simple random sampling – is a form of probability sampling method which involves random selection of elements from a population and it ensures that the probability of each case being selected from the population is known and is usually equal for all cases. Bryman and Bell, (2003); Singh and Masuku, (2014) define simple random sampling as each element of the population having an equal chance of being selected into the sample drawn. In the case of this research study, every element of the sample, that is, farmers, has an equal chance of being selected. This is because the method is easy to implement, can be easily understood and used, and the samples allow one to project sample results to the entire population.

Stratified sampling–this method of probability sampling involves dividing the entire population of elements into sub populations, called strata, then selecting elements separately from each subpopulation (Levy and Lemeshow, 2013; Singh and Masuku, 2014). The researcher controls sample size in strata, increased statistical efficiency, provides data to represent and analyse subgroups, enables use of different methods in strata. Increased error will result if subgroups were selected at different rates, expensive. This method is very useful especially when the population is large and several characteristics occur within that population.

Cluster sampling - This is a probability design where a sample of clusters is first selected and decide on which sampling units to include in the sample further study (Singh and Masuku, 2014)). The method provides unbiased estimate of population parameters. It is also economically more efficient than simple random and easy to do without a population list.

Purposive sampling - Palinkas, Horwitz, Green, Wisdom, Duan and Hoagwood (2015) mention that the principle of selection in this non-probability sampling is that respondents are selected as it is expected that they are representative of the population of interest and meet the specific needs of the research study. There is low likelihood that the sample will be representative enough. The method is mainly used in qualitative research and when dealing with very small samples such as in case study research and when one wishes to select cases that are particularly informative (Neuman, 2000).

3.5.2 Sample size determination

Epi Info 7.2.1.0, is a software package that was used to calculate sample size. Sample size was determined at 95.0 % Confidence Level with a 5.0 % Margin of Error 5.0 %. A total of 414 farm households were selected as the sample and used in the study from the two districts. Sample size per ward ranged between 26 and 57, with an average of about 41 households per ward (Table 3.1).

Table 3.1: Distribution of sample households in Goromonzi and Murewa districts District Ward Identification

Number

Household Sample size

Percent frequency (%)

Goromonzi

2 39 9.4

4 54 13.1

5 26 6.3

11 57 13.8

12 54 13.0

Murewa

4 40 9.7

11 36 8.7

14 30 7.2

27 30 7.2

28 48 11.6

Total 414 100

A total of 4 FGDs were successfully conducted at 4 sites (2 in each of the districts) and participants comprised extension staff from LPD, AGRITEX and Veterinary Services Division, lead farmers, local agro dealers, NGO partners, and local leaders. Of the 55 participants to the FGDs, 56.4 % were females, whilst the balance were males. Table 3.2 shows the number of participants at each of the FGDs and where it was held in the respective districts. Attendance at FGDs, had more women than their male counterparts. LPD, AGRITEX and Veterinary Services Division are government departments responsible for the provision of extension services to farmers on livestock production, crops and animal health respectively. In every ward, each department is represented by an officer. NGOs also implemented agriculture related projects in the study area and they had a fair understanding of the communities and their livelihood activities. With four FGDs, the number was adequate and it enabled consolidated qualitative data and drew themes from it, making use of NVivo software package.

Table 3.2: Sites and participant composition for FGDs in Goromonzi and Murewa

District Site Males Females Total

Goromonzi Showgrounds 6 6 12

Chikwaka Milk Collection Centre 8 7 15

Murewa Muchinjike School 6 9 15

Mahohwa Business Centre 4 9 13

Total 24 31 55

For Key Informant Interviews, a total of five (representing 38.5 % of registered and participating companies) seed companies were selected using simple random sampling from the list that was availed by the Zimbabwe Seed Traders Association (ZSTA) where interviews were conducted. The companies have had years of operation in the seed industry ranging from 5 to 70. They also have a good national coverage, with two of the companies being represented internationally. Industry players (input suppliers and traders) who usually serve the farmers were also selected randomly to conduct interviews for an industry analysis.

These include locally-based agro dealers and those located in the city and they interact with farmers for agricultural inputs supply and other household requirements. They have an average branch network of 2 across the country and have been in operation for an average of 12 years. Relevant heads of government department (research institutes, regulatory services-Seed Services, Extension Services-AGRITEX and LPD Divisions) were also interviewed during the study. These are public institutions that depend solely on the fiscus for their operations and service delivery.